JP3960932B2 - Digital signal encoding method, decoding method, encoding device, decoding device, digital signal encoding program, and decoding program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a encoding method for compressing the information of a digital signal without causing distortion and selecting a hierarchically configured sampling rate, a decoding method, an encoder, a decoder, and programs for them. <P>SOLUTION: A down sampling section 13 converts the sampling frequency of digital signal e.g., from 96 kHz to 48 kHz for each frame, the converted signal is compression-encoded and the result is outputted as the main code Im, an up-sampling section 16 converts the local signal corresponding to the main code Im into a signal of a sampling frequency of e.g., the original 96 kHz, an error signal between the converted signal and the received digital signal is produced, a rearrangement/encoding section 18 applies rearrangement to bits of a sample string of the error signal, and the result is outputted as an error code Pe. A decoder side obtains a high fidelity reproduction signal by using the main code Im and the error code Pe or obtains the reproduction signal by using only the main code Im. <P>COPYRIGHT: (C)2005,JPO&amp;NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an encoding method for converting a digital signal such as voice, music, or an image into a code compressed to a lower information amount, a decoding method thereof, an encoding device, a decoding device, and a program thereof.
[0002]
[Prior art]
There are irreversible encoding that allows distortion and reversible encoding that does not allow distortion as methods of compressing information such as sound and image. As irreversible compression, various methods are known as standards of ITU-T (International Telecommunications Union-Telecom Standardization) and ISO / IEC MPEG (International Organization for Standardization / International Electrotechnical Commission Moving Picture Experts Group). By using these lossy compression methods, it is possible to compress the original digital signal to 1/10 or less while suppressing a slight distortion. However, the distortion depends on the encoding condition and the input signal, and deterioration of the reproduction signal may be a problem depending on the application.
[0003]
On the other hand, as a reversible compression method capable of completely reproducing the original signal, universal compression coding often used for compression of computer files and texts is known. This can be compressed for any signal and can be compressed up to about 1/2 for text and the like, but the compression effect is only about 20% even when applied directly to audio or image data.
By combining lossy encoding with a high compression rate and reversibly compressing the error between the reproduced signal and the original signal, reversible compression at a high compression rate is possible. This combination compression method is proposed in Patent Document 1. This method is described in detail in the above document, but will be briefly described below.
[0004]
In the encoder, a digital input signal (hereinafter also referred to as an input signal sample sequence) is sequentially divided by a frame division unit into frame units each comprising, for example, 1024 input signal samples. The digital signal is lossy compression encoded for each unit. This encoding may be any system suitable for the input signal as long as the original digital input signal can be reproduced to some extent at the time of decoding. For example, if the digital input signal is speech, speech coding recommended as the ITU-T G.729 standard can be used, and if it is music, Twin VQ (Transform-Domain) adopted in MPEG-4. Weighted Interleaved Vector Quantization) can be used, and the lossy encoding method described in the above document can also be used. This lossy compression code is locally decoded, and an error signal between the local signal and the original digital signal is generated. Actually, local decoding is not necessary, and an error between the quantized signal obtained when generating the lossy compression code and the original digital signal may be obtained. The amplitude of this error signal is usually much smaller than the amplitude of the original digital signal. Therefore, the amount of information can be reduced by reversibly compressing and encoding the error signal rather than reversibly compressing and encoding the original digital signal.
[0005]
In order to increase the efficiency of the lossless compression encoding, the bit positions of all the samples in the frame of the sample sequence in which the absolute value of the error signal is represented (binary number of polarity and absolute value), that is, the MSB, For each second MSB,..., LSB, a bit string in which bits are concatenated in the sample sequence direction (time direction) is generated. That is, bit array conversion is performed. For convenience, the bit string consisting of 1024 bits at the same bit position connected to each other will be referred to as “coordinate bit string”. In contrast, for convenience, a bit string of one word representing an amplitude value including the polarity of each sample is referred to as an “amplitude bit string”. Since the error signal has a small amplitude, one or a plurality of consecutive bits from the most significant bit of each sample often becomes “0”. Therefore, the lossless compression encoding efficiency of the error signal can be improved by expressing the equivalence bit string generated by concatenating these bit positions with a predetermined code.
[0006]
These coordinate bit strings are lossless compression encoded. As lossless compression coding, for example, entropy coding such as Huffman coding or arithmetic coding using a case in which there is a sequence in which the same code (1 or 0) continues or a sequence that frequently appears is used. Can be used.
On the decoding side, the lossless compression code is decoded, and the bit sequence is inversely converted with respect to the decoded signal. That is, the equivalence bit string is converted into the amplitude bit string for each frame, and the obtained error signal is sequentially converted. Played. Further, the lossy compression code is decoded, the decoded signal and the reproduced error signal are added, and finally, the addition signals for each frame are sequentially connected to reproduce the original digital signal sequence.
[0007]
[Patent Document 1]
Patent Application Publication 2001-44847
[0008]
[Problems to be solved by the invention]
An object of the present invention is to provide an encoding method, a decoding method, an encoding device, a decoding device, and a program thereof that can compress information without distortion and select a hierarchical sample rate. It is in.
[0009]
[Means for Solving the Problems]
  According to this invention,At least one of the stepwise hierarchy of sampling frequencies and the stepwise hierarchy of quantization accuracy, which are attributes of the digital signal, encode a predetermined digital signalThe digital signal encoding method isHigher frequency is higher, lower is lower, higher quantization accuracy is higher, lower is lower.
  (a) Signal to be encoded and its signalThanattributeHierarchy ofSubordinateThe faithGenerating a differential signal from the signal or its deformation signal;
  (b) lossless encoding the difference signal;
including.
  According to this invention,At least one of the stepwise hierarchy of sampling frequencies and the stepwise hierarchy of quantization accuracy, which are attributes of the digital signal, encode a predetermined digital signalThe digital signal encoder isHigher frequency is higher, lower is lower, higher quantization accuracy is higher, lower is lower.
  The signal to be encoded and its signalThanattributeHierarchy ofSubordinateThe faithDifferential signal generating means for generating a differential signal with the signal or its modified signal;
  Differential signal lossless encoding means for losslessly encoding the difference signal;
Including.
[0010]
  According to this invention,At least one of the stepped hierarchy of sampling frequencies and the stepped hierarchy of quantization accuracy, which are attributes of digital signals, is predetermined.A decoding method for decoding a code of a digital signal is as follows:Higher frequency is higher, lower is lower, higher quantization accuracy is higher, lower is lower.
  (a) decoding an input code to generate a differential signal;
  (b) The difference signal and the difference signalThanattributeHierarchy ofSubordinateRecoveryGenerating a decoded signal by synthesizing the signal or its modified signal;
Including.
  According to this invention,At least one of the stepped hierarchy of sampling frequencies and the stepped hierarchy of quantization accuracy, which are attributes of digital signals, is predetermined.The digital signal decoding device is:Higher frequency is higher, lower is lower, higher quantization accuracy is higher, lower is lower.
  Differential signal decoding means for decoding the input code to generate a differential signal;
  The difference signal and the difference signalThanattributeHierarchy ofSubordinateRecoveryA signal synthesis means for synthesizing the encoded signal or its modified signal to generate a decoded signal;
Including.
[0011]
  According to this invention,At least one of the stepped hierarchy of sampling frequencies and the stepped hierarchy of quantization accuracy, which are attributes of digital signals, is predetermined.A computer-executable encoding program describing a digital signal encoding processing procedure,Higher frequency is higher, lower is lower, higher quantization accuracy is higher, lower is lower.
  (a) Signal to be encoded and its signalThanattributeHierarchy ofSubordinateThe faithGenerating a differential signal from the signal or its deformation signal;
  (b) lossless encoding the difference signal;
including.
  According to this invention,At least one of the stepped hierarchy of sampling frequencies and the stepped hierarchy of quantization accuracy, which are attributes of digital signals, is predetermined.A computer-executable program that describes a decoding processing procedure for decoding a code of a digital signal.Higher frequency is higher, lower is lower, higher quantization accuracy is higher, lower is lower.
  (a) decoding an input code to generate a differential signal;
  (b) The difference signal and the difference signalThanattributeHierarchy ofSubordinateRecoveryGenerating a decoded signal by synthesizing the signal or its modified signal;
Including.
[0012]
DETAILED DESCRIPTION OF THE INVENTION
First embodiment
A first embodiment of the present invention will be described with reference to FIG. In the figure, the sampling rate (frequency) of the signal is also indicated by a symbol. The digital signal from the input terminal 11 is divided by the frame dividing unit 12 in units of frames, for example, every 1024 samples, and the digital signal for each frame is divided by the downsampling unit 13 at the first sampling frequency F.₁Lower sampling frequency F from this digital signal₂Are converted into digital signals. In this case, the second sampling frequency F₂So that no aliasing signal is generated by sampling of the₂/ 2 or more high frequency components are removed.
[0013]
Second sampling frequency F₂The digital signal is subjected to irreversible or lossless compression encoding in the encoding unit 14 and is output as the main code Im. When irreversible compression encoding is performed by the encoding unit 14, for example, the main code Im is decoded by the local decoding unit 15, and the decoded local signal is second sampling frequency F by the upsampling unit 16.₂From the local signal of the first sampling frequency F₁Are converted into local signals. Alternatively, when irreversible encoding is performed so that the quantization error is minimized in the encoding unit 14, the quantized signal obtained at that time is the same as the output signal of the local decoding unit 15. In this case, the local decoding unit 15 is not necessary. When the encoding unit 14 performs lossless encoding, since the output of the local decoding unit 15 is the same as the input signal of the encoding unit 14, the local decoding unit 15 is omitted, and is indicated by a two-dot chain line in the figure. As described above, the input signal of the encoding unit 14 may be supplied to the upsampling unit 16. In any case, the signal given to the upsampling unit 16 corresponds to the main code Im, and will be referred to as a local signal for the sake of convenience in the description of the following embodiments, but the local decoding unit is not used in this way. However, the same applies to other embodiments.
[0014]
The error calculation unit 17 uses the first sampling frequency F₁The difference between the local signal and the digital signal having the first sampling frequency branched from the frame dividing unit 12 is calculated as an error signal and supplied to the array conversion / encoding unit 18. The processing of the array conversion / encoding unit 18 will be described later, but includes a bit array conversion unit and a lossless encoding unit, and encodes the error signal into an error code Pe that can be restored correctly, that is, reversibly. The error code Pe and the main code Im from the array conversion / encoding unit 18 are formatted in a format required by the output unit 19 and output to the output terminal 21.
[0015]
The code sequence signal output from the encoding device 10 according to the present invention may be transmitted to the decoding device 40 through a transmission path, or may be temporarily stored in a recording medium and later read out from the recording medium. It may be given to the decoding device 40. When transmitting the code sequence signal through the transmission path, for example, the output unit 19 assigns the priority to the main code Im and the error code Pe for each predetermined length (for example, one to a plurality of frames) as necessary. It is packetized and output sequentially. When storing the code sequence in the recording medium, for example, the main code Im and the error code Pe are combined for each frame to form a series of combined code sequences, and output as a plurality of parallel bits according to the interface of the connected device Or output as a single bit string. In the following description, a case where the main code Im and the error code Pe are output as packets will be described as an example.
[0016]
In the decoding device 40, the received packet from the input terminal 41 is separated into the main code Im and the error code Pe in the input unit 42, and the main code Im corresponds to the encoding unit 14 of the encoding device 10 in the decoding unit 43. The second sampling frequency F is irreversibly or reversibly decoded by the decoding process.₂The decoded signal is obtained. The decoded signal having the second sampling frequency is up-sampled by the up-sampling unit 44 to obtain the first sampling frequency F.₁Is converted into a decoded signal. At this time, set the sampling frequency to F₂In order to make it higher, an interpolation process is performed and a local signal is obtained.
[0017]
The separated error code Pe is subjected to a process of reproducing an error signal by the decoding / array inverse transform unit 45. The specific configuration and processing of the decoding / array inverse transform unit 45 will be described later. The sampling frequency of the recovered error signal is the first sampling frequency F₁The error signal and the local signal from the upsampling unit 44 are added by the adding unit 46 and supplied to the frame synthesizing unit 47 as a reproduced digital signal. In the frame synthesizing unit 47, the digital signals for each reproduced frame are sequentially combined and supplied to the output terminal 48. In a more realistic configuration, as shown by a broken line, a loss detection unit 49 that detects a loss of a packet with an error code Pe and a loss correction unit 58 that corrects a decoded error signal sample based on the packet loss detection are decoded and arranged. Although provided on the output side of the inverse conversion unit 45, these will be described in detail later with reference to FIGS.
[0018]
According to this configuration, it is possible to reproduce a high quality signal having the same sampling frequency as that of the original digital signal by using both the main code Im and the error code Pe. In addition, when encoding and outputting with a packet, a high priority is given to the packet with the main code Im, so that a signal with a relatively high quality can be reproduced even when the packet with the error code Pe is lost. In addition, if a user does not require such a high-quality data signal, providing only a main code Im based on a signal whose sampling frequency is lower than that of the original digital signal provides a high-quality signal with a small amount of information. can do. For example, when digital signals are transmitted over a network, that is, on the transmitting side, only the main code Im or an error from the main code Im depending on the network conditions (route communication capacity and traffic) or according to the request on the receiving side There is a degree of freedom in selecting whether to send both of the signs Pe.
When lossless encoding is performed as the encoding unit 14, the processing similar to the processing performed by the array conversion / encoding unit 18 may be performed. In that case, the decoding unit 43 performs the decoding process in the same manner as the decoding / array inverse conversion unit 45.
[0019]
Second embodiment
In the second embodiment of the present invention, the sampling frequency of the data signal is hierarchized in multiple stages, and it is possible to selectively provide more types of quality signals.
As shown in FIG. 2 with the same reference numerals assigned to the portions corresponding to FIG. 1, in the second embodiment, the first sampling frequency F from the error calculation unit 17 is shown.₁Of the error signal is down-sampled by the down-sampling unit 22 to obtain the first sampling frequency F.₁Third sampling frequency F lower but higher than the second sampling frequency_ThreeIs converted into an error signal. For example, in the downsampling unit 13, the sampling frequency F of the input signal₁Sampling frequency F₂Get the signal. In the downsampling unit 22, the sampling frequency F of the error signal₁Is reduced to half the sampling frequency F_ThreeThe error signal is obtained. That is, the sampling frequency is F₁= 4F₂And F₁= 2F_ThreeDecide on the relationship.
[0020]
Third sampling frequency F from the downsampling unit 22_ThreeThe error signal is irreversibly or losslessly encoded by the encoding unit 23 and output as a slave code Ie. The sub code Ie is decoded by the local decoding unit 24 to obtain a third sampling frequency F._ThreeThe local signal of the third sampling frequency is up-sampled by the up-sampling unit 25, and the first sampling frequency F is output.₁Are converted into local signals. An error between the local signal of the first sampling frequency and the error signal of the first sampling frequency from the error calculation unit 17 is calculated as an error signal by the error calculation unit 26, and this error signal is supplied to the array conversion / encoding unit 18. Then, an error code Pe is generated by an array conversion / encoding unit described later. Similarly to the local decoding unit 15, the local decoding unit 24 can be omitted. For example, when the encoding unit 23 performs lossy encoding, the input signal of the encoding unit 23 is quantized so as to minimize the error. What is necessary is just to give the up-sampling unit 25 the quantized signal obtained by the processing to be converted. Alternatively, when the encoding unit 23 performs lossless encoding, an input signal of the encoding unit 23 may be given to the upsampling unit 25. Similarly, in the following other embodiments, when a configuration in which the local decoding units 15 and 24 are omitted is possible, those blocks are indicated by broken lines. The output unit 19 packetizes the main code Im, the subcode Ie, and the error code Pe, and outputs the information data with priority as necessary.
[0021]
In the decoding device 40, the input unit 42 separates the main code Im, the subcode Ie, and the error code Pe from the received packet, the main code Im is supplied to the decoding unit 43, and the error code Pe is the decoding / array inverse conversion unit 45. The same processing as that for the main code Im and the error code Pe in the decoding unit 43 and the decoding / array inverse conversion unit 45 shown in FIG.₂Main signal and sampling frequency F₁Error signal is obtained.
The sub code Ie is decoded by the decoding unit 27, and the third sampling frequency F_ThreeThe decoded slave signal is reproduced. The decoding unit 27 performs a decoding process corresponding to the encoding method of the encoding unit 23 of the encoding device 10. Third sampling frequency F_ThreeThe up-sampling unit 52 outputs the decoded signal of the first sampling frequency F₁The decoded signal having the first sampling frequency and the decoded signal having the first sampling frequency from the upsampling unit 44 are added by the adding unit 43, and the added first sampling frequency F is added.₁The first sampling frequency F from the decoded signal and the decoding / array inverse transform unit 45₁Are added by the adder 46 and supplied to the frame synthesizer 47 as a reproduced digital signal.
[0022]
In the upsampling unit 44, in the case of the relationship of the sampling frequency in the encoding device, the sampling frequency F₂4 times the sampling frequency F₁In the upsampling unit 52, the sampling frequency F_ThreeIs doubled and sampling frequency F₁To.
According to this configuration, if all information, that is, Im, Ie, and Pe, can be obtained correctly, a high sampling frequency F₁The original digital signal can be obtained. When the reproduction error signal cannot be obtained, the second sampling frequency F from the decoding unit 43 as shown by the broken line in the figure.₂The up-sampling unit 54 converts the decoded signal of the third sampling frequency F_ThreeThis signal and the decoded signal from the decoding unit 27 are added by the adding unit 55 and supplied to the frame synthesizing unit 47 as a reproduced digital signal, so that the quality is slightly lower than that of the original digital signal. Is the sampling frequency F_ThreeThe digital signal of the quality can be obtained from a code obtained by high-efficiency compression coding.
[0023]
To further increase the coding efficiency, the decoded signal of the main code Im, that is, the second sampling frequency F from the decoding unit 43₂Only the decoded signal can be supplied to the frame synthesizing unit 47 as a reproduced digital signal.
For example, as the original digital signal, the first sampling frequency F₁Is a 192kHz music signal and the third sampling frequency F_ThreeIs 96kHz and the second sampling frequency F₂Is 48 kHz, generally, even a reproduced digital signal with a sampling frequency of 48 kHz can obtain a high quality comparable to that of a compact disc (CD). In other words, high quality information can be provided with a small amount of information, and a user who desires a 96 kHz reproduced digital signal with a higher sampling frequency can use a main code Im and a subcode Ie to achieve a higher compression ratio and higher than CD. A quality signal can be provided. The user's decoding device 40 who wants a signal having a higher sampling frequency can reproduce the original digital signal of 192 kHz by using Im, Ie and Pe.
[0024]
Modification of the second embodiment
A modification of the second embodiment in which the sampling frequency is changed in multiple stages will be described with reference to FIG. In FIG. 3, parts corresponding to those in FIG. In this example, the encoding apparatus 10 supplies a digital signal for each frame to the encoding unit 14 through a plurality of stages of downsampling units. In the figure, the first sampling frequency F is shown in the case of a two-stage configuration using the downsampling unit 13 and the downsampling unit 27.₁The sampling frequency of the output signal of the downsampling unit 13 to which the above signal is input is the third sampling frequency F in the above example._ThreeAnd its third sampling frequency F_ThreeThe sampling frequency of the output signal of the downsampling unit 27 to which the above signal is input is the second sampling frequency F₂It is. Sampling frequency F of local signal obtained by decoding main code Im from encoding unit 14₂The upsampling unit 16 is the sampling frequency of the input signal of the downsampling unit 27 immediately before the encoding unit 14, that is, the third sampling frequency F._ThreeAre converted into local signals. In the case of the relationship of the sampling frequency, the downsampling units 13 and 27 respectively convert the sampling frequency to half. This third sampling frequency F_ThreeThe error signal between the local signal and the input signal of the down-sampling unit 27 is calculated as an error signal by the error calculation unit 52, and this error signal is encoded by the encoding unit 23 by irreversible or lossless, preferably high compression encoding. Is output as a secondary code Ie.
[0025]
This secondary code Ie is decoded by the local decoding unit 24 and is then subjected to a third sampling frequency F._ThreeThe local signal is obtained. The local signal and the input signal of the downsampling unit 27 are added by the adding unit 29, and the added local signal having the third sampling frequency is converted by the upsampling unit 25 into an adding local signal having the first sampling frequency. The error between the added local signal and the digital signal branched from the output from the frame dividing unit 12 is calculated as an error signal by the error calculating unit 17, and this error signal is supplied to the array conversion / encoding unit 18 and supplied to the error code Pe. Is generated. The error code Pe, the main code Im, and the sub code Ie are combined by the output unit 19 and output.
[0026]
In the coding apparatus 10 of the modified embodiment shown in FIG. 3, as in the case of the coding apparatus in FIGS. 1 and 2, the local decoding sections 15 and 24 are not used, but the coding sections 14 and 23, respectively. Is supplied to the upsampling unit 16 and the addition unit 29 (when the encoding units 14 and 23 perform lossy encoding), or the input signals of the encoding units 14 and 24 are respectively input to the upsampling unit 16 and You may comprise so that it may give to the addition part 29 (when the encoding parts 14 and 23 perform lossless encoding).
[0027]
In the decoding device 40, the input packet from the terminal 41 is separated into the main code Im, the sub code Ie, and the error code Pe by the input unit 42, and the decoding unit 43, the decoding unit 51, and the decoding / array inverse conversion unit 45 respectively. The local signal and the error signal are reproduced as in the case shown in FIG. Second sampling frequency F from the decoding unit 43₂Is decoded by the upsampling unit 44 in this example at the third sampling frequency F._ThreeIs converted into a decoded signal of the third sampling frequency F from the decoding unit 51._ThreeAre added by the adder 53. The added decoded signal is converted into a decoded signal having the first sampling frequency by the upsampling unit 52, and the decoded signal and the first sampling frequency F from the decoding / array inverse converting unit 45 are converted.₁The error signal is added to the frame synthesis unit 47 as a reproduced digital signal.
[0028]
If sufficient information for reproducing the error signal cannot be obtained, or if the error code Pe is not inputted, the third sampling frequency F from the adder 53 is used._ThreeAre added to the frame synthesizer 47 as a reproduced digital signal. Further, when only the main code Im is obtained, the second sampling frequency F from the decoding unit 43 is obtained.₂The decoded signal is supplied to the frame synthesizing unit 47 as a reproduced digital signal.
In the second embodiment shown in FIGS. 2 and 3, the sampling frequency is converted into two stages, but the same encoding and decoding can be performed by converting the sampling frequency more than three stages.
[0029]
Array conversion / encoding unit
A specific example of the array conversion / encoding unit 18 in the embodiment of each encoding apparatus shown in FIGS. 1, 2, and 3 will be described with reference to FIG. The error signal from the error calculator 17 (26 in FIG. 2) is supplied to the auxiliary information generator 18E. For each frame, the effective digit number detection unit 18E5 in the auxiliary information generation unit 18E detects the number of digits representing the maximum absolute value of the error signal samples in the frame as the effective digit number Fe. In addition, each bit of the error signal sample within the effective number of digits is extracted by the bit array conversion unit 18A as a coordinate bit string.
[0030]
The equidistant bit string from the bit array conversion unit 18A is divided into transmission unit or recording unit data by the transmission / recording unit dividing unit 18B. These divided transmission record unit data are losslessly encoded by the lossless compression encoding unit 18C as necessary, and are provided to the auxiliary code adding unit 18D as the error data code Ine. The auxiliary code adding unit 18D adds an auxiliary code Inx from an auxiliary information encoding unit 18F described later to the error data code Ine and outputs it as an error code Pe.
An example of bit array conversion is shown in FIG. 5A. The amplitude bit string of each error signal sample expressed as a polarity code and an absolute value is shown in each vertical column on the left side in FIG. 5A, and one frame of the amplitude bit string is sequentially arranged in the sample direction. In order to make it easier to understand the state of one amplitude bit string, the amplitude bit string DV (k) straddling the amplitude is indicated by a bold line, k represents the time in the frame, for example, k = 1, 2,. is there. In this example, the polarity code of each amplitude bit string DV (k) is adjacent to the MSB of the absolute value, and in the figure, the polarity code is positioned immediately above the MSB (Most Significant Bit).
[0031]
An error signal expressed as an absolute value of polarity is supplied to the effective digit number detection unit 18E5, and a place where "1" is present in the digit closest to the MSB in one frame of the amplitude bit string of the error signal is detected, LSB (Least Significant Bit) : The least significant bit) to that digit is obtained as the effective digit Fe. Only the partial LBP within the effective number of digits and the polarity code in the error signal of one frame are converted into a coordinate bit string. That is, it is not necessary to convert the partial HBP from the digit above the effective digit number to the MSB into a coordinate bit string.
As shown on the right side of FIG. 5A, in this example, such sample arrangement data is first converted into a bit string obtained by concatenating only the polarity bits (signs) of the numerical values of the amplitudes of the samples (amplitude bit strings) in the time direction in the frame. Extract as a bit string. Next, a coordinate bit string in which only the most significant digit within the effective number Fe is concatenated in the frame is extracted, and in the same manner, a coordinate bit string concatenated in the time direction for each digit (corresponding bit position) is sequentially extracted. Finally, a coordinate bit string in which only LSBs are concatenated in the frame is extracted. One example of these extracted equidistant bit strings is indicated as DH (i) with a thick frame in the right horizontal arrangement in FIG. 5A. i indicates the bit position in the amplitude bit string before the array conversion of the bits constituting the coordinate bit string. The above bit arrangement conversion does not change the contents of each bit constituting the bit string.
[0032]
In the case of performing bit array conversion of a sample string when each error signal sample is expressed by a two's complement and a positive / negative integer, for example, as shown in FIG. 5B, the absolute value of the sample is shown as an amplitude bit string for one frame. The portion of the digit higher than the digit representing the maximum value (indicated by the portion HBP in FIG. 5B) is all “0” if the amplitude bit string is a positive value, and “1” if the amplitude bit string is a negative value. "Become. The number of digits of the other partial LBP is detected by the effective digit number detection unit 18E5 of FIG. 4 as the effective digit number Fe. Only the significant digit portion LBP and the bit position (digit) adjacent thereto, that is, the polarity code, need only be converted into a coordinate bit string.
[0033]
The transmission record unit dividing unit 18B divides the data into transmission record unit data for each equal bit string DH (i) or for each of a plurality of adjacent bit strings DH (i) adjacent to each other. In this case, in one frame, data composed of one coordinate bit string and data composed of a plurality of coordinate bit strings may be mixed as transmission recording unit data. Each transmission record unit data divided in this way is losslessly compressed and encoded by the lossless compression unit 18C, and is given to the auxiliary code adding unit 18D as an error data code Ine.
For example, as shown in FIG. 5C, the output unit 19 stores the error code of the transmission record unit data in the payload PYL, and adds the header HD to the payload PYL. For example, the header HD is provided with a packet number PKTN including, for example, a frame number and a transmission record unit data number (number in output order) in the frame, and a priority PRIO and a data length DTL of the packet as necessary. The error signal sample sequence can be reconstructed for each frame.
[0034]
The data length DTL is not required if the data length of the transmission recording unit data (payload) PYL is fixed, but is required because the data length differs depending on the packet when compressed by the lossless compression unit 18C. Furthermore, generally, an error detection code ED such as a CRC code for detecting whether or not an error has occurred in the entire packet is added at the end to form one packet PKT. The main code Im and the subcode Ie are packetized in the same manner, and the error code Pe, the main code Im, and the subcode Ie packet PKT are sequentially output to the output terminal 21.
When priority is given to the packet PKT, higher priority is given to the packet PKT including the corresponding transmission recording unit data closer to the MSB side. For example, the priority is set to about 2 to 5 levels. The highest priority is given to the equivalence bit string of the polarity code, and then the priority is given in the order of the bit string representing the main code Im and the bit string representing the sub code Ie.
[0035]
Returning to the explanation of FIG. 4, the effective digit number Fe detected by the effective digit number detection unit 18E5 is encoded by the auxiliary information encoding unit 18F and transmitted. Further, in the example of FIG. 4, a parameter sequence LPC that expresses a spectral envelope by a spectral envelope calculation unit 18E4 from a sample sequence of an error signal for each frame is obtained as a linear prediction coefficient by, for example, linear prediction analysis, and also by a power calculation unit 18E1 The average power PW of the error signal for each frame is calculated. Alternatively, an error signal is input to the inverse filter 18E2 configured based on the linear prediction coefficient sequence obtained by the spectrum envelope calculation unit 18E4 and normalized by the spectrum envelope, and then flattened, and the average power of the flattened signal Is obtained by the power calculator 18E3. The linear prediction coefficient LPC and the average power PW are also quantized by the auxiliary information encoding unit 18F with a low bit of, for example, about 30 to 50 bits / second, and a code representing these quantized values is output as the auxiliary code Inx. The auxiliary code Inx obtained by encoding the effective digit number Fe, the spectral envelope parameter string LPC, and the average power PW is supplied to the output unit 19, and the representative packet of each frame, for example, transmission record unit data including the polarity code is stored. It is added in the packet or output as an independent packet.
[0036]
The above-described array conversion / encoding unit described the case where the maximum number of significant digits of a sample is detected for each frame and the bits in the significant digits are array-converted, but the efficiency is somewhat lower than that, Instead of detecting significant digits, all bits from the least significant bit to the most significant bit of the sample sequence may be encoded by bit array conversion processing.
[0037]
Decoding / array inverse transform unit
A specific example of the decoding / array inverse transform unit 45 corresponding to the above-described array transform / encoding unit 18 is shown in FIG. The decoding / array inverse transform unit 45 includes a separation unit 45A, a lossless decompression unit 45B, a transmission record unit integration unit 45C, and a bit array inverse transform unit 45D. The missing correction unit 58 includes an auxiliary information decoding unit 58D, a switch 58A, a missing information correction unit 58B, and a digit alignment unit 58C.
[0038]
The packet of the error code Pe separated by the input unit 42 is separated into an error data code Ine and an auxiliary code Inx by the separation unit 45A. The error data code Ine is supplied to the lossless decompression unit 45B, and the auxiliary code Inx is supplied to the auxiliary information decoding unit 58D of the missing correction unit 58. The auxiliary information decoding unit 58D decodes the effective digit number Fe of the frame, the parameter sequence LPC indicating the spectral envelope and the code indicating the average power PW, and supplies the effective digit number Fe to the digit aligning unit 58C, and the spectral envelope parameter sequence LPC. The average power PW is supplied to the missing information correction unit 58B.
[0039]
The error data code Ine is losslessly decoded by the lossless decompression unit 45B, and the error data of the obtained transmission recording unit is obtained by the transmission recording unit integration unit 45C based on the packet number. They are integrated as shown in the arrangement of equidistant bit strings on the right side of FIG. 5A. The integrated equidistant bit string is converted into an amplitude bit string, that is, a sample string (waveform) by the bit array inverse conversion unit 45D. At this time, if the transmission recording unit data is created from what each sample represents with a polarity code and an absolute value, the bit array inverse conversion unit 45D reverses the bit array conversion described with reference to FIG. As shown on the left side of FIG. 5A, the equivalence bit sequence shown on the right side of FIG. 5A is converted into an amplitude bit sequence and output as an error signal sample sequence. This inverse array conversion is to extract the bits belonging to the same sample in the encoding device 10 in each error data equivalence bit string from the transmission storage unit integration unit 45C to form an amplitude bit string of one sample.
[0040]
When the transmission record unit data is based on a direct conversion from an amplitude bit string expressed in two's complement to a coordinate bit string, the amplitude of the coordinate bit string shown on the right side in FIG. 5B is shown on the left side. This is converted into an array of bit strings, but the processing is the same as the array inverse conversion in the case of the sample composed of the above-described polarity value and absolute value. The error signal sample from the bit array inverse conversion unit 45D is supplied to the digit alignment unit 58C. The digit matching unit 58C performs digit matching for each amplitude bit string in accordance with the decoding effective digit number Fe. That is, “0” is added to the upper part of the amplitude bit string so as to correspond to the digit part HBP of FIG. 5A so that the number of bits (number of digits) of the original amplitude bit string is obtained. When the sample is expressed in two's complement, for example, “0” is added to the digit portion HBP in FIG. 5B if the polarity sign is positive, and “1” is added if it is negative. The amplitude bit string thus aligned is output as a reproduction error signal sample string (that is, a decoded error signal sample string).
[0041]
If a packet loss has occurred, the missing packet number is detected from the packet number of the received packet by the loss detection unit 49, whereby the switch 58A is switched and the amplitude bit string from the bit array inverse conversion unit 45D is aligned. Without being directly supplied to the unit 58C, it is supplied to the missing information correcting unit 58B, the missing information is corrected for the amplitude bit string (sample), and then supplied to the digit aligning unit 58C.
[0042]
Correction in the missing information correction unit 58B is performed by estimating missing information from known information. If some packets, packets with low-priority LSB side bits are lost, the numerical value corresponding to the missing part cannot be determined, so a small value, for example, 0, or the minimum and maximum values that the missing part can take The waveform must be replayed using the median value. In this case, the accuracy of the determined number of bits on the MSB side can be maintained, but the auditory distortion is large. The reason for this is that the energy of the original sound spectrum is often biased in the low range, whereas the distortion component due to missing bits has a nearly flat spectrum shape, so that the high frequency component is larger than the original sound, and the listener will hear it when played back. Sounds like noise. Therefore, the value of the uncertain waveform is corrected so that the spectrum of the uncertain component approximates the average spectrum or the spectrum determined every frame. As a result, the spectrum of the corrected distortion component has a small high frequency component, and the distortion is masked by the original sound, thereby improving the quality.
[0043]
In other words, the spectrum obtained from information other than the missing information of the frame approximates the average spectrum of the past several frames or the determined spectrum obtained as a result of decoding the auxiliary information as described later. In addition, correction for missing information is performed. A preferred method for this correction will be described later. As a simple correction method, the missing information correction unit 58B may smooth the input reproduction sample sequence through a low-pass filter to remove high-frequency noise components. If the spectrum shape (envelope) of the original signal is known in advance, the cutoff characteristic of the low-pass filter is selected so as to attenuate the high-frequency component from the cutoff frequency according to the characteristic. Alternatively, as described above, an average spectrum may be obtained, or the cutoff characteristic may be adaptively changed in accordance with the shape of the spectrum determined for each frame.
[0044]
In this way, since the decoding / array inverse transform unit 45 can correct the missing information due to the packet loss, the LSB side packet is not sent intentionally as necessary, and the decoding / array inverse transform is performed even if the encoding compression efficiency is increased. The unit 45 can perform reversible decoding, or can perform reproduction with an error that does not cause an audible problem.
Alternatively, all combinations of possible values of missing information (bits) are added to each sample value to create correction sample string (waveform) candidates, and the spectral envelopes of these candidates are obtained. A correction sample sequence (waveform) candidate corresponding to the one closest to the decoded spectrum envelope is output as a correction sample sequence to the digit alignment unit 58C. 4 and 6, the lossless compression unit 18C and the lossless decompression unit 45B may be omitted.
In the above description of the decoding / array inverse transform unit, the processing corresponding to the case where the encoding device 10 detects the significant digits of the sample and the bits in the effective digits are array-transformed has been described. When all the bits of the sample sequence are array-converted without detecting significant digits, digit alignment is not necessary in the decoding device 40.
[0045]
Correction by auxiliary information
When creating correction sample string candidates using all combinations of possible values of missing information, if the missing information (bits) increases, the number of correction sample string (waveform) candidates increases significantly and the amount of processing increases. There is a risk of becoming unrealistic. The processing of the missing information correction unit 58B and its functional configuration to prevent such a problem will be described below.
[0046]
FIG. 7 shows an example of the processing procedure, and FIG. 8 shows an example of the functional configuration. First, the provisional waveform generator 58B1 reproduces the provisional waveform (provisional sample string) in the frame using only the confirmed bits input from the array conversion unit 45D (S1). In the reproduction of the provisional waveform, missing bits are fixed to 0, for example, or set to an intermediate value between the minimum value and the maximum value that the missing bits can take. For example, if the lower 4 bits are missing, any of the levels from 0 to 15 is a correct value, but is set to 8 or 7.
[0047]
Next, the spectrum envelope of the provisional waveform is calculated by the spectrum envelope calculation unit 58B2 (S2). For example, the spectral envelope can be estimated by performing all-pole linear prediction analysis used in speech analysis on the provisional waveform. The error calculation unit 58B3 compares the estimated spectrum envelope with the spectrum envelope of the original sound sent as auxiliary information, that is, the spectrum envelope decoded by the auxiliary information decoding unit 58D, and the error is within a predetermined allowable range. For example, the switch SW1 is controlled to output the provisional waveform as a corrected reproduction error signal (S3).
[0048]
In step S3, when the error between the estimated spectral envelope shape and the decoded spectral envelope shape is larger than the allowable range, first, the inverse characteristic of the estimated spectral envelope is given to the provisional waveform (S4). Specifically, the parameter representing the spectral envelope obtained in step S2 is set in, for example, an all-pole linear prediction inverse filter (all zero type) 58B4, and the provisional waveform given from the provisional waveform generator 58B1 through the switch SW2 is set. By passing through the inverse filter 58B4, the spectrum of the provisional waveform is flattened to obtain a flattened signal. The average power of the flattened signal is calculated by the power calculator 58B5 and corrected from this average power and the average power PW decoded from the auxiliary information decoder 58D (output of the flattened power calculator 18E1 in FIG. 4). The amount is calculated by the correction amount calculation unit 58B6, for example, the ratio or difference between them is taken, and the amplitude correction is performed on the output power value of the inverse filter 58B4 by the power correction unit 58B7 based on the correction amount. That is, the output of the inverse filter 58B4 is multiplied or added by the correction amount to match the output power value of the power correction unit 58B7 with the decoded power value (S5).
[0049]
Next, the spectral envelope characteristic of the auxiliary information is given to the flattened signal whose amplitude has been corrected to correct the spectral envelope (S6). That is, a spectrum correction waveform is created through the output of the power correction unit 58B7 in the all-pole type synthesis filter 58B8 using the parameter LPC representing the decoded spectrum envelope of the auxiliary information. The resulting spectral envelope of the waveform is close to the original error signal.
However, since this spectrum correction waveform may be inconsistent with the already determined digit bit, it is corrected to the correct value by the correction unit 58B9 (S7). For example, when the lower 4 bits of the amplitude value with 16-bit precision are unknown, there are 16 uncertain values for the values that can be taken by each sample, but the values are corrected to the values closest to the spectrum correction waveform. In other words, if the sample value corrected for each sample is out of the possible range, it is corrected to the limit value of the possible range. For example, if the upper 12-bit corrected sample value is larger than the correct 12-bit sample value, the upper 12 bits of the corrected sample value are corrected to the correct sample value, and all the lower 4 bits of the corrected sample value are set to “1” (upper limit). If it is small, set all to "0" (lower limit). With this modification, all the bits whose amplitude values are determined are matched, and at the same time, a waveform having a spectrum envelope close to that of the original error signal can be reproduced.
[0050]
By using this corrected waveform as a provisional waveform in step S1, the processing after step S2 can be repeated. In addition, when the number of significant decoding digits varies from frame to frame, there are cases where samples subject to processing by the linear envelope analysis of the spectrum envelope calculation unit 58B2, the inverse filter 58B4, and the synthesis filter 58B8 span the current frame and the past frame. is there. In this case, even if the object of processing is the current frame, it is necessary to perform analysis and filter processing after aligning the number of significant digits of the past frame with the number of significant digits of the current frame. If the number of significant digits of one past frame is N digits less than the number of significant digits of the current frame, the amplitude value is reduced by shifting the sample of the past frame, for example, by N digits lower, and reducing the amplitude value. Match the number to the number of significant digits in the current frame. Conversely, if the number of significant digits of one previous frame is M digits greater than the number of significant digits of the current frame, the amplitude of the sample of the past frame is temporarily shifted to the upper M digits by, for example, floating point display. Increase the value to match the number of significant digits to that of the current frame. In this case, the accuracy of the amplitude value of the sample of the past frame is lowered when the information overflows from the register due to the upper shift and the information loss is large, so the sample of the past frame is not used or the sample of the current frame is not used. The correction process may be omitted.
[0051]
As shown by the broken line in FIG. 7, such correction of the effective number of digits is performed before the effective processing for correcting the number of digits (S2 ′) described above when necessary for the analysis processing in step S2, and the reverse of step S4. If necessary for the filter processing, the effective digit number correction is performed before that (S4 '), and if necessary for the synthesis filter processing in step S6, the effective digit number correction is performed before that (S6'). Also, in FIG. 8, the significant digits decoded from the auxiliary information decoding unit 58D as shown by the broken line for those requiring samples of past frames in the spectrum envelope calculation unit 58B2, the inverse filter 58B4, and the synthesis filter 58B8. The number Fe is also supplied, and after the spectrum envelope calculation unit 58B2, the inverse filter 58B4, and the synthesis filter 58B8 perform the process of matching the number of significant digits for the sample of the past frame with the number of significant digits of the current frame, the original process To be done.
[0052]
Note that the waveform (sample value) is premised on an integer value, but it is treated as a real number in the filter calculation, and the output value of the filter must be converted to an integer. In the case of a synthesis filter, the result differs depending on whether it is converted into an integer for each sample or later in a batch for each frame, but both are possible.
As shown by broken lines in FIG. 7 and FIG. 8, after the provisional waveform is flattened in step S4, the flattened provisional waveform (flattened signal) is first passed through the synthesis filter 58B8 and the spectrum envelope corrected. A constituent sample series (waveform) may be obtained (S5 '), the amplitude of the spectrum envelope corrected waveform may be corrected by the power correction unit 58B7' (S7 '), and the process may proceed to step S7. In this case, the average power of the spectrum envelope corrected waveform from the synthesis filter 58B8 is calculated by the power calculation unit 58B5 ′, and this average power and the decoding power PW of the auxiliary information (the output of the power calculation unit 18E in FIG. 4). The correction amount is calculated by the correction amount calculation unit 58B6 ′, and the amplitude correction is performed by the power correction unit 58B7 ′ on the output of the synthesis filter 58B8 based on the correction amount.
[0053]
After step S3 in the process shown in FIG. 7, the filter coefficient of one synthesis filter 58B8 ′ combining the inverse filter 58B4 of the spectrum envelope estimated in step S2 and the synthesis filter 58B8 of the decoded spectrum envelope of the auxiliary information is obtained. The composite spectrum envelope calculation unit 58B10 may synthesize a waveform obtained by correcting the spectrum envelope through the provisional waveform in the synthesis filter 58B8 ′ in which the coefficient is set, and may perform amplitude correction on the waveform with the corrected spectrum envelope. . Further, when the bit array conversion unit 18A in the encoding device 10 does not detect the significant digit number Fe shown in FIGS. 5A and 5B and converts the entire amplitude bit string into an equivalent bit string, the significant digit number detection unit 18E5, The digit alignment unit 58C in the decoding device 40 related thereto may be omitted. Furthermore, it does not necessarily have to be divided into transmission recording units, and does not have to be packetized. In the case of packetization, the primary code Im, the secondary code Ie, and other codes in the first to third embodiments are also packetized.
[0054]
In this specification, packets in one frame are intentionally removed in order to adjust the amount of information, so that all the packets in one frame are not input to the decoder, or the switching center due to traffic congestion in the communication network For example, if a packet is missing due to not sending out some packets, or due to a transmission path failure, recording / playback device abnormality, etc., or there is an error in the input packet, the transmission record unit data cannot be decoded. When a packet cannot be used, or when a certain packet is abnormally delayed, it is collectively referred to as a packet loss.
According to the first and second embodiments described above, the original digital signal is encoded by converting its sampling frequency, and the error signal is output as at least a coordinate bit string as the sampling frequency of the original signal. A quality signal can be reproduced.
[0055]
Third embodiment
In the above-described embodiments of FIGS. 1, 2, and 3, the example in which the error signal from the error calculation unit 17 or 26 is array-converted and encoded in the array conversion / encoding unit 18 has been described. Array conversion / encoding may be performed. FIG. 9 shows a configuration in which the example is applied to the encoding device 10 of FIG. 1 and a configuration of the decoding device 40 corresponding thereto.
In this configuration, a prediction error generation unit 31 is provided between the error calculation unit 17 and the array conversion / encoding unit 18 in the encoding device 10 of FIG. 1, and a decoding / array inverse conversion unit 45 and an addition unit in the decoding device 40 are provided. 46 is provided with a predictive synthesis unit 56, and the other parts are the same as in FIG.
[0056]
As shown in FIG. 10A, for example, the prediction error generation unit 31 includes a prediction analysis unit 31A, a sample register 31B, a linear prediction unit 31C, an integerization unit 31D, and a subtraction unit 31E. For example, a plurality of samples of the previous error signal from the error calculation unit 17 are supplied from the sample register 31B to the linear prediction unit 31C, and the linear prediction coefficient LPC based on the spectral envelope parameter sequence from the prediction analysis unit 31A is applied to these samples. Are convolutionally calculated to obtain a linear prediction value. This linear prediction value is converted to an integer value by the integerization unit 31D, and the difference between the integer prediction value and the current sample of the error signal from the error calculation unit 17 is obtained by the subtraction unit 31E to obtain the prediction error signal Spe. The prediction error signal Spe is input to the array conversion / encoding unit 18.
[0057]
Alternatively, as shown in FIG. 10B, the prediction error generation unit 31 includes a prediction analysis unit 31A, a linear prediction unit 31C, an integerization unit 31D, and a subtraction unit 31E, and predicts and analyzes the error signal from the error calculation unit 17. The unit 31A performs linear prediction analysis to obtain the prediction coefficient LPC, and the linear prediction unit 31C performs a convolution operation on the corresponding samples of the prediction coefficient LPC and the error signal to obtain a prediction signal. The prediction signal is converted into an integer by the integerization unit 31D, and a difference between the integer prediction signal and the input error signal is obtained as a prediction error signal Spe by the subtractor 31E. The obtained prediction error signal Spe is given to the array conversion / encoding unit 18. Further, the coefficient code Ic corresponding to the quantized value of the prediction coefficient LPC obtained by the prediction analysis unit 31A is given to the output unit 19 as shown by the broken line in FIG.
[0058]
In each of the above-described embodiments, the encoding device 10 and the decoding device 40 can also function by causing the computer to execute an encoding program and a decoding program, respectively. In these cases, the lossless encoding program and the lossless decoding program are downloaded to the program memory of the computer from a CD-ROM, a flexible magnetic disk, etc. or through a communication line.
The prediction error signal Spe thus obtained is subjected to bit array conversion in the array conversion / encoding unit 18 in the same manner as described above, encoded to generate an error code Pe, and supplied to the output unit 19. The output unit 19 packetizes the error code Pe into the main code Im and, if necessary, the coefficient code Ic, and outputs the packet from the output terminal 21.
[0059]
In the decoding device 40, the separated error code Pe from the input unit 42 is decoded by a decoding / array inverse transform unit 45 to obtain a coordinate bit string, and one frame of these coordinate bit strings is arranged in an amplitude bit string. Then, the prediction error signal is reproduced. An error signal is reproduced from the prediction error signal by prediction synthesis in the prediction synthesis unit 56. The prediction synthesis unit 56 corresponds to the configuration of the prediction error generation unit 31 of the encoding device 10. That is, the prediction synthesis unit 56 in the decoding device 40 when the prediction error generation unit 31 has the configuration shown in FIG. 10A includes, for example, a linear prediction unit 56A, an addition unit 56B, and a prediction analysis unit, as shown in FIG. 11A. 56C and an integerizing unit 56D.
[0060]
The prediction analysis unit 56C determines the prediction coefficient so that the error power between the prediction signal generated by the linear prediction unit 56A and the reproduction error signal from the addition unit 56B is minimized, and the prediction coefficient is added by the linear prediction unit 56A. The prediction signal is output from the prediction signal generation unit 56A by performing a convolution operation on the plurality of past reproduction error signal samples from the unit 56B. The prediction signal is converted into an integer value by the integer converting unit 56D, and the prediction signal of the integer value and the prediction error signal from the decoding / array inverse transform unit 45 are added by the adding unit 56B, and the error signal is reproduced and output.
When the prediction error generation unit 31 of the encoding device 10 has the configuration shown in FIG. 10B, the prediction error generation unit 56 in the decoding device 40 includes, for example, a prediction signal generation unit 56A and an addition unit 56B as shown in FIG. 11B. And an integer converting unit 56D and a coefficient decoding unit 56E.
[0061]
The coefficient code Ic separated by the input unit 42 is decoded by the coefficient decoding unit 56E, and the linear prediction coefficient obtained by this decoding is convolved with the prediction error signal from the decoding / array inverse transform unit 45 in the prediction signal generation unit 56A. The prediction signal is generated by the calculation, and the obtained prediction signal is converted into an integer value by the integer converting unit 56D, and the prediction signal of this integer value and the prediction error signal from the decoding / array inverse transform unit 45 are added to the adding unit 56B. And an error signal is output.
The sampling frequency of the error signal reproduced in this way is the first sampling frequency F.₁This error signal and the first sampling frequency F from the upsampling unit 44₁The decoded signal is added by the adder 46 to reproduce the digital signal, which is supplied to the frame synthesizer 47. The frame synthesizing unit 47 sequentially connects the digital signals reproduced for each frame and outputs them to the output terminal 48.
[0062]
According to this configuration, for example, the sampling frequency F of the digital signal input to the input terminal 11₁Is a 96 kHz music signal, when all information of the main code Im, the packet Pe, and in some cases the coefficient code Pc is input to the decoding device 40, the digital signal having a sampling frequency of 96 kHz is faithful to the original signal. Can be played. For example, when a signal with a sampling frequency of 48 kHz is sufficient for some users, the sampling frequency is halved in the down-sampling unit 13, and if the main code Im at that time is provided, a code with a high compression rate can be supplied. it can. That is, the encoding efficiency can be increased. In this case, the decoding device 40 may supply the decoded signal having the second sampling frequency from the decoding unit 43 to the frame synthesis unit 47 as a reproduced digital signal.
[0063]
In this way, it is possible to provide an encoded signal having a quality according to the user's request. However, the high-frequency component has been removed by the downsampling unit 13, and therefore the error signal from the error calculation unit 17 is relatively large. This error signal is supplied to the array conversion / encoding unit 18 as it is for encoding. Then, the amount of information becomes large. However, in the third embodiment shown in FIG. 9, a prediction error signal of an error signal is generated and supplied to the array conversion / encoding unit 18, so that the error signal component is output with a considerably small amount of information. Can do.
[0064]
Sampling frequency F from the error calculator 17₁Error signal of the input signal is down-sampled by the down-sampling unit 13 and the frequency F₁/ 4 of the signal from which the component higher than / 4 is removed is the sampling frequency F of the upsampling unit 16₁Is up-sampled and subtracted from the original input signal. As a result, the low frequency component is removed, the high frequency component remains, and the high frequency component as shown in FIG. Sampling frequency F₁The error signal bandwidth of is F₁/ 2. Therefore, as shown by a broken line in FIG. 9, the frequency axis inversion unit 32 is inserted on the output side of the error calculation unit 17 to thereby generate the frequency F₁The frequency axis is inverted around / 4 so that the low frequency component becomes a large error as shown in FIG. 12B, for example. In order to perform the frequency axis inversion in the time domain, it is only necessary to apply a series in which the polarity is alternately inverted to +1 and −1 to the error signal sample. The error signal that has been frequency axis inverted in this way is supplied to the prediction error generator 31.
[0065]
In the above frequency axis inversion, (-1) for each sample amplitude value of the error signal e (t) to be invertedⁿMultiply (n is an integer value indicating the sample number). For this purpose, the sign of the amplitude value for each sample may be inverted. In that case, the frequency domain coefficient E (f) (f is the frequency) is inverted on the frequency axis and E (F₁/ 2−f). Where F₁Is the sampling frequency of the input signal. Sampling frequency after downsampling is F₁/ 2, and the frequency band targeted for lossy encoding is 0 to F₁Up to / 4, the error signal high frequency (F₁/ 4 to F₁/ 2) is not affected by lossy compression, so the error signal component subjected to frequency axis inversion is low frequency (from 0 to F₁/ 4) is the main. Therefore, the high frequency component is converted into a low frequency component that hardly contributes to randomness. Compression efficiency is improved by reversibly compressing a prediction error obtained by linearly predicting this. Here, a lossless compression code by lossless compression coding is output. The linear prediction coefficient obtained by linear prediction is quantized and a corresponding prediction coefficient code is output.
[0066]
In the decoding apparatus 40, as shown by a broken line, a frequency axis inversion unit 57 is inserted after the prediction synthesis unit 56 and the frequency axis is inverted in the time domain by the same method as the frequency axis inversion unit 32 described above. The error signal spectrum shown in FIG. 12B is converted into the error signal spectrum shown in FIG. 12A, that is, supplied to the adder 46 as an error signal similar to the error signal from the error calculator 17 of the encoding device 10.
[0067]
That is, on the decoding side, the input lossless compression code Pe is losslessly decoded by the decoding / array inverse transform unit 45 to decode the prediction error Spe, and the prediction coefficient code Ic separated by the input unit 42 is handled by the coefficient decoding unit 56E. Play the prediction coefficient LPC. A prediction signal is obtained by performing linear prediction with the prediction coefficient LPC in which the prediction error is reproduced, and the prediction signal is inverted by the time axis inversion unit 57 to reproduce the error signal. In frequency axis inversion, (-1) for each sample amplitude value of the prediction signal p (t) that is also the inversion targetⁿMultiply (n is an integer value indicating the sample number). For this purpose, the sign of the amplitude value may be inverted for each sample. The frequency domain coefficient P (f) (f is the frequency) is inverted on the frequency axis and P (F₁/ 2−f). In that case, the prediction signal is low frequency (0 to F₁/ 4) is the main error signal obtained by frequency axis inversion.₁/ 4 to F₁/ 2) is the main.
In this way, it has been experimentally confirmed that when the error signal with the sampling frequency increased is frequency axis inverted to generate the prediction error signal, the performance is better than when the frequency axis inversion is not performed.
[0068]
Fourth embodiment
FIG. 13 shows a fourth embodiment of the present invention, in which parts corresponding to those in FIG. 9 are given the same reference numerals. The difference from FIG. 9 is that in the encoding device 10, the error signal supplied to the prediction error generating unit 31 is converted into a third sampling frequency F by the downsampling unit 33._ThreeIn other words, the error signal is converted into the error signal, that is, the sampling frequency is lowered and supplied to the prediction error generation unit 31. This third sampling frequency F_ThreeIs the second sampling frequency F₂Is preferably equal to In this case, the error signal supplied to the downsampling unit 33 is supplied with the frequency axis inversion unit 32 being inverted in frequency axis.
[0069]
  In the prediction error generation unit 31, for example, the error signal from the downsampling unit 33 input as shown in FIG.AThe linear prediction analysis is performed using the linear prediction coefficient, and the error signal from the downsampling unit 33 is calculated using the linear prediction coefficient.linearForecastMeasuring sectionThe prediction signal obtained by processing at 31C is converted into an integer value by the integer converting unit 31D. In this case, the upsampling unit 31F converts the predicted signal, which is an integer, into the first sampling frequency F.₁The first sampling frequency F₁The difference between the predicted signal and the error signal from the frequency axis inverting unit 32 is taken by the subtracting unit 31E and supplied to the array conversion / encoding unit 18 as a predicted error signal.
[0070]
In the decoding device 40, the configuration of the prediction synthesis unit 56 is changed, and the first sampling frequency F from the decoding / array inverse transform unit 45 is changed.₁The reproduction prediction error signal of the third sampling frequency F is downsampled by the downsampling unit 56F._ThreeThe prediction error signal is converted into a prediction error signal, and this signal is subjected to a convolution operation with the linear prediction coefficient decoded by the coefficient decoding unit 56E and the prediction signal generation unit 56A to generate a prediction signal, which is converted to an integer value by the integerization unit 56D . This integer prediction signal is supplied to the first sampling frequency F by the upsampling unit 56G.₁Is added to the reproduction prediction error from the decoding / array inverse transform unit 45 in the adder 56B to obtain an error signal. This error signal is frequency axis inverted by the frequency axis inverter 57 and supplied to the adder 46.
[0071]
As the prediction error generation unit 31 of the encoding device 10, the one shown in FIG. 10A may be used. In this case, it is natural that the upsampling unit 31F is provided on the output side of the integer conversion unit 31D. . When this change is made, the predictive synthesis unit 56 of the decoding device 40 uses the configuration shown in FIG. 11A, and a downsampling unit 56F is provided on the signal input side of the predicted signal generation unit 56A. 56G is provided on the output side of the integerizing unit 56D.
Thus, by generating the prediction error signal by lowering the sampling frequency of the error signal, the error signal becomes only a low-frequency component, that is, a component having mainly a large level in the error signal shown in FIG. 12B, and this band is narrow. Since a prediction error signal of the signal is generated, the calculation processing amount is so small or the accuracy of the obtained prediction signal is high.
[0072]
In each of the above-described embodiments, the encoding device 10 and the decoding device 40 can also function by causing the computer to execute an encoding program and a decoding program, respectively. In these cases, the lossless encoding program and the lossless decoding program are downloaded to the program memory of the computer from a CD-ROM, a flexible magnetic disk, etc. or through a communication line.
As described above, according to the third and fourth embodiments of the present invention, if the main code Im can be correctly decoded and the error signal can be correctly reproduced, a high-quality signal having a high sampling frequency and a high band can be reproduced. Even if an error signal is not obtained or cannot be reproduced satisfactorily, a relatively high quality signal can be reproduced by decoding the main code. If a user does not require a signal with such a high quality, the encoding efficiency can be improved by providing only the main code Im. Moreover, by providing an error signal to a user who requests an ultra-high quality signal, the request can be satisfied. At that time, the error signal is provided as a prediction error signal to increase the encoding efficiency. be able to.
[0073]
Example 5
2D hierarchy
In the first to fourth embodiments described above, the same as the original sound signal, which represents an error between the sign (main code Im) of the signal obtained by down-sampling the input digital signal to a lower sampling frequency and the original sound signal resulting from the encoding. Outputs error code Pe at the sampling frequency and allows the user to select whether to use only the main code Im for decoding or to use both the main code Im and the error code Pe for decoding according to the required quality An example was given. That is, these embodiments show a case where a signal having a sampling frequency that is two layers is used as a signal to be encoded.
[0074]
Hereinafter, in this fifth embodiment, a signal is a combination of M kinds of amplitude resolutions of the sample (also called amplitude word length or quantization accuracy, expressed in bits) and N kinds of sampling frequencies (sampling rates). It is possible to encode and generate digital signals in all layers by forming an M × N two-dimensional layer. FIG. 14A shows an example of a combination of encoded digital signals in the case of two-dimensional hierarchical encoding of the digital signals. In this example, 3 × 3 hierarchies are shown with M = 3 kinds of amplitude word lengths of 16 bits, 20 bits, and 24 bits and N = 3 kinds of sampling frequencies of 48 kHz, 96 kHz, and 192 kHz. In FIG. 14A, the amplitude word length (number of bits) is shown in the downward direction with the most significant bit MSB of the sample word as a reference, and the sampling frequency is shown in the horizontal direction.
[0075]
As shown in FIG. 14B, for the upper 16 bits excluding the lower 8 bits in a digital signal having an amplitude word length of 24 bits, a code A encoded at a sampling frequency of 48 kHz and a component higher than the component encoded by the code A The frequency component is hierarchized into a code B encoded at a sampling frequency of 96 kHz and a code C encoded at a sampling frequency of 192 kHz for a frequency component higher than the component encoded by the code B.
For a 20-bit word length signal in which the lower 4 bits are further added to the 16-bit word length, the lower 4-bit component, that is, the residual component obtained by subtracting the 16-bit word length component from the 20-bit word length signal, The residual signals having sampling frequencies of 48 kHz, 96 kHz, and 192 kHz are hierarchized into codes D, E, and F, respectively. For a 24-bit word length signal in which the lower 4 bits are added to the 20-bit word length, the lower 4-bit component, that is, the residual component obtained by subtracting the 20-bit word length component from the 24-bit word length signal is sampled. The residual signals of frequencies 48 kHz, 96 kHz, and 192 kHz are hierarchized into codes G, H, and I, respectively. That is, for a signal having a word length of 16 bits or more, code hierarchization is performed for each sampling frequency.
[0076]
As described above, by using the codes A to I encoded according to the two-dimensional hierarchical nine kinds of coding conditions of the amplitude word length (amplitude resolution, quantization accuracy) and the sampling frequency, three kinds of amplitude word lengths are obtained. Nine types of digital signals, which are all combinations of the three types of sampling frequencies, can be output. Generally, a digital signal hierarchized into M × N types can be generated by a combination of M types of amplitude word lengths and N types of sampling frequencies. That is, the codes shown in FIG. 15 may be used for each combination of sampling frequency and amplitude word length. For example, in the case of a sampling frequency of 96 kHz and an amplitude word length of 24 bits, codes A, B, E, and H may be used.
[0077]
  Next, an encoding method for generating these codes A to I will be described with reference to the functional configuration shown in FIG. In the following description of the embodiments, the first, second,..., Mth amplitude resolution is called in order from the lowest resolution with respect to the M types of amplitude resolution, and any one of them is referred to as the mth amplitude resolution. This is called resolution. m is 1 ≦ m ≦MAn integer in the range Similarly, with respect to N types of sampling frequencies, the first, second,... n is an integer in the range of 1 ≦ n ≦ N. Further, a digital signal having an mth amplitude resolution and an nth sampling frequency is referred to as an mth and nth digital signal.
[0078]
The mth and nth digital signals S of the original sound of the combination of the sampling frequency and the amplitude word length required to generate the codes A to I_{m, n}Is m, n sound source 60_{m, n} M = 1, 2, 3 corresponding to the m-th amplitude word length (quantization accuracy). In this example, m = 1 is 16 bits, m = 2 is 20 bits, m = 3 is a case of 24 bits, and n = 1, 2, 3 corresponding to the nth sampling frequency (sampling rate). In this example, n = 1 is 48 kHz, n = 2 is 96 kHz, and n = 3 is 192kHz.
If a digital signal of a certain condition is not prepared, it is created from the higher-order digital signal. At least the third and third digital signal S_3,3That is, a digital signal source 60 with an amplitude word length of 24 bits and a sampling frequency of 192 kHz_3,3The third and third digital signals S are prepared_3,3From other sources 60 by downsampling or truncating the lower bits (lower 4 bits or 8 bits in this example)_{m, n} A digital signal (m ≠ 3, n ≠ 3) is generated.
[0079]
1st, 1st sound source 60_1,1 1st and 1st digital signal S_1,1Is the first and first compressor 61_1,1The first and first codes A are generated and output. The first and first digital signals S_1,1Is the accuracy converter 62_1,1 Therefore, the first quantization accuracy is converted to a second quantization accuracy higher than this. For example, the first and first digital signals S_1,1Is a code absolute value expression, 0 is added to the lower order, and in this example, 4 bits are added._2,1 2nd and 1st digital signal S_2,1Are the second and first precision conversion signals having the same quantization accuracy (same amplitude word length). Second and first sound source 60_2,1 2nd and 1st digital signal S_2,1Is the second and first subtractor 63_2,1The second and first precision conversion signals are subtracted by the second and first error signals Δ_2,1Is generated and this second and first error signal Δ_2,1Is the second and first compression part 61_2,1 The second and first codes D are generated and output.
[0080]
The first and first digital signals S_1,1Is the first and first upsampling section 64_1,1Thus, the sampling frequency is converted and generated into first and second upsampling signals having a second sampling frequency higher than the first sampling frequency. In this example, the sampling frequency is converted from 48 kHz to 96 kHz. For example, as shown in FIG._1,1A sample indicated by a broken line is inserted in the middle between adjacent samples in the sample row indicated by a solid line. The samples indicated by the broken lines are as close as possible to the samples when the original signal is sampled at the second sampling frequency to obtain a digital signal having the first amplitude word length. For example, as shown in FIG. 17B, the first and first digital signals S_1,1Are sequentially delayed by the delay units D1 and D2 in the sampling period, and the input samples and the output samples of the delay unit D2 are multiplied by the weights W1, W2, and W3 by the multipliers 641, 642, and 643, respectively, and the multiplication results are obtained. Sample US to add and insert in the adder 644₁Get. That is, the first and first digital signals S_1,1On the other hand, for example, linear interpolation is performed by the interpolation filter shown in FIG.₁Is generated.
[0081]
First and second sound source 60_1,21st and 2nd digital signal S_1,2Is the first and second subtractor 63_1,2 1st and 2nd upsampling signal US₁Is subtracted from the first and second error signal Δ_1,2Is generated. This first and second error signal Δ_1,2Is the first and second compression part 61_1,2The first and second codes B are generated and output.
Symbol E is the first and second sound sources 60._1,21st and 2nd digital signal S from_1,24 bits "0" in the lower order of the first and second precision converter 62_1,2To generate a second and second precision converted signal having an amplitude word length of 20 bits, and the second and second precision converted signal is converted into a second and second subtractor 63._2,2In the second and second sound source 60_2,22nd and 2nd digital signal S from_2,2Subtracted from the second and second error signal Δ_2,2And the second and second error signal Δ_2,2The second and second compression parts 61_2,2Obtained by compression encoding.
[0082]
Symbol H is the third and second sound source 60._3,23rd and 2nd digital signal S from_3,2And second and second sound source 60_2,22nd and 2nd digital signal S from_2,2The error signal Δ with the precision-converted signal_3,2Obtained by compression encoding. Symbol C is the first and third sound sources 60._1,31st and 3rd digital signal S from_3,1And the first and second sound source 60_1,21st and 2nd digital signal S from_1,2Up-sampled signal US₂The first and third error signal Δ_1,3Obtained by compression encoding. Symbol F is the second and third sound source 60._2,32nd and 3rd digital signal S from_2,3And the first and third sound source 60_1,31st and 3rd digital signal S from_1,3The error signal Δ with the precision-converted signal_2,3Obtained by compression encoding. Similarly, the symbol I is the third and third sound sources 60._3,33rd and 3rd digital signal S from_3,3And the second and third sound source 60_2,32nd and 3rd digital signal S from_2,3The error signal Δ with the precision-converted signal_3,3Obtained by compression encoding.
[0083]
The generation of these codes A to I will be generally described. Only for the combination of m = 1 and n = 1, the first and first sound sources 60_1,1 To 1st and 1st digital signal S_1,11st and 1st compression part 61_1,1The first and first codes A are generated by compression encoding. For combinations of m and n in the range of 1≤m≤M-1,1≤n≤N, the m, nth digital signal S_{m, n}Is converted to an m + 1, n-th accuracy conversion signal having an m + 1-th quantization accuracy higher than the m-th quantization accuracy._{m, n}To generate m + 1, n sound source 60_{m + 1, n}M + 1, n digital signal S_{m + 1, n}M + 1, n precision conversion signal to m + 1, n subtraction unit 63_{m + 1, n} And the remainder is m + 1, n error signal Δ_{m + 1, n}The m + 1, n error signal Δ_{m + 1, n}M + 1, n compressor 61_{m + 1, n} To generate the (m + 1) th code.
[0084]
For m and n combinations in the range of m = 1, 1 ≦ n ≦ N−1, the m, n upsampling unit 64 converts the m, n digital signal to the (n + 1) th sampling frequency higher than the nth sampling frequency._{m, n}Upsampling to generate the m, n + 1 upsampling signal, and the m, n + 1 sound source 60_{m, n + 1}The m, n + 1th upsampling signal is converted from the m, n + 1th digital signal to the mth, n + 1th subsampling unit 63._{m, n + 1}And the remainder is m, n + 1 error signal Δ_{m, n + 1}This m, n + 1 error signal Δ_{m, n + 1}M, n + 1 compression unit 61_{m, n + 1}To generate the mth, n + 1th code.
1st and 1st digital signal S_1,1Since the energy bias is large, the first and first compression parts 61_1,1For example, predictive coding, transform coding, or compression coding combined with high compression coding is possible. As a specific example, FIG. 18A shows an example of a lossless compression encoder capable of variable compression. This technique is shown, for example, in Japanese Patent Application Publication No. 2001-144847.
[0085]
As shown in FIG. 18A, in the encoder 61, the time series of the input digital signal is sequentially divided into frame units composed of, for example, 1024 digital signals (that is, samples of 1024 points) by the frame dividing unit 61A. The digital signal is irreversibly compressed and encoded by the irreversible quantization unit 61B for each frame unit. This encoding may be any system suitable for the input signal as long as the original digital input signal can be reproduced to some extent at the time of decoding. For example, as described above, if the digital input signal is voice, ITU-T voice coding can be used, and if it is music, MPEG-4 AUDIO option TwinVQ can be used. An encoding method can also be used. The lossy compression code I (n) is locally decoded by the inverse quantization unit 61C, and an error signal between the local signal and the original digital signal is generated by the difference circuit 61D. However, the error signal can be obtained by using the quantized signal obtained by performing the irreversible quantization in the irreversible quantization unit 61B, for example, as described for the local decoding unit 15 in FIG. The quantization unit 61C can be omitted. This error signal represents a quantization error by the irreversible quantization unit 61B, and its amplitude is usually much smaller than the amplitude of the original digital signal. Therefore, the amount of information can be reduced by lossless compression encoding of the quantized error signal rather than lossless compression encoding of the digital signal.
[0086]
In order to further increase the efficiency of the lossless compression encoding, the error signal, that is, the digital sample string is array-converted by the array converter 61E. The processing of the array conversion unit 61E is the same as that described with reference to FIG. However, here, a case is shown in which the conversion of all bits is performed without detecting the significant digits of the sample. A bit in the frame is extracted by extracting the bits in the frame across the samples at the same bit position of each sample of the quantization error signal from the difference circuit 61D, that is, for each MSB, second MSB,. The lossless encoding unit 61F performs lossless compression encoding and outputs the code I (e) together with the quantization code I (n) by the irreversible quantization unit 61B.
[0087]
First and second compression part 61_1,2, First and third compression unit 61_1,3As the first and second error signals Δ_1,2, First and third error signal Δ_1,3Since there is energy only in the upper half of the frequency band, compression encoding may be performed after predicting a signal or performing conversion such as processing in the array conversion unit 61E in FIG. 18A. Compression unit 61_2,1, 61_3,1, 61_2,2, 61_3,2, 61_2,3And 61_3,3For example, an encoder obtained by removing the lossy quantization unit 61B, the inverse quantization unit 61C, and the difference circuit 61D from the encoder in FIG. 18A, that is, the lossless encoder 61 shown in FIG. 19A can be used. These compression parts 61_2,1, 61_3,1, ..., 61_2,3, 61_3,3When the error signal input to is sufficiently small, it becomes close to noise and large compression cannot be expected. Therefore, compression encoding may be performed to a code representing only 0 in this frame.
[0088]
First and first upsampling unit 64_1,1First and second upsampling unit 64_1,2If the number of taps of the interpolation filter used in (the number of multipliers in FIG. 17B, 3 in the example of FIG. 17B) is not known in advance on the decoding side, auxiliary information indicating the number of taps is shown in FIG. As indicated by the broken line, the auxiliary information encoding unit 65_1,2, 65_1,3Thus, the first and second auxiliary codes and the first and third auxiliary codes are output in association with the first and second codes and the first and third codes, respectively. An example of the number of taps of the interpolation filter and the auxiliary code is shown in FIG. 20A. As the number of taps of the interpolation filter, a large value is selected when high-precision decoding is performed on the decoding side, and a small value is selected when high-precision decoding is not required. The tap coefficient may be fixed, in which case there is no need to send an auxiliary code.
[0089]
Next, a decoding apparatus corresponding to the encoding apparatus in FIG. 16 will be described with reference to FIG. 1st, 1st code A, 2nd, 1st code D, 3rd, 1st code G, 1st, 2nd code B, 2nd, 2nd code E, 3rd, 2th code H, 1st, 3rd code C, 1st The second and third codes F and the third and third codes I are the first and first decompressors 80, respectively._1,1, 2nd and 1st extension 80_2,13rd, 1st extension 80_3,1, First and second extension 80_1,2, 2nd and 2nd extension 80_2,23rd, 2nd extension 80_3,21st, 3rd extension 80_1,3, Second and third extension 80_2,3And the third and third extension 80_3,3And decompression decoding is performed respectively. As a result, the first and first digital signals S_1,1And error signal Δ_2,1, Δ_3,1, Δ_1,2, Δ_2,2, Δ_3,2,, Δ_1,3, Δ_2,3, Δ_3,3Is obtained. It should be noted that the m-th and n-th expansion units 80 other than m = 1 and n = 1_{m, n}, The mth and nth error signals Δn of the mth quantization accuracy and the nth sampling frequency_{m, n}Are decompressed and decoded. These mth and nth extension parts 80_{m, n}Is the corresponding m-th and n-th compression unit 61._{m, n}The m-th and n-th codes that have been compression-encoded in (1) are decompressed and decoded.
[0090]
For m and n pairs in the range of 1 ≦ m ≦ M-1 and 1 ≦ n ≦ N, the mth and nth expansion sections 80_{m, n}The mth and nth digital signals S of the mth quantization accuracy and the nth sampling frequency that are decompressed by_{m, n}The mth and nth precision conversion unit 81_{m, n}The quantization accuracy (amplitude word length) is converted into an (m + 1) -th and n-th accuracy conversion signal of (m + 1) -th quantization accuracy, nExtension part 80_{m + 1, n}M + 1, n error signal Δ_{m + 1, n}The m + 1, n adder 82_{m + 1, n}The (m + 1) th digital signal S with the (m + 1) th quantization accuracy (amplitude word length) and the nth sampling frequency is added._{m + 1, n}Is played.
For example, the first and first extension parts 80_1,1Decompressed first and first digital signals S_1,1Is the first and first precision converter 81_1,1In this way, the second and first precision converted signal with 0 added to the lower order and the amplitude word length of 20 bits is generated. This second and first precision converted signal is sent to the second and first decompression unit 80._2,1The second decoded error signal Δ_2,1And the second and first adder 82_2,1Is added to the second and first digital signals S_2,1Is played.
[0091]
Also, for m, n pairs in the range of m = 1, 1 ≦ n ≦ N−1, the first, n extension unit 80_{1, n}1st and nth digital signal S_{1, n}Is the first, n upsampling unit 83_{1, n}The first and n + 1 upsampling signals are converted into the first and n + 1 upsampling signals having the (n + 1) th sampling frequency._{1, n + 1}1st quantization accuracy, 1st, 1st + 1 error signal Δ of 1st + 1 sampling frequency_{1, n + 1}First, n + 1 adder 82_{1, n + 1}And the first and n + 1 digital signals S having the first quantization accuracy and the (n + 1) th sampling frequency._{1, n + 1}Is played.
For example, the first and first extension parts 80_1,1Decompressed first and first digital signals S from_1,1Is the first and first upsampling unit 83_1,1Thus, the first sampling frequency is converted into the first and second upsampling signals converted into the second sampling frequency. The first and second upsampling signals are supplied to the first and second expansion units 80._1,2The first and second error signals Δ_1,2First and second adder 82_1,2The first and second digital signals S_1,2Is played.
[0092]
Upsampling unit 83_1,1, 83_1,2When the number of taps of the interpolation filter used in the above is not known in advance, the first, second auxiliary code and the first, third auxiliary code input in association with the first, second code B and the first, third code C respectively. Auxiliary code decoder 85_1,2, 85_1,3The number of taps as auxiliary information is decoded by the upsampling unit 83 corresponding to each tap number._1,1, 83_1,2Set to
1st, 1st extension 80_1,1Is the first and first compression units 61 in the encoding device of FIG._1,1For example, the compression unit 61_1,1When the encoder 61 shown in FIG. 18A is used, the expansion unit 80_1,1The decoder 80 shown in FIG. 18B is used.
[0093]
In the decoder 80, the lossless compression unit 80A decodes the lossless compression code I (e), and the array inverse transform unit 80B performs a process reverse to the array transform unit 61E in the encoder 61 (for the decoded signal). For example, a process of array conversion from an equidistant bit string to an amplitude bit string, which is the reverse of the process described with reference to FIGS. In addition, the irreversible compression code I (n) is decoded by the inverse quantization unit 80B, the decoded signal and the reproduced quantization error signal are added by the addition unit 80D, and finally, the frame synthesis unit 80F. The added signals for each frame are sequentially connected to reproduce the original digital signal sequence.
[0094]
That is, the lossless compression code I (e) in the first and first codes A is losslessly decoded, and a plurality of samples representing the code absolute value of the bit string consisting of the corresponding bit position in the frame from the decoded bit string are quantized in the frame. Reproduced as an error signal, irreversibly decodes the irreversible compression code I (n) in the first and first codes A to generate a local reproduction signal, and adds the reproduction signal and the quantization error signal to the first , 1 Digital signal S_1,1Play.
Extension part 80_1,2, 80_1,3Is the compression part 61_1,2, 61_1,3A decoding method corresponding to this encoding method can be used, and a predictive decoding method, a transform decoding method, or the like can be used. The decoding method for the encoding method used in the corresponding compression unit is also used for the other decompression units. When the compression unit has the configuration shown in FIG. 19A, the decompression unit has a configuration in which the inverse quantization unit 80C and the addition unit 80D are removed from the decoder 80 in FIG. 18B, that is, the configuration shown in FIG. 19B.
[0095]
According to the configuration of the encoding device shown in FIG. 16, various digital signals having combinations of various amplitude resolutions (amplitude word lengths) and various sampling frequencies (sampling rates) are uniformly encoded as a two-dimensional hierarchy. In addition, it is possible to perform compression coding with high efficiency as a whole, and to provide various combinations of digital signals requested by users with a small amount of data.
According to the configuration of the decoding apparatus shown in FIG. 21, the desired ones in the digital signals of various combinations of quantization accuracy and sampling frequency are unified from the codes encoded by the encoding apparatus shown in FIG. Can be decrypted automatically.
[0096]
Depending on the user, all combinations of the m-th and n-th digital signals S shown in FIG._{m, n}Do not need. First and first decompression units 80 for decoding codes A and B in the decoding apparatus shown in FIG._1,1First and first upsampling unit 83_1,1, First and second extension 80_1,2, First and second adder 82_1,2And a first and first precision conversion unit 81 for decoding the code D or E or C and F_1,1, 2nd and 1st extension 80_2,1And second and first adder 82_2,1Alternatively, the first and second precision converter 81_1,2, 2nd and 2nd extension 80_2,2And second and second adder 82_2,2Alternatively, the first and second upsampling units 83_1,21st, 3rd extension 80_1,3, First and third adder 82_1,3, First and third precision converter 81_1,3, Second and third extension 80_2,3And the second and third adders 82_2,3It may be provided with at least.
[0097]
In each of the embodiments shown in FIGS. 16 and 21, the number M of types of quantization accuracy and the number N of types of sampling frequencies are not limited to three, and the number of layers can be increased or decreased to increase the number of hierarchies. May be reduced. Similarly, for N, the number of hierarchies may be increased or decreased.
Sixth embodiment
The mth and nth digital signals S of various combinations of quantization accuracy and sampling frequency in FIG._{m, n}Sound source 60_{m, n}Is prepared in advance as described above, and the digital signal of each sound source is different from that obtained by simply down-sampling and truncating the lower bits of the m-th and n-th digital signals. Noise (fixed dither signal) may be added, and various conversions and adjustments such as deviations in amplitude and sampling (sample point position) may be performed. It is generally unknown in advance how this conversion or adjustment is.
[0098]
Therefore, in the sixth embodiment, in the encoding apparatus of FIG. 16, a digital signal having a lower amplitude resolution or a lower sampling frequency is converted into an accuracy conversion unit 62._{m, n}Or upsampling section 64_{m, n}Converts to a digital signal of higher amplitude resolution (quantization accuracy, amplitude word length) or higher sampling frequency by subtracting unit 63_{m + 1, n}(Or 63_{m, n + 1}Subtracting part 63_{m + 1, n}(Or 63_{m, n + 1}) Output error signal Δ_{m + 1, n}(Or Δ_{m, n + 1}) Is provided so as to be as small as possible.
For example, as shown in FIG._{m, n}The mth and nth digital signals from the precision conversion unit 62_{m, n}As described above, the m-th quantization accuracy (amplitude word length, amplitude resolution) is converted into the (m + 1) -th quantization accuracy. The m + 1, n accuracy conversion signal that has been level-adjusted by the gain adjustment unit 66A and further level-adjusted (gain) has its sample position adjusted by the timing adjustment unit 66B, and its sample position has been adjusted. The subtracting unit 63 performs subtraction on the m + 1, nth digital signal by the n precision conversion signal.
[0099]
The m + 1, n-th error signal Δ that is the subtraction result from the subtracting unit 63._{m + 1, n}Is branched and input to the error minimizing unit 66C. The error minimizing unit 66C receives the (m + 1) th error signal Δ._{m + 1, n}The level adjustment amount in the gain adjustment unit 66A and the sample position adjustment amount in the timing adjustment unit 66B are controlled so that the amount of information after compression is minimized. For this purpose, the error signal may be compression-coded and the information amount of the obtained error code may be compared. As a simple method for approximating the comparison of the information amount, the power of the error signal may be compared, and the gain and the sample position may be determined so that the power is minimized. In the following embodiments, an example in which the error signal power is minimized will be described. For example, as the level adjustment amount and the sample position adjustment amount, a plurality of predetermined values are respectively stored in a storage unit (not shown) in the error minimizing unit 66C and auxiliary codes indicating those values as shown in FIGS. 20B and 20C, for example. Correspondingly stored as a table, the error signal Δ_{m + 1, n}Each one with the smallest power is selected. An auxiliary code representing the selected level adjustment amount and sample position adjustment amount is output. The predetermined level adjustment amount and the sample position adjustment amount may be stored in association with the auxiliary code by using a set of these values as one table instead of as separate tables.
[0100]
When the power minimization of the error signal is achieved, the m + 1, n compression unit 61_{m + 1, n}Compression instruction signal is output to the (m + 1) th, nth compression unit 61._{m + 1, n}Is the m + 1, nth error signal Δ_{m + 1, n}Is compressed and encoded. Further, the error minimizing unit 66C supplies an auxiliary code representing the level adjustment amount and the sample position adjustment amount at that time to the auxiliary code generation unit 69, and the auxiliary code generation unit 69 inputs the input level adjustment amount and the auxiliary code of the sample position adjustment amount. Are combined and output as the m + 1, n auxiliary code in association with the m + 1, n code.
Similarly, as shown by a broken line in FIG. 22 and in parentheses, the mth and nth digital signals are converted into upsampling units 64._{m, n}Then, the sampling frequency is upsampled to the (n + 1) th sampling frequency, and the level of the m, n + 1th upsampling signal generated thereby is adjusted by the gain adjusting unit 66A as in the previous case, and further the timing adjusting unit In 66B, the sample position is adjusted and supplied to the subtracting unit 63. The subtracting unit 63 outputs the m, n + 1th digital signal S._{m, n + 1}The m, n + 1 upsampling signal is subtracted from the m, n + 1 error signal Δ_{m, n + 1}The gain adjusting unit 66A and the timing adjusting unit 66B are adjusted by the error minimizing unit 66C so that the power of the error adjusting unit 66C is minimized. The minimized m, n + 1 error signal Δ_{m, n + 1}Is the m, n + 1 compression unit 61._{m, n + 1}The auxiliary code corresponding to the position amount is encoded by the auxiliary code generation unit 65 and is associated with the m, n + 1 code as the m, n + 1 auxiliary code. Is output. Mth and nth upsampling unit 64_{m, n}When the number of taps of the interpolation filter is output, the auxiliary information encoding unit 65 also encodes the number of taps of the interpolation filter as the m, n + 1 auxiliary code.
[0101]
FIG. 20B shows a correspondence example between the auxiliary code and the adjustment gain, and FIG. 20C shows a correspondence example between the auxiliary code and the adjustment sample position amount (sample point movement amount). For example, as shown in FIG. 20D, these auxiliary codes are arranged in the order of presence / absence code C11 indicating the presence / absence of auxiliary information, gain code C12, movement amount code C13, and tap number code C14, and are referred to as m, n + 1 auxiliary codes. The In FIG. 22, the gain adjusting unit 66A and the timing adjusting unit 66B may be interchanged. In other words, either the gain adjustment or the time shift adjustment may be performed first. Alternatively, either the gain adjustment unit 66A or the timing adjustment unit 66B may be omitted. Further, the generation of such an auxiliary code in the error minimizing unit 66C may be performed for each frame. For example, when a fixed dither signal is added to the m-th and n-th digital signals, and this is known in advance, the fixed dither signal is derived from the m + 1, n-precision conversion signal or the m, n + 1-th upsampling signal. Subtract and subtract 63_{m, n + 1}(Or 63_{m, n + 1}And the fixed dither signal is encoded and output as the (m + 1) th, n-th auxiliary code.
[0102]
As described above, when the encoding apparatus adjusts the lower digital signal, specifically, for example, the (m + 1, n) th precision conversion signal, the decoding apparatus is provided with an adjustment unit, and the auxiliary code is decoded into the auxiliary information. Similarly, it is necessary to adjust the accuracy conversion signal. An example in that case is shown in FIG. This is a case where the precision conversion signal for the m, n digital signal is adjusted by the adjustment unit 87, and the m + 1, n auxiliary code associated with the m + 1, n code is decoded by the auxiliary information decoding unit 88. Then, the obtained auxiliary information, in this example, the adjustment gain and the sample position amount are generated and supplied to the deformation control unit 87C of the adjustment unit 87.
[0103]
On the other hand, the m-th and n-th digital signals subjected to decompression decoding are converted into the m-th and n-th precision converter 81._{m, n} Is converted into an (m + 1) th and nth accuracy conversion signal having the (m + 1) th quantization accuracy. Addition unit 82 sequentially_{m + 1, n}Configured to be supplied to. The gain adjusted in the gain adjustment unit 87A is set by the deformation control unit 87C, and the delay time corresponding to the decoded sample position amount is set in the timing adjustment unit 87B by the deformation control unit 87C. Therefore, the (m + 1) th and (n) th accuracy conversion signals are the level adjusted by the corresponding gain adjusting unit 66A and timing adjusting unit 66B (FIG. 22) in the encoding device, the same level as the adjusted sample position, and the same sample. The same deformation as the encoding side is performed. Thus, the (m + 1) th and n-th accuracy conversion signals whose levels and sample positions are adjusted are converted into the (m + 1) th and n-th decompression unit 80._{m + 1, n}The decoded m + 1, n error signal Δ_{m + 1, n}Adder 82_{m + 1, n}Is added. Therefore, adder 82_{m + 1, n}M + 1, n digital signal S reproduced from_{m + 1, n}Is the m + 1, n excitation 60 in the encoding device._{m + 1, n}M + 1, n digital signal S_{m + 1, n}Will be the same.
[0104]
The m, n + 1 auxiliary code associated with the m, n + 1 code is also used when the m, n + 1 digital signal is reproduced by using the upsampled m, n digital signal. Is input to the up-sampling unit 83, as shown by the broken line and parentheses in FIG._{m, n}Is converted into an m, n + 1 upsampling signal of the (n + 1) th sampling frequency, and this m, n + 1 upsampling signal is sequentially passed through the gain adjusting unit 87A and the timing adjusting unit 87B and added. Part 82_{m, n + 1}Supplied to. The m, n + 1 auxiliary code is decoded by the auxiliary information decoding unit 88, and the decoded adjustment gain, the gain according to the sample position amount, the delay time are changed by the deformation control unit 87C, and the gain adjustment unit 87A and the timing adjustment unit 87B. Respectively. Accordingly, the m, n + 1th error signal Δ obtained by decompressing the mth, n + 1th upsampling signal whose level is adjusted and the sample position is adjusted._{m, n + 1}Adder 82_{m, n + 1}And the m, n + 1th digital signal S is added._{m, n + 1}Is played.
[0105]
The gain adjustment unit 87A and the timing adjustment unit 87B may be interchanged. One of the gain adjustment unit 87A and the timing adjustment unit 87B may be omitted. If there is a fixed dither signal as the decoded information of the auxiliary code, it is subtracted from the m + 1, n precision conversion signal or the m, n + 1 upsampling signal.
The encoding apparatus and the encoding method shown in FIG. 22 and the decoding apparatus and the decoding method shown in FIG. 23 itself constitute an embodiment of the invention, whereby the quantization accuracy prepared in advance is set. Therefore, it is possible to perform lossless compression coding with high efficiency and high accuracy with respect to digital signals of at least two sound sources having various combinations of sampling frequencies, and lossless decoding of the codes with high accuracy.
[0106]
The encoding apparatus and the encoding method shown in FIG. 22 may have a two-dimensional multi-hierarchical configuration of quantization accuracy and sampling frequency like the encoding apparatus and the encoding method shown in FIG. Similarly, the decoding apparatus and the decoding method shown in FIG. 23 can have a two-dimensional multi-hierarchical structure as shown in FIG.
The coding apparatus shown in FIGS. 16 and 22 and the decoding apparatus shown in FIGS. 21 and 23 may be configured to cause a computer to function by executing a program. In this case, for example, with regard to a decoding device, a decoding program may be downloaded into a computer from a recording medium such as a CD-ROM or a magnetic disk or through a communication line, and the decoding program may be executed by the computer. .
[0107]
In order to explain the effect of the present invention, for example, three types of musical sound distribution forms shown in FIG. 24 are compared. In other words, in order to meet the different requirements of sampling frequency and quantization accuracy (amplitude resolution),
A. A music signal is encoded and stored by scalable encoding to which the present invention is applied. That is, for example, code sequences A to I shown in FIG. 14A are prepared. In response to a request from the client terminal, these codes are selected and combined and transmitted to the client terminal.
B. An encoded sequence of a combination corresponding to a request from the client terminal for each signal of a combination of a plurality of sampling frequencies and a plurality of quantization accuracy, for example, nine types of sound source signals in FIG. Select one of them according to the request and transmit it to the client terminal.
C. Stores only the compression code of the signal with the highest sampling frequency and the highest quantization accuracy, and transmits to the client terminal after decoding, sampling frequency conversion, quantization accuracy conversion, and re-encoding upon request from the client terminal .
[0108]
On the client terminal side, the received code sequence is decoded, and in the form A using the present invention, the digital signal is reconstructed by performing upsampling and accuracy conversion. In forms B and C, the decoded signal is immediately reconstructed.
In the form B, the capacity of the compression code string in the server becomes considerably large, and in the form C, the calculation amount becomes large. In the form A using the present invention, for example, as shown in FIG. 15, the compression code with the highest sampling frequency and the highest amplitude resolution includes the compression code with the low sampling frequency and the low amplitude resolution. It can be used and the total capacity can be reduced.
[0109]
Although the present invention is applied to a music digital signal in the above description, it can also be applied to an image digital signal.
As described above, according to the fifth and sixth embodiments, encoding with different amplitude accuracy and sampling rate requirements, particularly reversible encoding, can be performed in a unified manner, and the compression rate of the entire system can be increased. .
Example 7
The seventh embodiment of the present invention will be described below. Also in this embodiment, for example, three kinds of quantization precisions of 16 types, 20 bits, and 24 bits as M kinds of quantization precisions, and three kinds of arbitrary quantization precisions and sampling frequencies of 48 kHz, 96 kHz, and 192 kHz as N kinds of sampling frequencies. An example of a two-dimensional hierarchical encoding of a digital signal that enables generation of a digital signal of a combination of the above will be described.
[0110]
FIG. 25 shows an example of decomposition of a 24-bit 192 kHz digital signal for two-dimensional hierarchical coding in the seventh embodiment into signals of each hierarchy and each code. For the upper 16 bits excluding the lower 8 bits in the digital signal having an amplitude word length of 24 bits, the sampling frequency is encoded for the code A encoded with a sampling frequency of 48 kHz and the frequency component higher than the component encoded by the code A. A code B encoded as 96 kHz and a code C encoded with a sampling frequency of 192 kHz for a higher frequency component than the component encoded by the code B are hierarchized in the sampling frequency direction.
[0111]
For a 20-bit word length signal obtained by adding the lower 4 bits to the 16-bit word length, the lower 4-bit component, that is, the residual component obtained by subtracting the 16-bit word length component from the 20-bit word length signal is similarly applied. A code D encoded with a sampling frequency of 48 kHz, a code E encoded with a sampling frequency of 96 kHz for a frequency component higher than the encoded component of the code D, and a frequency component higher than the component encoded with the code E It is hierarchized into a code F encoded with a frequency of 192 kHz. For a 24-bit word length signal including the lower 4 bits with respect to the 20-bit word length, the lower 4-bit component, that is, the residual component obtained by subtracting the 20-bit word length component from the 24-bit word length signal is similarly sampled. A code G encoded as 48 kHz, a code H encoded with a sampling frequency of 96 kHz for a frequency component higher than the component encoded by the code G, and a sampling frequency for a frequency component higher than the component encoded with the code H Is hierarchized into a code I encoded as 192 kHz.
[0112]
Thus, by using the codes A to I encoded according to the two-dimensional hierarchical M × N encoding conditions of the amplitude word length (amplitude resolution, quantization accuracy) and the sampling frequency, M types of quantization are used. It is possible to output M × N types of digital signals that are all combinations of accuracy and N types of sampling frequencies. That is, the usage code (1) shown in FIG. 26 may be used for each combination of each sampling frequency and amplitude word length. For example, in order to encode a digital signal having a sampling frequency of 96 kHz and a quantization accuracy of 24 bits, codes A, B, D, E, G, and H may be used.
As described above, in this embodiment, a digital signal having a quantization accuracy of 16 bits and a sampling frequency of 48 kHz is basically encoded, and a difference from a signal having a lower quantization accuracy or a lower sampling frequency is obtained for a signal higher in the hierarchy. Since the signal component is encoded, the signal having the m-th quantization accuracy and the n-th sampling frequency can be expressed by a simple combination of codes as shown in the used code example (1) in FIG.
[0113]
FIG. 27 shows a functional mechanism of the encoding apparatus that performs the two-dimensional hierarchical encoding shown in FIGS. Each compression unit 61 in FIG._{m, n}The input signal is a hierarchical signal obtained by hierarchically decomposing one original sound (in this example, a digital signal having an amplitude word length of 24 bits and a sampling frequency of 192 kHz) into a plurality of types of quantization accuracy and a plurality of types of sampling frequency signals. is there.
First, a digital signal whose amplitude word length is 24 bits from the sound source 60 and whose sampling frequency is 192 kHz is the bit division unit 71. The 24-bit amplitude word length of each sample is the upper 16 bits, the lower 4 bits, and the lower order. It is divided into a plurality of bit sections such as 4 bits. Upper 16 bits are downsampled 72_1,3 The sampling frequency is downsampled to 96 kHz, and the output is further downsampled by 72_1,2 The sampling frequency is downsampled to 48 kHz, and the compression unit 61_1,1 , And lossless compression encoding is performed, and a code A is output. When used as a 16-bit signal, the last four digits of 20 bits are not simply removed, but a small noise such as rounding or dithering may be added. In this case, the error component signal of the produced 16-bit signal and 20-bit signal is separated. The amplitude is not 4 bits but may be 5 to 6 bits, but it is used with the number of bits remaining increased. The other processes are the same, and the following examples are also the same.
[0114]
Downsample section 72_1,2 Output from the upsampling section 73_1,1 Sampling frequency up to 96kHz, upsampling output and downsampling section 72_1,3The difference from the output from is the error signal Δ_1,2As the subtraction part 74_1,2 And the difference signal Δ_1,2Compression part 61_1,2 The lossless compression encoding is performed, and the code B is output.
Downsample section 72_1,3 Output from the upsampling section 73_1,2The sampling frequency is upsampled to 192kHz._1,2And the difference between the divided 16-bit signal from the bit divider 71 and the error signal Δ_1,3As the subtraction part 74_1,3 And its error signal Δ_1,3Compression part 61_1,3 The code C is output after lossless compression encoding.
[0115]
The lower 16 bits of the upper 16 bits of the signal from the dividing unit 71_2,3, 72_2,2 The sampling frequency is set to 48 kHz, and the compression unit 61_2,1 Is subjected to lossless compression encoding and a code D is output, and the downsampling unit 72_2,3 Output and downsampling section 72_2,2Output upsampling part 73_2,1Subtracting unit 74 with upsampling output by_2,2Is the error signal Δ_2,2And its error signal Δ_2,2Compression part 61_2,2 The code E is output by lossless compression encoding with the downsampling unit 72_2,3 Output sampling unit 73_2,2 The subsampling unit 74 subtracts the difference between the upsampling output from the 4-bit signal from the bit division unit 71_2,3 Error signal Δ_2,3And its error signal Δ_2,3Compression part 61_2,3 To perform lossless compression encoding and output code F.
[0116]
Similarly to the above, the least significant 4 bits of the signal from the bit division unit 71 are converted into the downsampling unit 72._3,3 , 72_3,2 , Upsampling part 73_3,1 , 73_3,2Subtracting unit 74_3,2 , 74_3,3 , Compression unit 61_3,1 , 61_3,2 , 61_3,3 To generate and output codes G, H, and I.
In FIG. 27, each upsampling unit performs the interpolation filter processing described in FIGS. 17A and 17B, for example, on the input signal. At this time, the corresponding subtraction unit 74_{m, n + 1}Output error signal Δ_{m, n + 1}The coefficients W1, W2, and W3 are determined so that the power of is minimized.
[0117]
For example, the subtraction unit 74_1,3 Output error signal Δ from_1,3Is a signal with an amplitude word length of 16 bits and a sampling frequency of 192 kHz and a band of 96 kHz, but the amplitude is small, and in particular, the component of 0 to 48 kHz is almost zero. For this reason, the compression unit 61_1,3 For example, a subtractor 74 using a predictive encoder 61 shown in FIG._1,3The linear prediction unit 61A performs linear prediction analysis on the error signal from the output signal, quantizes the obtained linear prediction coefficient, outputs a code Ic corresponding to the quantized value, and uses the prediction coefficient to predict the input error signal. Is generated. The prediction signal is converted into an integer by the integer conversion unit 61B, the difference from the input error signal is obtained as a prediction error signal by the subtraction unit 61C, and the prediction error signal is losslessly compressed and encoded by the lossless compression unit 61D. can do. Other compression units can also be efficiently compressed using predictive coding or the like.
[0118]
As an encoding procedure, in the above, first, each sample of a signal having a quantization accuracy of 24 bits and a sampling frequency of 192 kHz is divided into three signals of 16 bits, 4 bits, and 4 bits, and then divided into layers. The signal with the quantized bit precision is layered at sampling frequencies of 48 kHz, 96 kHz, and 192 kHz. First, the input digital signal is layered at the sampling frequency, and then the amplitude word length of the sample for the error signal of each layer May be divided. That is, for example, as shown in FIG. 29, a digital signal having an amplitude word length of 24 bits and a sampling frequency of 192 kHz from the sound source 60 is first converted into a downsampler 72._ThreeDownsampling to a sampling frequency of 96kHz, and the downsampled signal is upsampled 73₂To upsample to a sampling frequency of 192kHz. The difference between this upsampled signal and the original signal from the sound source 60 is calculated as an error signal Δ₁As the subtraction part 74₁Generate with
[0119]
Downsampling section 72_ThreeDownsampling the output from 72₂Downsampling to a sampling frequency of 48kHz, and the downsampled signal is upsampled 73₁To upsample to 96kHz sampling frequency. This upsampled signal and downsampler 72_ThreeThe difference from the output signal from the error signal Δ₂As the subtraction part 74₂Generate with Subtraction unit 74₁, 74₂Each error signal from downsampler 72₂Each of the output signals of the bit divider 71₁, 71₂, 71_ThreeThen, each sample is divided into upper 16 bits, its lower 4 bits, and further lower 4 bits, and these divided signals are losslessly encoded by the compression unit. The same reference numerals are assigned to the compression units in FIG. 29 corresponding to the compression units in FIG.
[0120]
Each compression unit 61 in FIG._{m, n}The input signal to the signal is hierarchically decomposed into one original sound (in this example, a digital signal with an amplitude word length of 24 bits and a sampling frequency of 192 kHz) into multiple types of amplitude resolution (quantization accuracy) and multiple types of sampling frequency signals This is a hierarchized signal.
Decoding device of seventh embodiment
FIG. 30 shows a functional configuration of the decoding apparatus according to the seventh embodiment. This is to decode M × N = 9 types of digital signals encoded by the encoding device shown in FIG. 27 or 29 and combining M types of quantization accuracy and N types of sampling frequencies. This is the case.
[0121]
Reference symbols A, B,..._1,1, 80_1,2 , 80_1,3 , 80_2,1 , 80_2,2 , 80_{2, Three} , 80_3,1 , 80_3,2, 80_3,3 Respectively, the input layered signal of each compression unit of the encoder is obtained. Each extension 80_{m, n}The decoding method can be performed by the same method as the lossless decoding unit 80A and the array inverse transform unit 80B of the decoder 80 shown in FIG. 18B.
Extension part 80_1,1 The decoded signal is a reproduction signal S of a digital signal having an amplitude word length of 16 bits and a sampling frequency of 48 kHz (hereinafter referred to as a 16b, 48 kHz digital signal)._1,1Is output as the upsampling unit 83_1,1 The sampling frequency is upsampled to 96kHz. This up-sampled signal is expanded by the expansion unit 80._1,2 The decoded error signal Δ_1,2And adder 82_1,2 16b, 96kHz digital signal S_1,2Is output as This 16b, 96kHz digital signal S_1,2Is upsampling part 83_1,2 The sampling frequency is upsampled to 192 kHz, and the expansion unit 80_1,3 The decoded error signal Δ_1,3And adder 82_1,3 16b, 192kHz digital signal S_1,3Is output as Playback 16b, 48kHz digital signal is expanded 80_2,1Decoded error signal Δ_2,1And adder 82_2,1 Is added to the playback 20b, 48kHz digital signal S_2,1Is output as
[0122]
Hereinafter, the digital signals S are combined by combining the decoded hierarchical signals in the same manner as described above._2,2, S_2,3, S_3,1, S_3,2, S_3,3Is played. Adder 82_{m, n}If the two sampling frequencies to be added are different from each other, signals having a lower sampling frequency are added together by matching the sampling frequencies by upsampling. Reference symbol 83 for upsample part_{m, n}N on the right side of the subscript indicates that the nth sampling frequency is upsampled to the (n + 1) th sampling frequency. For example, the right subscript n = 1 indicates that the sampling frequency is upsampled from 48 kHz to 96 kHz, and the subscript n = 2 indicates that the sampling frequency is upsampled from 96 kHz to 192 kHz. In short, a highly accurate signal is reconstructed by upsampling the layered partial signals and concatenating bits in the amplitude direction.
[0123]
When the quality of the decoded signal required on the decoding side does not require high quality (for example, a digital signal with a quantization accuracy of 24 bits and a sampling frequency of 192 kHz), a digital signal of the required quality (quantization accuracy and sampling frequency) Further, signals with higher quantization accuracy and higher sampling frequency can be omitted. For example, with respect to the maximum number of quantization bits 24, a hierarchized signal of the least significant 4 bits component or a hierarchized signal used only for reproducing a signal having a high sampling frequency can be omitted.
For example, when transmitting on a network, if a different packet is assigned to each code A,..., I, and a higher priority information is given to a lower-level (ie, lower-order) code for output, the network Resources can be used effectively. For example, all information can be transmitted in a normal state, but it is sufficient to preferentially transmit only the code A in the event of a network failure or congestion.
[0124]
Example 8
In the eighth embodiment of the present invention, as shown in FIG. 31, the sampling frequency is hierarchized by a 16-bit quantization accuracy signal as in the seventh embodiment, but the hierarchy for the quantization accuracy of 16 bits or more. Conversion is performed for each sampling frequency. In other words, for signals with a quantization accuracy of 20 bits, the signals D, E, and F are generated by encoding the signals of the sampling frequencies 48 kHz, 96 kHz, and 192 kHz for the residual components obtained by subtracting the signal components of the quantization accuracy of 16 bits from this. For a signal with quantization accuracy of 24 bits, codes G, H, and I are generated by encoding signals of sampling frequencies of 48 kHz, 96 kHz, and 192 kHz, respectively, obtained by subtracting a signal component with quantization accuracy of 20 bits. .
[0125]
Using these codes A,..., I, digital signals having various amplitude resolutions (quantization accuracy) and various sampling frequencies can be reproduced. A code used for reproduction of each digital signal is shown as a use code (2) in FIG. For example, a signal having a sampling frequency of 192 kHz and a quantization accuracy of 20 bits includes a code A obtained by encoding a signal having a sampling frequency of 48 kHz and a quantization accuracy of 16 bits, and a code obtained by encoding a signal having a sampling frequency of 96 kHz and a quantization accuracy of 16 bits. B and a code C obtained by encoding a signal having a sampling frequency of 192 kHz and a quantization accuracy of 16 bits.
The sound source 60 (60 in the encoding apparatus of the eighth embodiment shown in FIG._3,324b, 192kHz digital signal S from_3,3An example in which digital signals having various sampling frequencies and amplitude word lengths shown in FIG. 26 are generated and encoded will be described.
[0126]
24b, 192kHz digital signal S_3,3Is the bit divider 71_3,3 Are divided into lower 4 bits and upper 20 bits for each sample, and the compression unit 61 converts the lower 4 bits into a signal._3,3 The signal of the upper 20 bits is generated in the code I by the bit dividing unit 71._2,3 Are subdivided into lower 4 bits and upper 16 bits, and the lower 4 bits of the signal are compressed by the compression unit 61._2,3The subtracting unit 63 for the signal of the upper 16-bit section generated in the code F_1,3Supplied to.
24b, 192kHz digital signal S_3,3Down sample section 72_3,3 Is down-sampled to a signal with a sampling frequency of 96 kHz, and this down-sampled signal is_3,2 , 71_2,2 In the same manner, the signal is sequentially divided into bit sections, and is divided into the least significant 4 bits signal, the next least significant 4 bits signal, and the most significant 16 bits signal._3,2 , 61_2,2 The 16-bit signal of the latter is generated by the subtractor 63._1,2 Supplied to.
[0127]
Downsample section 72_3,2 The 24b, 96kHz digital signal with a sampling frequency of 96kHz is downsampled 72_3,2In addition, the 24b, 48kHz digital signal is further down-sampled to a signal with a sampling frequency of 48kHz._3,1 , 71_2,1 In the same manner, the signal is sequentially divided into bit sections, and the least significant 4 bits signal, the next least significant 4 bits signal, and the most significant 16 bits signal are compressed. The two 4 bits signal and the 16 bits signal are compressed. Part 61_3,1 , 61_2,1, 61_1,1 Are generated into codes G, D, and A, respectively.
Furthermore, the 16b, 48kHz digital signal is upsampled 73_1,1 Is up-sampled to a signal with a sampling frequency of 96 kHz, and this up-sampled signal and the bit divider 71_2,2 The difference from the 16-bit signal is the error signal Δ_1,2As subtractor 63_1,2The error signal generated by_1,2And code B is generated. Bit division unit 71_2,216-bit signal from the upsampling unit 73_1,2Is up-sampled to a signal with a sampling frequency of 192 kHz, and this up-sampled signal and the bit divider 71_2,3The difference from the 16-bit signal is the subtractor 63._1,3Error signal Δ_1,3As compression part 61_1,3Is encoded into code C. Each compression unit in FIG. 31 may be compression-encoded by the same method as each compression unit shown in FIG.
[0128]
16b, 48kHz digital signal S generated by down-sampling from 24b, 192kHz original sound digital signal_1,1For example, in the case of an audio signal or a music signal, the energy is biased to a lower frequency band, so_1,1 For example, predictive coding, transform coding, or compression coding combined with high compression coding is possible. Specifically, for example, the encoder 61 shown in FIG. 18A can be used.
Compression unit 61_1,2, Compression unit 61_1,3As the input error signal Δ_1,2And Δ_1,39 has energy only in the upper half 24 kHz to 48 kHz and 48 kHz to 96 kHz of the frequency band 0 to 48 kHz and 0 to 96 kHz, so that the prediction error is obtained by inverting the frequency axis of the error signal as described in the embodiment of FIG. The prediction error may be compression encoded. Alternatively, after performing conversion such as processing in the array conversion unit 61E in FIG. 18A, compression encoding may be performed. Compression unit 61_2,1, 61_3,1, 61_2,2 , 61_3,2, 61_2,3 And 61_3,3 For example, an encoder obtained by removing the lossy quantization unit 61B, the inverse quantization unit 61C, and the difference circuit 61D from the encoder 61 of FIG. 18A, that is, the lossless encoder 61 shown in FIG. 19A can be used. These compression parts 61_2,1, 61_3,1,…, 61_2,3, 61_3,3 When the error signal input to is sufficiently small, it becomes close to noise and large compression cannot be expected. Therefore, compression encoding may be performed to a code representing only 0 in this frame.
[0129]
Upsample section 73_1,1 , Upsampling part 73_1,2 In the case where the number of taps of the interpolation filter used in is not known in advance on the decoding side, the auxiliary information representing the number of taps is represented by the auxiliary information encoding unit 65 as indicated by a broken line in FIG._1,2, 65_1,3Are respectively encoded and output as first and second auxiliary codes and first and third auxiliary codes in association with the first and second codes B and the first and third codes C, respectively. Examples of the number of taps of the interpolation filter and auxiliary codes are the same as those in FIG. 20A.
Note that the sound source of each digital signal to be encoded is shown by a broken line block 60 in FIG._2,3, 60_1,3,…, 60_1,1As indicated by, each may exist independently. In that case, each digital signal has a corresponding bit divider 71._3,3, 71_2,3, 71_3,2, 71_2,2, 71_2,1, 71_2,1Or subtracting unit 63_1,3, 63_1,2Or compression unit 61_1,1 To supply. Digital signal S_1,1~ S_2,3Among them, a sound source exists using a digital signal from the existing sound source, and a sound source without a sound source is created from the higher-order digital signal using a bit division unit and a downsampling unit. That is, as shown by a broken line in FIG._2,3 , 75_1,3 , 75_3,2 , 75_2,2 , 75_1,2 , 75_3,1 , 75_2,1 , 75_1,1 Each selecting unit selects a digital signal if a corresponding digital signal sound source is present, and selects a signal of a bit division unit or a downsampling unit immediately above it if not. For example, the selection unit 75_2,3If there is a sound source of a 20b, 192 kHz digital signal, that digital signal is selected._3,3 The bit division unit 71 selects the upper 20 bits from_2,3To supply. Selector 75_3,2 In this case, if a 24b, 96 kHz digital signal sound source is present, that signal is selected._3,3The bit-splitting unit 71 selects the down-sampled signal from_3,2 To supply. Others are the same.
[0130]
It is to be noted that this encoding method is generally described below as an encoding method hierarchized by M types of quantization accuracy and N types of sampling frequencies in the same manner as described above.
Mth and Nth digital signals S of Nth sampling frequency with at least Mth quantization accuracy_{M, N}Sound source 60_{M, N}Shall be obtained from
For a set of m, n in the range m = 1, 2 ≦ n ≦ N, input digital signal S_{m, n}Or digital signal S_{m + 1, n}Digital signal S generated by dividing from_{m, n}And the m, n-1th digital signal S_{m, n-1}Signal S generated by upsampling_{m, n}The difference is subtracted by 63_{m, n}The mth and nth error signal Δ_{m, n}As the compression unit 61_{m, n}The mth and nth codes are generated by compression encoding.
[0131]
For m, n pairs in the range m = M, 2 ≦ n ≦ N, the m, nth digital signal S_{m, n}Down-sampled and the m, n-1 digital signal S_{m, n-1}Is generated. For m, n pairs in the range of 2 ≦ m ≦ M and 1 ≦ n ≦ N, the mth and nth digital signals with quantization accuracy m and sampling frequency n are sampled with m−1, which has a quantization accuracy lower than m. M-1, n digital signal S of frequency n_{m-1, n}And the m, n error signal Δ, which is the error between the m-1, n digital signal and the m, n digital signal._{m, n}And the mth and nth error signal Δ_{m, n}The mth and nth compression unit 61_{m, n}The lossless compression encoding is performed to generate the mth and nth codes.
For the set m = 1, n = 1, the mth, nth digital signal S with the mth quantization accuracy divided from the m + 1, nth digital signal._{m, n}Or the input mth and nth digital signal S_{m, n}Are compressed and encoded to generate the m-th and n-th codes.
[0132]
This encoding method uses the signal S of the highest layer to be encoded._{M, N}First, a digital signal whose sampling frequency falls in order of N-1, N-2, ..., is generated by down-sampling, while maintaining the amplitude resolution M, and then the quantization accuracy is increased for each sampling frequency. It is layered.
Next, a decoding apparatus corresponding to the encoding apparatus shown in FIG. 31 will be described with reference to FIG. Code A, Code D, Code G, Code B, Code E, Code H, Code C, Code F, and Code I are decompression units 80, respectively._1,1, 80_2,1, 80_3,1, 80_1,2, 80_2,2, 80_3,2, 80_1,3, 80_2,3And 80_3,3And decompression decoding is performed respectively. These extension parts 80_{m, n}Is the corresponding compression part 61_{m, n}The m-th and n-th codes that have been compression-encoded in (1) are decompressed and decoded.
[0133]
As in the previous embodiment, for example, a digital signal having a quantization accuracy of 24 bits and a sampling frequency of 192 kHz is expressed as a 24b, 192 kHz digital signal. Extension part 80_1,116b, 48kHz digital signal S_1,1Is output as is and the precision conversion unit 81_1,1Then, 4 bits of 0 are added to the lower order, and a 20b, 48 kHz accuracy conversion signal having an amplitude word length of 20 bits is generated. This accuracy conversion signal is supplied to the expansion unit 80._2,120b, 48kHz error signal Δ_2,1And adder 82_2,120b, 48kHz digital signal S_2,1Is played.
Extension part 80_1,1Decompressed 16b, 48kHz digital signal S from_1,1Is upsampling part 83_1,1The sampling frequency is converted to 96 kHz, and this 16b, 96 kHz upsampled signal is_1,2The decompression decoded 16b, 96kHz error signal is added to the adder 82._1,216b, 96kHz digital signal S_1,2Is played.
[0134]
Expressed in general terms, the expansion unit 80 is used for m, n pairs in the range of 1 ≦ m ≦ M-1, 1 ≦ n ≦ N._{m, n}The mth and nth digital signals of the mth quantization accuracy and the nth sampling frequency that have been decompressed and decoded by the_{m, n}Then, the quantization accuracy (amplitude word length) is converted and generated into an m + 1, n-precision conversion signal of m + 1-th quantization accuracy, and the m + 1, n-precision conversion signal is converted into an expansion unit 80._{m + 1, n} Adder 82 to the m + 1, n residual signal subjected to decompression decoding of_{m + 1, n}Quantization accuracy (amplitude word length) m + 1, m + 1, n digital signal S with nth sampling frequency_{m + 1, n}Is played.
Further, for the m, n pair in the range of m = 1, 1 ≦ n ≦ N−1, the expansion unit 80_{m, n}The m-th and n-th digital signals of the upsampling unit 83_{m, n}Is converted to an m, n + 1 upsample signal of the (n + 1) th sampling frequency, and the m, n + 1 upsample signal is generated by the decompression unit 80._{m, n + 1}Is added to the mth, n + 1th error signal of the (n + 1) th sampling frequency._{m, n + 1}And the mth, n + 1th digital signal S with the mth quantization accuracy and the (n + 1) th sampling frequency._{m, n + 1}Is played. It should be noted that the expansion unit 80 other than m = 1 and n = 1_{m, n}, The mth and nth error signals of the mth quantization accuracy and the nth sampling frequency are decompressed and decoded.
[0135]
Upsample section 83_1,1, 83_1,2When the number of taps of the interpolation filter used in the above is not known in advance, the first and second auxiliary codes and the first and third auxiliary codes respectively input in association with the code B and the code C are respectively the auxiliary code decoding unit 85._1,2, 85_1,3The number of taps as auxiliary information is decoded by the upsampling unit 83 corresponding to each tap number._1,1, 83_1,2Set to
Extension part 80_1,1Is the compression part 61_1,1For example, the compression unit 61_1,1When the encoder 61 of FIG. 18A is used,_1,118B is used.
[0136]
Extension part 80_1,2, 80_1,3Is the compression part 61_1,2, 61_1,3 A decoding method corresponding to this encoding method can be used, and a predictive decoding method, a transform decoding method, or the like can be used. The decoding method corresponding to the encoding method used in the corresponding compression unit is also used for the other decompression units. When the compression unit has the configuration shown in FIG. 19A, the configuration shown in FIG. 19B is used as the decompression unit.
According to the configuration shown in FIG. 31, various digital signals having combinations of various quantization accuracies (amplitude resolution, amplitude word length) and various sampling frequencies (sampling rates) are uniformly encoded as a two-dimensional hierarchy. In addition, it is possible to perform compression coding with high efficiency as a whole, and to provide various combinations of digital signals requested by users with a small amount of data.
[0137]
According to the configuration shown in FIG. 32, the desired one in the digital signal of various combinations of quantization accuracy and sampling frequency is uniformly decoded from the code encoded by the encoding device shown in FIG. can do.
Note that some users do not necessarily need the m-th and n-th digital signals in all combinations shown in FIG. In the decoding apparatus shown in FIG.
Extension part 80_1,1, Upsampling part 83_1,1, Extension part 80_1,2, Adder 82_1,2When,
{Accuracy converter 81_1,1, Extension part 80_2,1And adder 82_2,1} And {precision converter 81_1,2, Extension part 80_2,2 And adder 82_2,2} And {Upsampling part 83_1,2, Extension part 80_1,3, Adder 82_1,3, Precision converter 81_1,3, Extension part 80_2,3And adder 82_2,3 } May be provided.
[0138]
Ninth embodiment
The ninth embodiment is based on the premise that there exists a sound source that outputs the mth and nth digital signals in which M kinds of amplitude word lengths (quantization accuracy) and N kinds of sampling frequencies (sampling rates) are combined. However, if any of the sound sources does not exist, it may be created from a digital signal of a higher layer as in the case of the encoding apparatus shown in FIG.
[0139]
As shown in FIG. 33, for the digital signal of the shortest amplitude word length, in this example 16 bits, the sampling frequency is hierarchized to a lower sampling rate, that is, a digital signal having a lower sampling frequency is set to the same sampling frequency as the digital signal. After up-sampling, the error signal with the up-sampled signal is encoded to obtain the codes B and C, and the digital signal with the lowest sampling frequency, in this example 48 kHz, is, for example, a 16-bit signal and 20 The codes D and G are constructed by sequentially using the error signal of the bit signal and the error signal of the 20-bit signal and the 24-bit signal.
[0140]
If there is a low-order signal in the sampling frequency direction or amplitude resolution direction, that is, a signal with a low sampling frequency or a short amplitude word length, there is room for two types of selection. That is, by comparing both error signals of the error of the digital signal and the digital signal of the lower sampling frequency and the error of the digital signal of the lower amplitude word length (amplitude resolution), an error with an attribute of which the power of the error signal is small A signal is selected and encoded, along with auxiliary information specifying the selected attribute. For example, 20b, 96kHz digital signal S_2,2For accuracy conversion unit 62_1,216b, 96kHz digital signal S_1,2An error signal with a signal in which 4 bits of 0 are added to the lower order of each sample or a 20b, 48kHz digital signal S_2,1The upsample part 64_2,1Generates an error signal with the signal up-sampled to 96 kHz, and selects the error signal with the smaller power._2,2Compression part 61_2,2And the auxiliary information representing the selected attribute is added to the auxiliary encoding unit 77._2,2Is output in association with the code E.
[0141]
Or the digital signal S_2,2Digital signal S with lower sampling frequency_2,1And low-order amplitude resolution (quantization accuracy) digital signal S_1,2The weighted sum signal is generated and the sum signal and the digital signal S_2,2The weighting coefficient is determined as auxiliary information so that the power of the error signal is minimized, and the auxiliary information that is the weighting coefficient and the error signal Δ_2,2Is encoded.
In FIG. 33, it is shown that the reproduction of the 20b, 96 kHz digital signal can be performed using the set of codes A, B, and E, or can be performed using the set of codes B, D, and E. That is, the auxiliary information indicating the selection indicates which decoding path of the white arrow and the black arrow in FIG. 33 is selected when the digital signal is reproduced. When a lower digital signal is selected and an error signal is generated and encoded in this way, codes necessary for reproducing each digital signal are as shown in the table of FIG.
[0142]
Encoder
An embodiment of the encoding apparatus of the ninth embodiment is shown in FIG. The mth and nth digital signals S of the original sound, each of which is a combination of the sampling frequency and quantization accuracy required to generate the codes A to I_{m, n}Sound source 60_{m, n} It is assumed that it is stored in Alternatively, the mth and nth digital signals may be input from the outside. This is the case of m = 1, 2, 3 and corresponds to the m-th amplitude word length (quantization accuracy). In this example, m = 1 is 16 bits, m = 2 is 20 bits, and m = 3 is 24 bits. N = 1, 2, 3 corresponding to the nth sampling frequency (sampling rate). In this example, n = 1 is 48 kHz, n = 2 is 96 kHz, and n = 3 is 192 kHz. Both m and n represent higher levels as the value increases. The m-th and n-th digital signals represent m-th quantization frequency digital signals with the m-th quantization accuracy. Further, the values of the mth quantization accuracy and the nth sampling frequency may be directly used to express, for example, a 16b, 96 kHz digital signal.
[0143]
If a digital signal of a certain condition is not prepared, it is created from the higher-order digital signal. At least the third and third digital signal S_3,3That is, a digital signal source 60 having an amplitude word length of 24 bits and a sampling frequency of 192 kHz._3,3Is prepared and this third and third digital signal S_3,3From other sources 60 by downsampling or truncating the lower bits (lower 4 bits or 8 bits in this example)_{m, n} Digital signal S (m ≠ 3, n ≠ 3)_{m, n}Is generated.
Sound source 60_1,116b, 48kHz digital signal S_1,1Is the compression part 61_1,1The code A is generated and output. The 16b, 48kHz digital signal is converted to the accuracy converter 62._1,1Thus, the first quantization accuracy (16 bits) is converted to a second quantization accuracy (20 bits) higher than this. For example, when a 16b, 48 kHz digital signal is represented by a code absolute value, 0 is added to the lower order, and in this example, 4 bits are added._2,1 20b, 48kHz digital signal S from_2,1And 20b, 48kHz accuracy conversion signal with the same quantization accuracy (same amplitude word length). Sound source 60_2,120b, 48kHz digital signal S_2,1Is the subtractor 63_2,1This 20b, 48kHz accuracy conversion signal is subtracted and the 20b, 48kHz error signal Δ_2,1And this error signal Δ_2,1Compression part 61_2,1Is compressed and encoded to generate and output a code D.
[0144]
16b, 48kHz digital signal S_1,1Is the upsample 64_1,1Thus, the sampling frequency is converted into a 16b, 96 kHz upsampling signal having a second sampling frequency (96 kHz) higher than the first sampling frequency (48 kHz). Sound source 60_1,216b, 96kHz digital signal S_1,2Is the subtractor 63_1,2The difference from the 16b, 96kHz upsampling signal is_1,2Is generated as This 16b, 96kHz error signal Δ_1,2Is the compression part 61_1,2Is compressed and encoded to generate and output code B.
Thus, there is no lower sampling frequency, i.e. the lowest sampling frequency digital signal, e.g. 24b, 48kHz digital signal S_3,1And 20b, 48kHz digital signal S_2,1Is encoded by compressing and encoding an error signal with a digital signal of quantization accuracy one layer lower than the digital signal at the same sampling frequency. Also, digital signals without lower quantization accuracy, such as 16b, 96kHz digital signal S_1,216b, 192kHz digital signal S_1,3Is encoded with a digital signal S with the same quantization accuracy and a lower sampling frequency._1,1, S_1,2The error signal is compressed and encoded.
[0145]
But 20b, 96kHz digital signal S_2,2If there is a lower input digital signal with respect to quantization accuracy and a lower input digital signal with respect to the sampling frequency, one of these is selected. That is, 20b, 96kHz digital signal S_2,2The sampling frequency is the next lower 20b, 48kHz digital signal S with the same amplitude word length._2,1The upsample part 64_2,1Use a 20b, 96kHz upsampled signal upsampled to a sampling frequency of 96kHz, or a 16b, 96kHz digital signal S with the same sampling frequency in the next lower amplitude word length (quantization accuracy)_1,2The precision converter 62_1,2The selection unit 76, which will be described later with reference to FIG. 36, uses either a 20b, 96kHz accuracy conversion signal with 0 added to the lower 4 bits._2,2Select the selected signal and 20b, 96kHz digital signal S._2,2Is the error signal Δ_2,2As subtractor 63_2,2Ask for. This error signal Δ_2,2The digital signal from the lower order of the attribute whose power is small is selected 76_2,2The auxiliary encoding unit 77 encodes information indicating which attribute signal is selected, and outputs an auxiliary code. This 20b, 96kHz error signal Δ_2,2Compression part 61_2,2Is compressed and encoded as code E.
[0146]
Similarly, 24b, 48kHz digital signal S_3,1The upsample part 64_3,124b, 96kHz upsampled signal, and 20b, 96kHz digital signal S_2,2The precision converter 62_2,2By adding “0” of 4 bits to the lower order, a 24b, 96 kHz accuracy conversion signal is obtained, and one of these signals is selected by the selection unit 76._3,2Select the selected signal and the 24b, 96kHz digital signal S._3,224b, 96kHz error signal Δ_3,2As subtractor 63_3,2Generated by the compression unit 61_3,2Let H be the symbol H.
20b, 192kHz digital signal S_2,3Similarly, the 20b, 96kHz digital signal S_2,2Up-sampled signal and 16b, 192kHz digital signal S_1,3Error signal Δ with one of the accuracy conversion signal of_2,3Is compressed and encoded to generate a code F. 24b, 192kHz digital signal S_3,3The two lower digital signals S_3,2, S_2,3Select one of 76_3,3And select the error signal Δ_3,3A code I is generated from
[0147]
Selector 76_2,2, 76_3,2, 76_2,3, 76_3,3A specific example of this is shown in FIG. In this example, the m, n + 1-th digital signal S for a set of m, n in the range of 2 ≦ m ≦ M, 1 ≦ n ≦ N−1._{m, n + 1}M, n digital signal S_{m, n}Upsample section 64_{m, n}Up-sampled to the m, n + 1th upsampled signal, and the m-1, n + 1th digital signal S_{m-1, n + 1}The precision converter 62_{m-1, n + 1}Thus, the accuracy is converted to the m, n + 1 accuracy conversion signal. These m, n + 1 upsample signals and m, n + 1 digital signals S_{m, n + 1}And m, n + 1 precision conversion signal and m, n + 1 digital signal S_{m, n + 1}Are calculated by the strain calculators 76A and 76B as mth, nth strain and m-1, n + 1 strain, respectively. The comparison unit 76C compares the powers of the mth and nth distortions and the m-1th and n + 1th distortions, and selects the mth, n + 1th upsample signal if the powers of the mth and nth distortions are smaller. If the power of the (m-1, n + 1) th distortion is smaller, the switch 76D is controlled so as to select the mth, n + 1th accuracy conversion signal.
[0148]
The signal selected by the switch 76D is the subtractor 63._{m, n + 1}M, n + 1 digital signal S_{m, n + 1}And m, n + 1 error signal Δ_{m, n + 1}Is generated, and this is the compression unit 61._{m, n + 1}Thus, compression coding is performed on the m, n + 1 code. In this way, the m, n + 1th digital signal S_{m, n + 1}And the mth and nth digital signals S_{m, n}And m-1, n + 1 digital signal S_{m-1, n + 1}The error signal with the smaller power of the two error signals is the m, n + 1 error signal Δ_{m, n + 1}Is generated as The auxiliary information indicating which signal is selected by the switch 76D is associated with the m, n + 1 code as the m, n + 1 auxiliary code by the auxiliary encoding unit 77. This auxiliary information is the mth, n + 1th digital signal S._{m, n + 1}On the other hand, the m, nth digital signal S with the sampling frequency immediately below_{m, n}Or the (m-1, n + 1) th digital signal S with the quantization accuracy immediately below_{m-1, n + 1}The m, n + 1 auxiliary code may be a total of 2 bits, ie, a code indicating the presence / absence of auxiliary information and a code indicating which one is selected. The error signal code and the auxiliary information may be integrated and output so that the +1 code can be distinguished.
[0149]
An embodiment of a decoding apparatus corresponding to the encoding apparatus in FIG. 35 is shown in FIG. Decoding of the digital signal with the lowest sampling frequency of 48 kHz is the same as that of the decoding apparatus shown in FIG. When a digital signal with lower quantization accuracy and lower sampling frequency than the digital signal to be decoded is reproduced, for example, a 20b, 96 kHz digital signal S_2,2Is reproduced, the reproduced 20b, 48kHz digital signal S_2,1The upsampling part 83_2,1In the selection section 87, the up-sampled signal is 20b, 96kHz._2,216b, 96kHz digital signal S supplied and regenerated to_1,2Is the precision converter 81_1,2Is converted to a 20b, 96kHz accuracy conversion signal by the selector 87._2,2Supplied to. Auxiliary decoder 86_2,2And the second and second auxiliary codes are decoded in accordance with the selection information indicated by the decoded auxiliary information._2,2Selects one of both input signals and adds 82_2,2 To give. Adder 82_2,2Is the extension 80_2,2Decoding of code E from 20b, 96kHz error signal Δ_2,2And selection part 87_2,220b, 96kHz digital signal S by adding the signals selected by_2,2Is played.
[0150]
In general, for m, n pairs in the range of 2 ≦ m ≦ M, 1 ≦ n ≦ N−1, the m, n + 1 auxiliary code is the auxiliary decoding unit 86._{m, n + 1}According to the auxiliary information decoded in step 1, the signal of any attribute, that is, the reproduced mth and nth digital signal S_{m, n}The upsample part 83_{m, n} The m-1, n + 1 upsampled signal or the regenerated m-1, n + 1 digital signal S_{m-1, n + 1}The precision converter 81_{m-1, n + 1}Any one of the m, n + 1th accuracy conversion signals is the selection unit 87._{m, n + 1}Selected. The selected signal is added by the adder 82._{m, n + 1}The m, n + 1 error signal Δ obtained by decompressing the m, n + 1 code in_{m, n + 1}And the m, n + 1th digital signal S_{m, n + 1}Is played.
The mth and nth digital signals S_{m, n}, M-1, n + 1 digital signal S_{m-1, n + 1}The reproduction method from the code is not limited to the method shown in FIG. In short, it is sufficient if there is a means for reproducing both of these digital signals.
[0151]
10th embodiment
In the ninth embodiment, a digital signal having the same sampling frequency and a lower quantization accuracy and a digital signal having a smaller error signal power among the digital signals having the same quantization accuracy and a lower sampling frequency are selected. However, the power of the error signal may be reduced by weighted addition of these two lower digital signals. That is, each selection unit 76 in FIG._{m, n}In the block of (2 ≦ m ≦ M, 2 ≦ n ≦ N), both input signals are weighted and added using the mixing unit instead of the selection unit so that the mixing unit is written in parentheses and output. . For example, mixing section 76_2,2Then, the upsampling part 64_2,120b, 96kHz upsampled signal and accuracy converter 62_1,220b, 96kHz accuracy conversion signal is weighted and added, and the 20b, 96kHz addition signal and 20b, 96kHz digital signal S_2,2And 20b, 96kHz error signal Δ_2,2Subtracting unit 63_2,2Generate with At that time, 20b, 96kHz error signal Δ_2,2Mixing section 76 to minimize the power of_2,2The set of weight coefficients used in is selected from a plurality of sets stored in a storage unit (not shown) and determined. In this way, a 20b, 96kHz error signal Δ that minimizes power_2,2Compression part 61_2,2And the code E is output.
[0152]
Mixing section 76_{m, n + 1}A specific example of this is shown in FIG. Upsample section 64_{m, n}M, n + 1 upsampled signal from, accuracy conversion unit 62_{m-1, n + 1}Further, the m, n + 1th accuracy conversion signal is multiplied by the weighting factors W1 and W2 of the sets selected by the multipliers 76G and 76H, respectively. These multiplication results are added by the adder 76J, and the m, n + 1 added signal and the m, n + 1 digital signal S are added._{m, n + 1}The difference between the subtractor 63 and the error signal_{m, n + 1}Is generated. The m, n + 1 error signal is branched and input to the control unit 76K. As described above, the control unit 76K has a plurality of sets of weighting factors W1, W2 determined in advance in a storage unit (not shown) as a table corresponding to codes representing these sets. 1 Error signal Δ_{m, n + 1}A set of weighting factors W1 and W2 is selected from the weighting factor table so as to minimize the power of and is supplied to the multipliers 76G and 76H. M, n + 1 error signal Δ when error signal power is the smallest_{m, n + 1}Compression part 61_{m, n + 1}And the code that designates the set of weighting factors (W1, W2) selected at that time is used as the m, n + 1 auxiliary code by the auxiliary encoding unit 79, and the error signal Δ_{m, n + 1}Are output in association with the mth, n + 1 codes.
[0153]
In general, encoding of a digital signal is performed by dividing it into frames (encoding unit time). However, the determination of the auxiliary information is not limited to each frame, and may be performed for each subframe constituting the frame.
In a decoding apparatus corresponding to the encoding apparatus using the mixing unit 76, the mixing unit 87 may be used instead of the selection unit 87 as shown in parentheses in FIG. The mixing unit 87 is the same as the configuration having the weighted addition shown in FIG. 38, that is, the configuration having the multiplying units 76G and 76H and the adding unit 76J. For example, the auxiliary decoding unit 86_2,2The same weighting factor table as that of the control unit 76K of FIG. 38 is provided in a storage unit (not shown), and a corresponding set of weighting factors based on the input auxiliary code, that is, a code indicating a combination of weights. W1 and W2 are extracted from the weight coefficient table, and the upsampling unit 83_2,120b, 96kHz upsampled signal and accuracy converter 81_1,2For the 20b, 96kHz accuracy conversion signal from_2,2Are multiplied by weights W1 and W2, respectively, and the multiplication results are added._2,2Decompressed 20b, 96kHz error signal Δ from_2,2And adder 82_2,220b, 96kHz digital signal S_2,2Play.
[0154]
Generally speaking, the upsampling unit 83_{m, n}M, n + 1 upsampled signal and accuracy conversion unit 81_{m-1, n + 1} For the m, n + 1 precision converted signal, the m, n + 1 auxiliary code is converted to the auxiliary decoding unit 86._{m, n + 1}A set of weighting factors W1 and W2 specified by the auxiliary code input in step_{m, n + 1}And the multiplication result is added, and the m, n + 1 addition signal and the expansion unit 80 are added._{m, n + 1}The m, n + 1 error signal Δ obtained by decoding the m, n + 1 code at_{m, n + 1}And the adder 82_{m, n + 1}The m, n + 1th digital signal S is added by_{m, n + 1}Play.
Modification of the tenth embodiment
The m-th and n-th digital signals of various combinations of quantization accuracy and sampling frequency in FIG. 35 are input as signals collected separately from the same sound field, or they are input to the sound source 60._1,1~ 60_3,3Are once stored and read from them. The digital signal of each sound source is the mth and nth digital signal S._{m, n}Unlike simple downsampling and lower bit truncation, noise (fixed dither signal) may be added, and various conversions and adjustments such as amplitude and sampling (sample point position) deviations may be made. There is a possibility that It is generally unknown in advance how this conversion or adjustment is.
[0155]
Therefore, in this modification, in the encoding apparatus shown in FIG. 35, a digital signal having a lower (n-1) th sampling frequency or a lower mth sampling frequency is compared with a digital signal having an nth sampling frequency and mth quantization accuracy. The digital signal with -1 quantization accuracy is transformed into a digital signal of the same level as described above so that the power of the error signal obtained by subtracting the lower digital signal from the upper digital signal is minimized.
For example, as shown in FIG. 22, the mth and nth digital signals S_{m, n}Is the accuracy converter 62_{m, n} As described above, the quantization accuracy (amplitude word length, amplitude resolution) is converted to the (m + 1) th quantization accuracy, and this m + 1, n accuracy conversion signal is leveled by the gain adjusting unit 66A in this example. The timing adjustment unit 66B adjusts the sample position of the m + 1, n accuracy conversion signal that has been adjusted and further adjusted in level (gain), and the m + 1, n accuracy conversion signal that has been adjusted in the sample position adjusts the sample position. m + 1, n digital signal S_{m + 1, n}Subtracting the subtracting unit 63_{m + 1, n}Done in These adjustments are the same as those described with reference to FIG.
[0156]
Further, when the encoding apparatus performs time / gain adjustment on the lower-order digital signal, specifically, for example, the (m + 1) th and n-th precision conversion signal, the decoding apparatus corresponds to the mth It needs to be done on + 1, n precision conversion signal. The example in that case can also have the same configuration as that described with reference to FIG. 23, and since the operation has already been described, description thereof is omitted here.
In the above modification, the lowest sampling frequency is 48 kHz in FIG. 35 and FIG. 37, and the lowest quantization accuracy is 16 bits in FIG. 35 and FIG. It is an example. In the case of using the selection unit or the mixing unit, the upsampling unit 64 is adjusted by the adjustment unit 76E as shown by the broken lines in FIGS._{m, n}M, n + 1th upsample signal and m, n + 1th digital signal S_{m, n + 1}In FIG. 36, one or both of the level adjustment by the gain adjustment unit 66A and the sample position adjustment by the timing adjustment unit 66B shown in FIG. 22 are performed, and the distortion calculation unit 76A and the switch unit 76D in FIG. 76G), and the adjustment unit 76F provides the accuracy conversion unit 62_{m-1, n + 1}The distortion calculation unit 76B (or the level adjustment and the sample position adjustment shown in FIG. 22 or both are performed between the m, n + 1 accuracy conversion signal and the m, n + 1 digital signal. Multiplier 76H). Also, the adjustment unit 76E, 76F outputs the m, n + 1 auxiliary code of one or both of the adjustment gain and the adjustment sample position amount. These m, n + 1 auxiliary codes may be output as one m, n + 1 auxiliary code together with the m, n + 1 auxiliary code from the auxiliary encoding unit 77. In the example shown in FIG. 38, values obtained by multiplying the adjustment gains in the adjustment units 76E and 76F and the weights W1 and W2 of the multiplication units 76G and 76H may be used as auxiliary information.
[0157]
In the decoding apparatus shown in FIG._{m, n}For example, as shown in FIG. 39, the m, n + 1 auxiliary code is decoded by the auxiliary information decoding unit 88. Upsample section 83_{m, n} And precision conversion unit 81_{m-1, n + 1}And selection part (mixing part) 87_{m, n + 1} Adjusters 87A and 87B are inserted between them, respectively, and these adjusters 87A and 87B have the same configuration as the adjuster 87 shown in FIG. Corresponding to one or both of the adjustment sample position amounts and the m, n + 1 upsample signal and the m, n + 1 accuracy conversion signal are respectively performed for level adjustment and / or sample position adjustment. Breaking selection part (mixing part) 87_{m, n + 1}Supplied to.
[0158]
In the coding apparatus shown in FIG. 35, the 20b, 96 kHz digital signal S_2,234, a set of codes A, D, and E can be used as shown in FIG. 34, and a set of codes A, B, and E can also be used. Therefore, an encoding method using a set with the least amount of information can be used. Similarly, 24b, 192kHz digital signal S_3,3For A, B, C, F, I, A, B, E, F, I, A, B, E, H, I, A, D, E, F, It is encoded by the one with the smallest total information amount among the six types of encoding of the set of I, the set of codes A, D, E, H, and I, and the set of codes A, D, G, H, and I. Therefore, a high compression rate can be obtained. For example, in the case of a 20b, 192 kHz digital signal, a set of codes A, B, C, and F, and a set of A, B, E, and F, as shown in the logical expression shown in FIG. , A, B, E, and F, and A, D, E, and F, can be encoded in four ways. In the case of a 24b, 96 kHz digital signal, encoding can be performed in three ways: a set of A, B, E, H, a set of A, D, E, H, and a set of A, D, G, H. . Of these, the transmission efficiency can be increased by using the group having the smallest total information amount (the group having the highest compression rate).
[0159]
The compression unit in the encoding device shown in FIG. 35 can also have the same configuration as the compression unit of the encoding device shown in FIGS. 27 and 31. Similarly, in the decoding device shown in FIG. The decompressor can have the same configuration as the decompressor in the decoding apparatus shown in FIGS.
As described above, in the encoder of the tenth embodiment, when any sound source does not exist or when there is only a digital signal sound source having the highest quantization accuracy and the highest sampling frequency, the signal of the existing sound source is obtained. Then, a digital signal having other quantization accuracy and sampling frequency is generated and encoded. 24b, 192kHz digitized signal S_3,3An example in which all digital signals are generated from FIG. 40 will be described with reference to FIG. 40 where only the parts corresponding to those in FIG. The sound source indicated by the broken line on the left side in FIG. 40 does not exist.
[0160]
24b, 192kHz digital signal S_3,3Is the digit part 67_3,3The lower 4 digits are removed and 20b, 192kHz digital signal S_2,3Is generated. The 20b, 192kHz digital signal is further reduced by 67_2,316b, 192kHz digital signal S without using the lower 4 bits_1,3Is generated. 24b, 192kHz digital signal S_3,3Down sample section 68_3,3The sampling frequency is downsampled to 96kHz, and the 24b, 96kHz digital signal S_3,2Is generated. This 24b, 96kHz digital signal S_3,2Is the digit part 67_3,2, 67_2,2In turn, without using the lower 4 bits, each 20b, 96kHz digital signal S_2,216b, 96kHz digital signal S_1,2Is generated. Similarly, 24b, 48kHz digital signal S_3,1, 20b, 48kHz digital signal S_2,116b, 48kHz digital signal S_1,1Down sample section 68_3,2, Digit part 67_3,1, 67_2,1Is generated by
[0161]
Another example of these digital signal generation methods is shown in FIG. In this case, as shown in FIG._3,3, 67_2,320b, 192kHz digital signal S_2,316b, 192kHz digital signal S_1,3Generate and downsample part 68_3,3, 68_3,224b, 96kHz digital signal S_3,2And 24b, 48kHz digital signal S_3,1, But in this example, the digit part 67_3,320b, 192kHz digital signal S_2,3, Digit part 67_2,316b, 192kHz digital signal S_1,3Each downsampler 68_2,3, 68_1,320b, 96kHz digital signal, 16b, 96kHz digital signal S_1,2Are generated, and these are further downsampled 68._2,2, 68_1,220b, 48kHz digital signal S_2,116b, 48kHz digital signal S_1,1Is generated. Other configurations in FIGS. 40 and 41 are the same as those in FIG.
[0162]
In the seventh to tenth embodiments described above, the number M of types of quantization accuracy and the number N of types of sampling frequencies are not limited to 3 each, and the value of M may be a different value. Similarly, the value of N is not limited to 3, and may be another value. In addition, each encoder and each decoder shown in each of the above-described embodiments may be configured to function by causing a computer to execute a program. In this case, for example, with regard to the decoder, the decoding program is downloaded from a recording medium such as a CD-ROM or a magnetic disk or through a communication line by a control means such as a CPU in the computer, and the decoding program is downloaded to the computer. Just execute.
[0163]
According to the seventh to tenth embodiments, for example, the musical sound distribution system described with reference to FIG. 24 can be executed.
According to the seventh to tenth embodiments described above, encoding with different amplitude quantization accuracy and sampling frequency can be executed in a unified manner, and the compression rate of the entire system can be increased.
11th embodiment
First, FIG. 42 shows two-dimensional hierarchization of digital signals used in the eleventh embodiment. A digital signal of M × N = 9 combinations, using three types of quantization types of 16 bits, 20 bits, and 24 bits as M types of quantization accuracy, and three types of 48 kHz, 96 kHz, and 192 kHz as N types of sampling frequencies. Can be generated.
[0164]
The upper 16 bits obtained by removing the lower 8 bits from the least significant bit in the digital signal having a quantization accuracy of 24 bits are encoded by the code A encoded at a sampling frequency of 48 kHz and the upper 16 bits by the code A. A code B encoded with a sampling frequency of 96 kHz for a frequency component higher than the component, a code C encoded with a sampling frequency of 192 kHz for a frequency component higher than the component encoded with the code B, and a sampling for a signal with an amplitude word length of 16 bits The frequency is hierarchized into a plurality. In other words, the sampling frequency is hierarchized by a 16-bit word length signal.
[0165]
For a signal in which the lower 4 bits are added to the 16-bit word length to obtain a 20-bit word length, the lower 4-bit component, that is, a residual component (error) obtained by subtracting the 16-bit word length component from the 20-bit word length signal. Signal), the error signal between the code D encoded at a sampling frequency of 48 kHz, a signal obtained by upsampling a signal having a 20-bit word length and a sampling frequency of 48 kHz to a sampling frequency of 96 kHz, and a digital signal having a 20-bit sampling frequency of 96 kHz is compressed Hierarchy of sampling frequency for a 20-bit word length signal by an encoded code J, a signal obtained by up-sampling a 20b, 96 kHz digital signal to a sampling frequency of 192 kHz, and a code K obtained by compression encoding an error signal of the 20b, 192 kHz digital signal Is realized.
[0166]
With respect to a signal having a 24-bit word length by adding lower 4 bits to the 20-bit word length, the lower 4-bit component, that is, a residual component obtained by subtracting the 20-bit word length component from the 24-bit word length signal ( Error signal), a code G encoded at a sampling frequency of 48 kHz, a signal obtained by up-sampling a 24b, 48 kHz digital signal to a sampling frequency of 96 kHz, a code L obtained by compression encoding the error signal of the 24b, 96 kHz digital signal, Hierarchical coding is performed in the frequency direction as a code M obtained by compression-coding an error signal between a signal obtained by up-sampling a 24b, 96 kHz digital signal at a sampling frequency of 192 kHz and the 24b, 192 kHz digital signal. That is, hierarchization for quantization accuracy of 16 bits or more is performed for each sampling frequency. The quantization accuracy and sampling frequency of the signals corresponding to codes A, B, C, D, and G in this hierarchization are the same as the corresponding codes in FIG. 25. For example, in this embodiment, the signal corresponding to the code L Includes signals corresponding to symbols B, E, and H in FIG. Similarly, the code M in this embodiment includes the codes C, F and I in FIG. 25, the code K in this embodiment includes the codes C and F in FIG. 25, and the code J in this embodiment is the code in FIG. B and E are included.
[0167]
Thus, by using the codes A to D, G, and J to M encoded according to the nine kinds of encoding conditions that are two-dimensionally hierarchized with the amplitude word length (amplitude resolution, quantization accuracy) and the sampling frequency. 43, all M × N = 9 types of digital signals of combinations of M = 3 types of amplitude word lengths and N = 3 types of sampling frequencies can be output. That is, the symbols shown in FIG. 43 may be used for each combination of sampling frequency and quantization accuracy. For example, in the case of a sampling frequency of 96 kHz and an amplitude word length of 24 bits, codes A, D, G, and L may be used.
Next, an encoding method for generating these codes A to D, G, and J to M will be described with reference to the functional configuration shown in FIG. The mth and nth digital signals of the original sound of the combination of the sampling frequency and the amplitude word length necessary for generating the codes A to D, G and J to M are the sound sources 60._{m, n} It is assumed that it is stored in m = 1, 2, 3 corresponding to the m-th amplitude word length (quantization accuracy). In this example, m = 1 is 16 bits, m = 2 is 20 bits, and m = 3 is 24 bits. N = 1, 2, 3 corresponding to the nth sampling frequency (sampling rate). In this example, n = 1 is 48 kHz, n = 2 is 96 kHz, and n = 3 is 192 kHz. If a digital signal of a certain condition is not prepared, it is created from the digital signal of the higher hierarchy. At least the third and third digital signal S_3,3That is, a digital signal source 60 having an amplitude word length of 24 bits and a sampling frequency of 192 kHz._3,3Is prepared, this third digital signal S_3,3From other sources 60 by downsampling or truncating the lower 4 bits or 8 bits as necessary._{m, n}A digital signal (m ≠ 3, n ≠ 3) is generated.
[0168]
Sound source 60_1,1 1st and 1st digital signal S_1,1Is the compression part 61_1,1The first and first codes A are generated and output. The first and first digital signals are also converted by the accuracy converter 62._1,1 Therefore, the first quantization accuracy is converted to a second quantization accuracy higher than this. For example, the first and first digital signals S_1,1Is a code absolute value expression, 0 is added to the lower order of the predetermined number of bits, in this example, 4 bits are added._2,1 2nd and 1st digital signal S_2,1Are the second and first precision conversion signals having the same quantization accuracy (same amplitude word length). Sound source 60_2,1 2nd and 1st digital signal S_2,1Is the subtractor 63_2,1The second and first precision conversion signals are subtracted by the second and first error signals Δ_2,1Is generated and this second and first error signal Δ_2,1Compression part 61_2,1 Is compressed and encoded to generate and output a code D. In other words, the lowest sampling frequency in a plurality of digital signals is the error signal from the signal obtained by converting the digital signal of quantization accuracy one layer lower than the digital signal to the same quantization accuracy (amplitude word length). Compress and encode. The third and first digital signals are similarly encoded, and a code G is generated.
[0169]
The first and first digital signals S_1,1Is the first and first upsampling section 64_1,1Thus, the sampling frequency is converted and generated into first and second upsampling signals having a second sampling frequency higher than the first sampling frequency. In this example, the sampling frequency is converted from 48 kHz to 96 kHz. For example, as described with reference to FIGS. 17A and 18B, a broken line sample for interpolating each adjacent sample in the solid line sample row of the first and first digital signals is inserted. Is done.
Sound source 60_1,21st and 2nd digital signal S_1,2Is the subtractor 63_1,2 The first and second upsampling signals are subtracted, and the first and second error signals Δ_1,2Is generated. This first and second error signal Δ_1,2Is the compression part 61_1,2The first and second codes B are generated and output.
[0170]
Other codes B, C, J, K, L, and M are similarly generated. The generation of these codes is generally described. For the combination of m and n, the lowest m = 1, n = 1 mth and nth digital signals are converted into mth and nth compression units 61._{m, n} To generate the mth and nth codes and output them.
M, n sets of m, n digital signals S in the range of 2 ≦ m ≦ M, n = 1_{m, n}For the m-1 and n-precision converted signals with the same quantization accuracy m, the m-1 and n-precision digital signals are converted from the m-1 and n-precision digital signals immediately below the m-th quantization accuracy. Conversion unit 62_{m-1, n} The subtractor 63 generates the difference between the mth and nth digital signals and the mth and nth precision converted signals._{m, n} And the m-th and n-th error signals are calculated by the compression unit 61._{m, n} To generate and output the m-th and n-th codes.
[0171]
For the m, n digital signal whose sampling frequency is not the lowest, that is, n ≧ 2, the m, n−1 digital signal immediately below the sampling frequency with the same quantization accuracy is used as the nth sampling frequency. -1 Upsample section 64_{m, n-1}The upsampling is performed to generate the mth and nth upsampling signals, and the mth and nth upsampling signals are subtracted from the mth and nth digital signals._{m, n}And the remaining m-th and n-th error signals are compressed by the compression unit 61._{m, n}To generate and output the m-th and n-th codes.
When the original sound signal is speech or music, generally, the first and first digital signals are largely biased in the low frequency region._1,1For example, predictive coding, transform coding, or compression coding combined with high compression coding is possible. Specifically, for example, the encoder 61 of FIG. 18A described above can be used.
[0172]
Compression unit 61_1,2, 61_1,3As described above, the first and second error signals and the first and third error signals, which are inputs to, are out of the band of the first and first error signals. Since there is energy only in the upper half of the frequency band, the signal may be predicted or may be compressed and encoded after performing conversion such as the processing in the array conversion unit 61E in FIG. 18A described above. Compression unit 61_2,1, 61_3,1, 61_2,2, 61_3,2, 61_2,3And 61_3,3For example, the combination of the prediction encoder and the lossless compression unit shown in FIG. 28 described above, or the lossy quantization unit 61B, the inverse quantization unit 61C, and the difference circuit 61D from the encoder 61 shown in FIG. The removed encoder, that is, the lossless encoder 61 shown in FIG. 19A can be used. These compression parts 61_2,1, 61_3,1, ..., 61_2,3, 61_3,3If the error signal input to is sufficiently small and the sequence is random like noise, the improvement of the compression rate cannot be expected. Therefore, in this frame, compression encoding may be performed to a code representing only 0.
[0173]
Each upsample section 64_{m, n}When the number of taps of the interpolation filter used in (see FIG. 17B) is not known in advance on the decoding side, the number of taps is shown in FIG._{m, n + 1}Are encoded and output as m, n + 1 auxiliary codes in association with the m, n + 1 codes, respectively. An example of the number of taps of the interpolation filter and the auxiliary code is shown in FIG. 20A, for example.
Next, a digital signal decoding method corresponding to FIG. 44 will be described with reference to FIG.
Reference numerals A, D, G, B, J, L, C, K, and M are expansion units 80, respectively._1,1, 80_2,1, 80_3,1, 80_1,2, 80_2,2, 80_3,2, 80_1,3, 80_2,3And 80_3,3And decompression decoding is performed respectively. These mth and nth extension parts 80_{m, n}Is the corresponding m-th and n-th compression unit 61._{m, n}The m-th and n-th codes that have been compression-encoded in (1) are decompressed and decoded.
[0174]
Extension part 80_1,1The first and first digital signals subjected to decompression decoding from the precision conversion unit 31_1,1In this way, the second and first precision converted signal with 0 added to the lower order and the amplitude word length of 20 bits is generated. This second and first precision converted signal is sent to the expansion unit 80._2,1The second decoded error signal Δ_2,1And adder 82_2,1Is added to the second and first digital signals S_2,1Is played.
Extension part 80_1,1Decompressed first and first digital signals S from_1,1Is upsampling part 83_1,1Thus, the first sampling frequency is converted into the first and second upsampling signals converted into the second sampling frequency. The first and second upsampling signals are supplied to the first and second expansion units 80._1,2The first and second error signals Δ_1,2Adder 82_1,2The first and second digital signals S_1,2Is played.
[0175]
Extension part 80_{m, n}The m-th and n-th digital signals having the m-th quantization accuracy and the n-th sampling frequency that have been decompressed and decoded by the m-th, n-th precision conversion unit 81 if n is the lowest value, that is, n = 1._{m, n}Then, the quantization accuracy (amplitude word length) is converted and generated into an m + 1, n-precision conversion signal of m + 1-th quantization accuracy, and the m + 1, n-precision conversion signal is converted into an expansion unit 80._{m + 1, n}Adder 82 to the m + 1, n error signal subjected to decompression decoding of_{m + 1, n}The (m + 1) th digital signal having the (m + 1) th quantization accuracy (amplitude word length) and the nth sampling frequency is reproduced.
Also, the expansion part 80_{m, n}If the sampling frequency of the m-th and n-th error signals from is higher than the minimum value, that is, n> 1, the m-th reproduction signal of the (n-1) -th sampling frequency immediately below is sampled with the same m-th quantization accuracy. , n-1 digital signal is the mth, n-1 upsampling unit 83._{m, n-1}Is converted into an mth and nth upsampling signal having an nth sampling frequency, and the mth and nth upsampling signals are added to the mth and nth error signals by an adder 82._{m, n}And the mth and nth digital signals having the mth quantization accuracy and the nth sampling frequency are reproduced. It should be noted that the expansion part 80 other than m = 1, n = 1_{m, n}, The mth and nth error signals of the mth quantization accuracy and the nth sampling frequency are decompressed and decoded.
[0176]
Each upsample part 83_{m, n}When the number of taps of the interpolation filter used in the above is not known in advance, the first and second auxiliary codes, the second and second auxiliary codes, and the second auxiliary code input in association with the codes B, J, L, C, K, and M, respectively. 3, 2 auxiliary code, 1st, 3rd auxiliary code, 2nd, 3rd auxiliary code and 3rd, 3rd auxiliary code respectively_1,2, 85_2,2, 85_3,2, 85_1,3, 85_2,3And 85_3,3Thus, the number of taps is decoded, and the upsampling unit 83 corresponding to each tap number corresponds to_1,1, 83_2,1, 83_3,1, 83_1,2, 83_2,2And 83_3,2Set to
Extension part 80_1,1Is the compression part 61_1,1For example, the compression unit 61_1,1When the encoder 61 of FIG. 18A is used,_1,1The decoder in FIG. 3 is used. That is, the lossless compression code in the code A is losslessly decoded, and a plurality of samples expressing the code absolute value of the bit string composed of the corresponding bit position in the frame from the decoded bit string is reproduced as an error signal of the frame. The lossless compression code is irreversibly decoded to generate a local reproduction signal, and the reproduction signal and the error signal are added to reproduce the first and first digital signals.
[0177]
Extension part 80_1,2, 80_1,3Is the compression part 61_1,2, 61_1,3A decoding method corresponding to this encoding method can be used, and a predictive decoding method, a transform decoding method, or the like can be used. The decoding method for the encoding method used in the corresponding compression unit is also used for the other decompression units. When the compression unit 61 has the configuration shown in FIG. 19A, the decompression unit 80 has a configuration in which the inverse quantization unit 80C and the addition unit 80D are removed from the decoder 80 in FIG. 18B, that is, the configuration shown in FIG. 19B. To do.
According to the configuration of the encoding device shown in FIG. 44, various digital signals having combinations of various quantization accuracies (amplitude resolution, amplitude word length) and various sampling frequencies (sampling rates) are formed as a two-dimensional hierarchy. Encoding can be performed uniformly, and compression coding can be performed with high efficiency as a whole, and various combinations of digital signals for providing a reproduction signal of the quality required by the user can be provided with a small amount of data. .
[0178]
According to the configuration of the decoding device shown in FIG. 45, the desired ones in the digital signals of various combinations of quantization accuracy and sampling frequency are unified from the codes encoded by the encoding device shown in FIG. Can be decrypted automatically.
Note that the m-th and n-th digital signals in all combinations shown in FIG. 44 are not necessarily required. For example, the decompression unit 80 in the decoder shown in FIG._1,1In addition, the upsampling unit 83_1,1, Extension part 80_1,2, Adder 82_2,1A first means comprising: an accuracy conversion unit 81_1,1, Extension part 80_2,1And adder 82_2,1A second means comprising: an accuracy conversion unit 81_1,2, 2nd and 2nd extension 80_2,2, 2nd and 2nd adder 82_2,2Upsampling unit 83_2,1, Extension part 80_2,2, Adder 82_2,2It may be provided with at least one of the third means.
[0179]
44 and FIG. 45, the number M of types of quantization accuracy and the number N of types of sampling frequencies are not limited to three, but may be other values.
The sound source 60 of the mth and nth digital signals of various combinations of quantization accuracy and sampling frequency in FIG._1,1~ 60_3,3Is prepared beforehand, the m-th and n-th digital signals of each sound source are different from those obtained by simply down-sampling the m-th and n + 1-th digital signals or by rounding down the lower bits. ), And various conversions and adjustments such as deviations in amplitude and sampling (sample point position) may be performed. It is generally unknown in advance how this conversion or adjustment is.
[0180]
In short, the encoding method according to the eleventh embodiment has one quantization accuracy and one sampling frequency for encoding digital signals having various quantization accuracies (amplitude resolution, amplitude word length) and various sampling frequencies (sampling rates). Digital signal encoding has the same quantization accuracy as this, generates an error signal with a signal with a lower sampling frequency but a closer digital signal upsampled to the same sampling frequency, and compresses the error signal Encoding is performed only by compressing and encoding the error signal with the upsampling signal as described above except for the digital signal having the lowest sampling frequency. For a digital signal with the lowest sampling frequency, an error signal is compressed and encoded with a signal whose accuracy is converted to the same quantization accuracy (same amplitude word length) from the closest digital signal with a lower quantization accuracy.
[0181]
Also, the decoding method of the eleventh embodiment generates an error signal by decompressing and decoding the compression code of the error signal for the digital signal to be decoded, and has the same quantization accuracy as the digital signal to be decoded and a low sampling frequency. The reproduced digital signal closer to is up-sampled to the same sampling frequency as the sampling frequency of the decoded error signal, and added to the decoded error signal to obtain a digital signal. The modification of FIGS. 22 and 23 applied in the embodiment of FIGS. 16 and 21 is applied to the embodiment of FIGS. 44 and 45, and the level adjustment of the sample and the sample position are applied to one or both of the upsampling signal and the accuracy conversion signal. May be adjusted.
[0182]
The coding apparatus shown in FIG. 44 and the decoding apparatus shown in FIG. 45 may be configured to cause a computer to execute a program and function. In this case, for example, with regard to a decoding device, a decoding program may be downloaded into a computer from a recording medium such as a CD-ROM or a magnetic disk or through a communication line, and the decoding program may be executed by the computer. .
Although the present invention is applied to a music digital signal in the above description, it can also be applied to an image digital signal.
According to the eleventh embodiment described above, encoding with different amplitude accuracy and sampling rate requirements, particularly reversible encoding, can be performed in a unified manner, and both individual compression performance and compression performance for all of the various encoding conditions are compatible. it can.
12th embodiment
FIG. 46 shows a conceptual configuration of the twelfth embodiment of the present invention. In this embodiment, 5-channel signal L5c: front left, R5c: front right, C5c: center, LS5c: left rear (surround), RS5c: right rear (surround), two-channel stereo signals L and R, It is assumed that signals of three types of channels of the mono-channel monaural signal M are hierarchically encoded. Both are recorded in the same space. The stereo signals L and R and the monaural signal M with a smaller number of channels are lower than the 5-channel signal, and the monaural signal M with a smaller number of channels (ie, one channel) is lower or pre-arranged with respect to the stereo signals L and R. Hierarchized to those recorded according to the standard.
[0183]
First, the monaural signal M is compression-coded alone. This compression encoding may be lossless encoding or lossy encoding. The stereo signals L and R are encoded by correcting the monaural signal M, subtracting the corrected signal M ′ from L and R, and reversibly compressing the differential signals L-M ′ and R-M ′, respectively. To do. At this time, auxiliary information related to correction is also losslessly encoded. When the auxiliary information itself is output as a code, it is not necessary to perform further encoding. Since the monaural signal M has a certain degree of correlation with the stereo signals L and R, in many cases, the difference signal can be smaller in amplitude than the signals L and R itself.
Although correction will be described later with reference to FIG. 52, amplitude adjustment by multiplying a signal sample value by a coefficient, adjustment of a sample position, or a combination thereof is performed. By such correction, the amplitude of an error signal to be encoded as described later is made as small as possible. Correction can be performed using auxiliary information for each frame. The auxiliary information related to the determined correction amount is also encoded.
[0184]
Further, the stereo signals L and R and the monaural signal M are used for improving the coding efficiency of the five channels. Under general recording conditions, the signal L5c, LS5c in the 5-channel signal and the stereo signal L have a strong correlation, and the signal R5c, RS5c in the 5-channel signal and the stereo signal R have a strong correlation. Differential encoding is performed using the strong correlation between C5c and monaural signal M. That is, the difference signals (L5c-L) and (LS5c-L) of the signal L5c or LS5c in the 5-channel signal from the stereo signal L are respectively losslessly compressed and encoded from the stereo signal R to the signal R5c or RS5c in the 5-channel signal. The difference signals (R4c-R) and (RS5c-R) are losslessly encoded. Further, the differential signal (C5c-M) of the signal C5c in the 5-channel signal from the monaural signal M is also losslessly encoded.
[0185]
FIG. 47 shows an example in which the concept of the twelfth embodiment shown in FIG. 46 is more specifically configured. Here, five-channel signals C5c, L5c, R5c, LS5c, and RS5c with a sampling frequency of 192 kHz and a sampling word length (quantization accuracy) of 24 bits are supplied by signal sources 10C5, 10L5, 10R5, 10LS, and 10RS, respectively. In this example, stereo signals L and R of 192 kHz and a sample word length of 24 bits are supplied from the signal sources 10L and 10R, and a monaural signal M having a sampling frequency of 48 kHz and a sample word length of 16 bits is supplied from the signal source 10M.
The signals L5c and LS5c in the 5-channel signal are subtracted by the subtracters 13L5 and 13LS, respectively, from the corrected stereo signals L ′ from the correction units 16L5 and 16LS described later with reference to FIG. (Also referred to as an error signal or a differential signal) is losslessly encoded by the compression encoders 11L5 and 11LS. The correction information determined by the correction units 16L5 and 16LS is losslessly encoded as auxiliary information by the auxiliary information encoding units 15L5 and 15LS. Similarly, the signals R5c and RS5c in the 5-channel signal are subtracted from the corrected stereo signal R ′ from the correction units 16R5 and 16RS by the subtractors 13R5 and 13RS, respectively, and the obtained residual signal is compressed and encoded. It is losslessly encoded by 11R5 and 11RS. The parameters determined by the correction units 16R5 and 16RS are losslessly encoded as auxiliary information by the auxiliary information encoding units 15R5 and 15RS. When the correction information is output as a code by itself, it is not necessary to further encode it by the auxiliary information encoding unit.
[0186]
The monaural signal M is up-sampled by the upgrade unit 62 from 48 kHz to 192 kHz, and each sample is shifted by 8 bits in the most significant bit direction, and the same number of 8 bits, that is, 8 bits of “0” is placed on the lower side. Added to upgrade to 24-bit samples. The upgraded monaural signal is supplied to the correction units 16C5, 16L, and 16R. The signal C5c and the stereo signals L, R in the 5-channel signal are subtracted from the corrected upgraded monaural signal M ′ from the correction units 16C5, 16L, 16R by the subtractors 13C5, 13L, 13R, respectively, and the obtained error The signal is losslessly compressed and encoded by the compression encoding units 11C5, 11L, and 11R. The monaural signal M is compression encoded by the compression encoding unit 11M. The encoding in the compression encoding unit 11M may be lossless encoding or lossy encoding.
[0187]
FIG. 48 is a specific example of a decoding apparatus corresponding to the encoding apparatus of FIG. The codes losslessly encoded by the compression encoders 11C5, 11L5, 11R5, 11LS, 11RS, 11L, and 11R in FIG. 47 are encoded by the decoding / decompressing units 30C5, 30L5, 30R5, 30LS, 30RS, 30L, and 30R, respectively. The decoded signal is given to the adders 32C5, 32L5, 32R5, 32LS, 32RS, 32L, 32R and corrected from the correction units 36C5, 36L5, 36R5, 36LS, 36RS, 36L, 36R. The signals M ′, L ′, R ′, L ′, R ′, M ′, and M ′ are added to generate original sound signals C5c, L5c, R5c, LS5c, RS5c, L, and R. The code from the compression coding unit 11M of the coding device is decoded by the decoding / decompression unit 30M with a decoding algorithm corresponding to the coding of the compression coding unit 11M in the coding device of FIG. Further, the auxiliary information encoded by the encoding device is decoded by the auxiliary information decoding units 35C5, 35L5, 35R5, 35LS, 35RS, 35L, and 35R with decoding algorithms corresponding to the respective encodings, and correction units 36C5, Awarded to 36L5, 36R5, 36LS, 36RS, 36L, 36R.
[0188]
The monaural signal decoded by the decompression decoding unit 30M is output as it is as a monaural signal M having a word length of 16 bits and a sampling rate of 48 kHz, and upgraded by the upgrade unit 81 to a word length of 24 bits and a sampling rate of 192 kHz, and a correction unit 36C5 , 36L, 36R. The correction units 36C5, 36L, and 36R described later with reference to FIG. 53 are upgraded by correction parameters (a gain coefficient k and a timing adjustment amount p described later) decoded by the auxiliary information decoding units 35C5, 35L, and 35R, respectively. The monaural signal M ′ is corrected and supplied to the adders 32C5, 32L, and 32R. The adders 32C5, 32L, and 33R output a 5-channel signal center signal C5c and stereo signals L and R.
[0189]
The correction units 36L5 and 36LS correct the output (stereo signal L) of the adder 32L with the correction parameters decoded by the auxiliary information decoding units 35L5 and 35LS, and give the corrected signal L ′ to the adders 32L5 and 32LS. Similarly, the correction units 36R5 and 36RS correct the output (stereo signal R) of the adder 32R with the correction parameters decoded by the auxiliary information decoding units 35R5 and 35RS, and the corrected signal R ′ is added to the adders 32R5 and 32RS. To give. Adders 32L5, 32R5, 32LS, and 32RS output 5-channel signals L5c, R5c, LS5c, and RS5c.
13th embodiment
FIG. 49 shows the concept of the thirteenth embodiment, which generates the sum and difference of L and R for two-channel stereo signals L and R. Under general recording conditions, the sum signal (L + R) has a larger amplitude than the difference signal (LR), and the correlation between the monaural signal M captured at one position and the center signal C5c of the 5-channel signal is large. There are many. Therefore, the difference between the sum signal (L + R) and the monaural signal M and the difference between the sum signal (L + R) and the center signal C5c are losslessly encoded, and at the same time, the difference signal (L-R) is losslessly encoded as it is. The monaural signal M is also subjected to lossless encoding or lossy encoding as it is. When taking the difference between the sum signal and the monaural signal, either the half value of the sum signal or the double value of the monaural signal is used. When taking the difference between the sum signal and the center signal, either the half value of the sum signal or the double value of the center signal is used. In either case, in order to obtain a half value and a double value, the bit string indicating the value may be shifted to the lower or higher order bit by bit.
[0190]
Furthermore, since the stereo signal L is generally highly correlated with the signals L5c and LS5c in the 5-channel signal and the stereo signal R is highly correlated with the signals R5c and RS5c in the 5-channel signal, the difference between the signals L5c, LS5c and the signal L and The difference between the signals R5c, RS5c and the signal R is losslessly encoded. In the following description, the difference signal (LR) and the sum signal (L + R) are encoded. However, even if one of them is divided by 2, the difference signal (LR) is encoded. Since the least significant bit is the same as the least significant bit of the sum signal (L + R), the one divided by 2 is doubled when restored (1 bit lower, that is, shifted in the MSB direction), and the least significant bit If the bit is the same as the least significant bit that is not divided by 2, the difference signal (LR) and the sum signal (L + R) can be reconstructed completely without distortion. At the time of signal reconstruction, decoding of the monaural signal M, decoding of the sum signal (L + R) and decoding of the difference signal (L-R) can reconstruct all of the 5-channel signal, stereo signal, and monaural signal.
[0191]
FIG. 50 shows a specific configuration example of the thirteenth embodiment based on the concept shown in FIG. Of the five channel signals, the configuration for encoding processing for signals L5c, R5c, LS5c, and RS5c is the same as in FIG. 47, but the encoding for center signal C5c is not a monaural signal but a sum signal (L The difference is that encoding is performed by taking the difference from + R). In FIG. 50, stereo signals L and R are subtracted by a subtractor 78S to generate a difference signal (L-R), and losslessly encoded by the compression encoding unit 11L. Further, the stereo signals L and R are added by an adder 78A, and the added output (L + R) is a monaural signal M upgraded by a subtractor 13M to a sample word length of 24 bits from the upgrade unit 62 and a sampling rate of 192 kHz. The error signal is losslessly encoded by the compression encoding unit 11R. The output signal (L + R) from the adder 78A is also corrected by the correction unit 16C5, supplied to the subtractor 13C5, and subtracted from the center signal C5c in the 5-channel signal. The configuration and operation of each correction unit are the same as those of the correction units 16C5, 16L5, 16R5, 16LS, 16RS, 16L, and 16R used in FIG. 47, and will be described later with reference to FIG.
[0192]
FIG. 51 shows a configuration of a decoding apparatus corresponding to the encoding apparatus of FIG. In this example, the monaural signal M having a sample word length of 16 bits and a sampling rate of 48 kHz decoded by the decompression decoding unit 30M is output as it is, and the upgrade unit 81 upgrades the signal to a sample word length of 24 bits and a sampling rate of 192 kHz. And supplied to the adder 32M. The adder 32M adds the decoded error signal from the decompression decoding unit 30R and the upgraded monaural signal M ′ to generate a stereo sum signal (L + R). This sum signal (L + R) is corrected by the correction unit 36C5 using the decoded auxiliary information from the decoding unit 35C5 (described later with reference to FIG. 53), and provided to the adder 32C5. The adder 32C5 adds the decoded error signal from the decompression decoding unit 30C5 and the corrected sum signal (L + R), and outputs a 5-channel center signal C5c.
[0193]
The difference signal (LR) decoded by the decompression decoding unit 30L and the sum signal (L + R) from the adder 32M are added by an adder 97A and divided by 2 to generate a stereo signal L. Subtracted by the subtractor 97S and divided by 2, a stereo signal R is generated. The processing for the error signal decoded by the decompression decoding units 30L5, 30R5, 30LS, and 30RS is the same as that shown in FIG. 50, and 5-channel signals C5c, L5c, R5c, LS5c, and RS5c are generated by the processing.
Each of the correction units 16C5, 16L5, 16R5, 16LS, 16RS, 16L, and 16R in FIGS. 47 and 50 has the same configuration. For example, as shown in FIG. This is basically the same as that described above, and includes a gain adjusting unit 16A, a timing adjusting unit 16B, and an error minimizing unit 16C. The channel signal from the signal source is multiplied by the coefficient k given from the error minimizing unit 16C by the gain adjusting unit 16A, and the timing adjusting unit 16B advances the shift amount p specified by the sample timing by the error minimizing unit 16C. The signal is shifted in the direction or the delay direction and given to a subtracter 13mn (representing 13C5, 13L5,...). The error minimizing unit 16C selects and determines the coefficient k and the shift amount p that minimize the power of the output error of the subtractor 13mn, for example, from a predetermined set of (k, p) values. An index representing the set of the determined coefficient k and shift amount p is given as correction information to the auxiliary information encoding unit 15mn (representing 15C5, 15L5,...) And output as an auxiliary code.
[0194]
Each of the correction units 36C5, 36L5, 36R5, 36LS, 36RS, and 36L in FIGS. 48 and 51 has the same configuration as shown in FIG. 23 as shown in FIG. It is basically the same, and is composed of a gain adjustment unit 36A and a timing adjustment unit 36B. The amplitude of the signal sample is determined by the gain adjustment coefficient k and the time shift amount p as a decoded correction parameter from the correction information decoding unit 35mn. The gain adjustment unit 36A multiplies by k, the timing of the sample is shifted by p by the timing adjustment unit 36B, and is supplied to the adder 32mn.
14th embodiment
FIG. 54 shows the concept of a fourteenth embodiment of the encoding method according to the present invention. In the fourteenth embodiment, inter-channel orthogonal transformation is performed on a 5-channel signal to calculate a differential signal from the signals of other channels. The orthogonal transform between channels corresponds to a conversion to a frequency domain across channels, and Nc × Nc for a vector whose number is the number of channels Nc whose elements are one sample of each channel at the same time. Equivalent to multiplying an orthogonal matrix. Examples of the inter-channel orthogonal transform include multiplication of a principal component analysis matrix or Hadamard matrix, a DCT (digital cosine transform) matrix, and a DFT (digital Fourier transform) matrix between channels.
[0195]
By this conversion, a vector of input samples is converted into a vector composed of frequency domain sample elements. In the following description, F0, F1, F2, F3, and F4 are assumed in order from the lowest frequency of the converted output sample element. The component F0 having the lowest frequency due to the orthogonal transformation is a component that is the sum of the five-channel input samples, and generally has higher power than the component having a higher frequency. For example, when the correlation between channels such as a multi-channel music signal is high, energy is concentrated on the low frequency side and the energy on the high frequency side is small. Therefore, when the inter-channel orthogonal transform is performed, the amplitude on the signal F0 side having the lowest frequency becomes large.
[0196]
A signal having the largest amplitude in the inter-channel orthogonal transform outputs F0 to F4, for example, F0, is expected to have a high correlation with the monaural signal M. In addition, it is expected that a signal having the second largest amplitude, for example, F1 has a high correlation with the stereo difference signal (L-R). Therefore, the monaural signal M is corrected and the difference from the orthogonal transformation output signal F0 having the largest amplitude is losslessly encoded, the difference signal (LR) is corrected and the difference from the orthogonal transformation output signal F1 having the second largest amplitude is obtained. Lossless encoding.
FIG. 55 shows an example of an encoding apparatus that specifically configures the concept of the encoding method of the fourteenth embodiment shown in FIG. The configuration and connection relationship of the correction units 16A and 16B in FIG. 55 are the same as those shown in FIG. 52, and in order to simplify the drawing, the connection from the output of the subtractor to the correction unit, and the auxiliary information encoding unit 15mn The notation is omitted. The 5-channel signals C5c, L5c, R5c, LS5c, and RS5c are subjected to inter-channel orthogonal transform by the orthogonal transform unit 19 to generate converted output signals F0 to F4. For the stereo signals L and R, the difference signal (L-R) and the sum signal (L + R) are generated by the subtractor 78S and the adder 78A, as in the case of FIG. The difference signal (L-R) is losslessly encoded by the compression encoding unit 11L.
[0197]
The monaural signal M is losslessly or irreversibly encoded by the compression encoder 11M, and the upgrade unit 62 upgrades the sampling frequency from 48 kHz to 192 kHz and the quantization accuracy from 16 bits to 24 bits. The upgraded monaural signal M is subtracted from the sum signal (L + R) by the subtractor 13M, and the error signal is losslessly compressed by the compression encoder 11R. The upgraded monaural signal M is corrected by the correction unit 16A, and the error signal obtained by subtracting the corrected signal from the signal F0 having the largest amplitude among the signals F0 to F4 by the subtractor 13A is a compression code. The encoding unit 11C5 performs lossless encoding.
[0198]
On the other hand, the difference signal (LR) is corrected by the correction unit 16B, subtracted from the signal F1 having the second largest amplitude among the signals F0 to F4 by the subtractor 13B, and the obtained error signal is encoded by the compression encoding unit 11C5. It becomes. The other converted output signals F2 to F4 are encoded by the compression encoding units 11R5, 11LS, and 11RS. It should be noted that the outputs F0, F1,... Of the inter-channel orthogonal transform unit 19 do not always have the maximum amplitude F0 depending on the input signal, and F1 does not have the second largest amplitude. It suffices to determine in advance for which frequency the error signal is generated in accordance with the tendency.
FIG. 56 shows a decoding apparatus corresponding to FIG. The signal decoded by the decompression decoding unit 30M is output as it is as a monaural signal M with a sampling frequency of 48 kHz and a quantization accuracy of 16 bits, and is upgraded by the upgrade unit 81 to a signal with a sampling frequency of 192 kHz and a quantization accuracy of 24 bits. . The error signal decoded by the decompression decoding unit 30R is added to the monaural signal M upgraded by the adder 32M to generate a sum signal (L + R). The sum signal (L + R) and the difference signal (LR) decoded by the decoding unit 30L are added and subtracted by the adder 97A and the subtractor 97S, respectively, and divided by 2, respectively. Generated.
[0199]
The upgraded monaural signal M and difference signal (LR) are corrected by the correction units 36A and 36B, respectively, and the correction outputs are added to the decoded signals from the decoding units 30C5 and 30L5 by the adders 32A and 32B, respectively. , F1 is generated. These signals F0, F1 and signals F2, F3, F4 decoded by the decoding units 30R5, 30LS, 30RS are orthogonally inverse-transformed by an interchannel orthogonal inverse transform unit 39, and time-domain 5-channel signals C5c, L5c, R5c, LS5c and RS5c are generated.
47 and 50, the 5-channel signal is described as having a sampling frequency of 192 kHz and an amplitude resolution of 24 bits. Compared to this, the sampling frequency of the monaural signal M is 48 kHz and the amplitude resolution is 16 bits. Although low, the monaural signal M is upgraded to a sampling frequency of 192 kHz and an amplitude resolution of 24 bits in the upgrade unit 62, and the difference from the center signal C5c of the 5-channel signal is losslessly encoded.
[0200]
According to the above-described embodiment, reversible encoding with different numbers of channels can be performed uniformly, and the compression ratio of the entire system is increased as compared with the case where each channel is encoded separately without performing differential encoding. be able to. That is, since the correlation between the 5-channel signal and the stereo signal can be removed by taking the difference between the 5-channel signal and the stereo signal, the code bit string has a smaller information amount than the total when the 5-channel signal and the stereo signal are independently compressed. Also, the amount of traffic on the network is monitored, and when the amount of traffic exceeds a predetermined value, for example, transmission of 5 channel signals is stopped and only stereo and monaural signals are transmitted. It is also possible to increase / decrease the number of channels according to the band fluctuation.
15th embodiment
A lossless encoding method that does not allow distortion is known as a method for compressing information such as sound and images. Depending on the application, the sampling frequency and quantization accuracy may differ, and as shown in the previous example, if multiple types of sampling rate and amplitude resolution signals are prepared in advance, depending on the application, preference, and network conditions Lossless compression encoding that can correspond to one desired set of one of a plurality of types of sampling frequencies and a plurality of types of amplitude resolution is possible. A fifteenth embodiment of such compression encoding will be described below.
[0201]
Here, as shown in FIG. 33, the sampling frequency and the amplitude quantization accuracy are two-dimensionally hierarchized to encode a signal, so that higher-level coding becomes lower-level coding. So that it can be expressed as The original sound with the specified sampling frequency and quantization accuracy can be reproduced, and can be hierarchically integrated into multiple types of encoding. In particular, the coding efficiency can be improved by combining the low frequency component of the signal of the sampling frequency of the lower layer and the high frequency component of the signal of the lower amplitude resolution, and selecting or synthesizing the difference from the original sound.
When the two-dimensional hierarchization of the sampling frequency and the quantization accuracy is performed with the hierarchical structure shown in FIG. 33, the quantization accuracy is the number of layers P = 3, and the sampling rate is the number of layers Q. = 3, 48, 96 and 192kHz respectively. P × Q = 9 types of original sounds A, B, C, D, E, F, G, H, and I are given and encoded with as little information as possible, and the original original sound is decoded without distortion. The attributes of the original sound signal are hierarchized into nine types of P × Q = 3 × 3, and the sampling frequency and the quantization accuracy allow the higher order signal to be configured using the lower order signal.
[0202]
A 16-bit quantization accuracy signal encodes an error signal from a signal after up-sampling a signal with the same quantization accuracy at a lower sampling frequency. The 48 kHz signal is obtained by converting the lower quantization accuracy signal to the same higher accuracy and encoding an error signal with the accuracy converted signal. When lower signals exist in the sampling frequency direction and the quantization accuracy direction, one of the two lower signals can be selected. For example, when encoding a signal E having a sampling frequency of 96 kHz and a quantization accuracy of 20 bits, an error is detected among the signal B having a sampling frequency of 96 kHz and a quantization accuracy of 16 bits and the signal D having a sampling frequency of 48 kHz and a quantization accuracy of 20 bits. You may choose the one where the power of a signal becomes the minimum.
[0203]
FIG. 57 shows an encoding apparatus according to the fifteenth embodiment. In this example, the original sound signal S with a sampling frequency of 192 kHz as the original sound and a quantization accuracy of 24 bits, 20 bits, and 16 bits, respectively._3,3, S_2,3, S_1,3, Output signal source 10_3,3, Ten_2,3, Ten_1,3Original signal S with a sampling frequency of 96kHz and quantization accuracy of 24bit, 20bit and 16bit respectively_3,2, S_2,2, S_1,2Output signal source 10_3,2, Ten_2,2, Ten_1,2Original signal S with a sampling frequency of 48kHz and quantization accuracy of 24bit, 20bit and 16bit respectively._3,1, S_2,1, S_1,1Output signal source 10_3,1, Ten_2,1, Ten_1,1Is provided. Each signal source 10_3,3, Ten_2,3, Ten_1,3Output original sound signal S from_3,3, S_2,3, S_1,3Is the difference module 13_3,3, 13_2,3, 13_1,3At each signal S_3,3, S_2,3, S_1,3The lower signal is upgraded to the same level and the original signal S_3,3, S_2,3, S_1,3And the compression encoding unit 11_3,3, 11_2,3, 11_1,3To lossless encoding and output.
[0204]
Similarly, signal source 10_3,2, Ten_2,2, Ten_1,2Original sound signal S from_3,2, S_2,2, S_1,2Is the difference module 13_3,2, 13_2,2, 13_1,2At each signal S_3,2, S_2,2, S_1,2The lower signal is upgraded to the same level and the original signal S_3,2, S_2,2, S_1,2And the compression encoding unit 11_3,2, 11_2,2, 11_1,2To lossless encoding and output. Signal source 10_3,1, Ten_2,1Original sound signal S from_3,1, S_2,1Is the difference module 13_3,1, 13_2,1At signal S_3,1, S_2,1Upgrade signal of lower order to signal S_3,1, S_2,1And the compression encoding unit 11_3,1, 11_2,1To lossless encoding and output. Signal source 10_1,1Original sound signal S_1,1Since there is no signal lower than that, the compression encoding unit 11 is used as it is._1,1Output with lossless encoding or lossy encoding.
[0205]
Each difference module 13 in the encoding device of FIG._3,3, 13_3,2, 13_2,3, 13_2,2The signal source 10_{m, n}Original sound signal S from (m = 2, 3; n = 2, 3)_{m, n}And the lower signal S_{m-1, n}Or S_{m, n-1}And the compression encoding unit 11_{m, n}Output to. Lower signal S_{m-1, n}Or S_{m, n-1}After upsampling, accuracy adjustment, signal source 10_{m, n}Original sound signal S from_{m, n}A signal as close as possible to. At this time, one of two types of signals, that is, a signal having the same sampling frequency and lower quantization accuracy and a signal having the same quantization accuracy and lower sampling frequency is selected. The signal selection information is output as auxiliary information.
[0206]
For example, the difference module 13_3,3Is the original sound signal S_3,3The same quantization accuracy as 24 bits and the sampling frequency is lower, that is, 96 kHz original sound signal S_3,2And the original sound signal S_3,3Is the same sampling frequency as 192 kHz, but the quantization accuracy is low, that is, 20-bit original sound signal S_2,3As will be described later with reference to FIG. 58, either one is selected and the original sound signal S is selected._3,3Difference with is generated. Low-frequency signal (original sound signal S_{m, n}Only low frequency components with a half-value of the sampling frequency as the upper limit), and high-frequency signals (original sound signal S) that are expected to have relatively small errors for low-order quantization accuracy signals._{m, n}Only high-frequency components with a lower limit of half the sampling frequency).
[0207]
Alternatively, in addition to the above selection, two types of signals can be synthesized. Synthesis includes averaging, weighted addition, weighted addition that varies with time, and the like. For example, as described later with reference to FIG._3,2, S_2,3Weighted average and original sound signal S_3,3The difference between and is generated and output. Difference module 13_2,3, 13_3,2, 13_2,2Is the same configuration.
Difference module 13_1,3, 13_1,2, 13_3,1, 13_3,2For those input source signal S_1,3, S_1,2, S_3,1, S_2,1Since there is no signal of lower sampling frequency, the original sound signal S of lower quantization accuracy_1,2, S_1,1, S_2,1, S_1,1Only given.
[0208]
In addition, instead of selecting the entire frame, it is possible to select the one in which the difference power is smaller for each subframe or for each of a plurality of frames. Difference module 13_1,3, 13_1,2, 13_3,1, 13_2,1Then, the signal S from the signal source_1,3, S_1,2, S_3,1, S_2,1On the other hand, the difference from each one signal lower than them is taken and given to the corresponding compression encoding unit.
FIG. 58 shows each difference module 13_3,3, 13_2,3, 13_3,2, 13_2,2Any one of the 13_{m, n}The input original sound signal S_{m, n}(m = 2, 3; n = 2, 3)_{m, n-2}And S_{m-1, n}Are input to the upsampling unit 13A and the accuracy conversion unit 13C, respectively. Lower-order signal S in upsampling unit 13A_{m, n-1}Is the upsampled original sound signal S_{m, n}And is supplied to the selector 13E through the low-pass filter 13B having a cutoff frequency whose upper limit is half the sampling frequency. Low-order signal S in precision converter 13C_{m-1, n}Is shifted by 4 bits in the upper bit direction, so that a 4-bit "0" is added to the signal S._{m, n}The same quantization accuracy as the original sound signal S_{m, n}Is supplied to the selector 13E through the high-pass filter 13D having a cutoff frequency with the half value of the sampling frequency as the lower limit. The signal selected by the selector 13E is input to the input signal S by the subtractor 13S._{m, n}And subtracted. The error minimizing unit 13F controls the selector 13E so as to select the signal with the smallest output error power of the subtractor 13S, and outputs selection information indicating which signal is selected as auxiliary information. The auxiliary information corresponds to the corresponding compression encoding unit 11 as indicated by a broken line in FIG._{m, n}And encoded with the error signal.
[0209]
FIG. 59 shows the original sound signal S_{m, n}Lower signal S_{m, n-1}And S_{m-1, n}Difference module for weighted average of 13_{m, n}A configuration example of (m = 2, 3; n = 2, 3) is shown. 58, weight multipliers 13G and 13H and an adder 13K are provided in place of the selector 13E in the difference module of FIG. 58, and the weight coefficients w1 and w2 set by the error minimizing unit 13F and the original sound signal S_{m, n}The output from the low-pass filter 13B and the output from the high-pass filter 13D that have a cutoff frequency with the upper limit of half the sampling frequency as the upper limit are multiplied, and the two multiplication results are added and subtracted by the adder 13K. Is provided to the container 13S. In the error minimizing unit 13F, a plurality of predetermined weight coefficients (w1, w2) are stored in a storage unit (not shown) as a weight coefficient table corresponding to the code for each set, and the error The minimizing unit 13F selects a set of weighting factors w1, w2 that minimizes the power of the output error of the subtractor 13S from the weighting factor table, and uses the code corresponding to the weighting factor w1, w2 of the set as auxiliary information. Output. Difference module 13 in FIG._1,3, 13_1,2, 13_3,1, 13_2,1The lower order signal is S_1,2, S_1,1, S_2,1, S_1,1Therefore, the upsampling unit 13A, the low-pass filter 13B, the selector 13E, and the error minimizing unit 13F in FIG. 58 are not necessary, and the output of the high-pass filter 13D is directly supplied to the subtractor 13S. Similarly, in the case of FIG. 59, these difference modules provide the output from the high-pass filter 13D as it is to the subtractor 13S.
[0210]
FIG. 60 shows a configuration example of a decoding apparatus corresponding to the encoding apparatus of FIG. The input codes corresponding to the sound source signals I, F, C, H, E, B, G, D, and A are each decoded together with the auxiliary information by the decompression decoding unit, and the decompression decoding unit 30_1,1The decoding output from is the lowest decoded original sound signal S_1,1Are output as the addition module 32_1,2, 32_2,1Given to. Other decryption units 30_3,3~ 30_2,1The decoding error signal of the addition module 32_3,3~ 32_2,1Given to. Addition module 32_3,3, 322_2,3, 32_3,2, 32_2,2Is the sum of each decoded error signal and one of the two upgraded original sound signals from the lower level or the two weighted averages to add the original sound signal S_3,3, S_2,3, S_3,2, S_2,2Is generated. Addition module 32_1,3, 32_3,1, 32_2,1Is the original sound signal S by adding the decoded error signal and the lower-grade decoded original sound signal._1,3, S_2,3, S_2,1, S_3,2Is generated.
[0211]
Adder module 32 in FIG._3,3, 32_2,3, 32_3,2, 32_2,2Any one of 32_{m, n}This is expressed as (m = 2, 3; n = 2, 3), and its configuration is shown in FIG. The larger the value of m or n, the higher the sampling frequency or quantization accuracy (that is, the higher attribute). This example is the difference module 13 of FIG._{m, n}Corresponds to the case where one of the two lower signals is selected and used. Lower level original sound signal S_{m, n-1}, S_{m-1, n}Are upsampler 32A and precision converter 32C._{m, n}Same sampling rate as S_{m, n}And is supplied to the selector 32E through the low-pass filter 32B and the high-pass filter 32D, respectively. The control unit 32F switches the selector 32E according to the selection information indicating which one of the two lower signals is selected as the decoded auxiliary information, and the adder 32 adds the selected signal to the decoded error signal to generate the original sound. Signal S_{m, n}Is generated. Other addition module 32_1,3, 32_1,2, 32_3,1, 32_2,1In FIG. 61, the upsampling unit 32A, the low-pass filter 32B, the selector 32E, and the control unit 32F are removed, and the output from the high-pass filter 32D is supplied to the adder 32S.
[0212]
FIG. 62 shows the addition module 32 in FIG._{m, n}A configuration of (m = 2, 3; n = 2, 3) corresponding to the difference module of FIG. 59 is shown. In this case, coefficient multipliers 32G and 32H and an adder 32K are provided instead of the selector 32E in FIG. Upgraded lower signal S_{m, n-1}, S_{m-1, n}Are multiplied by the multipliers 32G and 32H, and the multiplication results are added by the adder 32K. The addition result is added by the adder 32 to the decompression decoding unit 30._{m, n}Is added to the decoding error signal from the original sound signal S_{m, n}Is generated. Other addition module 32_1,3, 32_1,2, 32_3,1, 32_2,1In FIG. 62, the upsampling unit 32A, the low-pass filter 32B, the multiplier 32G, and the adder 32K are removed, and the output of the multiplier 32H is supplied to the adder 32S.
[0213]
58 and 59, as shown in FIGS. 63 and 64, the output side of the upsampling unit 13A and the accuracy conversion unit 13C is a low-pass filter 13B1, a high-pass filter 13B2, and a low-pass filter, respectively. 13D1 and high-pass filter 13D2 are provided, and a signal S with a lower sampling rate is provided._{m, n-1}For subordinate quantization accuracy signal S_{m-1, n}On the other hand, after the upgrade to the upper level, the cut-off frequency is separated into a low frequency component and a high frequency component from the half value of the higher sampling frequency, and the output error power from the subtractor 13 is small among these filter outputs. The pair of filter outputs is determined by the error minimizing unit F13 and selected by the selector 13E (FIG. 63). Or, as shown in FIG. 64, the outputs of all filters 13B1, 13B2, 13D1, and 13D2 are multiplied by weighting factors w11, w12, w21, and w22 by multipliers 13G1, 13G2, 13H1, and 13H2, and the multiplication results are added. It is also possible to perform weighted averaging by adding in the unit 13K, and determine the weighting factors w11, w12, w21, w22 in the error minimizing unit 13F so that the power of the output error of the subtracter 13 is minimized. In this case, the error minimizing unit 13F is provided with a storage unit (not shown), and the values of a plurality of sets of weighting coefficients (w11, w12, w21, w22) correspond to the codes representing the respective sets. A table stored in the storage unit as a table may be searched for and determined from the set that minimizes the power of the error signal, and a code corresponding to the set may be output.
[0214]
61 and 62 in the decoding apparatus shown in FIGS._{m, n}Similarly to FIGS. 63 and 64, the output of the upsampling unit 32A is converted to the signal S by the low-pass filter 32B1 and the high-pass filter 32B2 as shown in FIGS._{m, n}As the cut-off frequency, the half-value of the sampling frequency is separated into two components, a high-frequency component and a low-frequency component. Similarly, the output from the precision conversion unit 32C is also converted to the signal S by the low-pass filter 32D1 and the high-pass filter 32D2._{m, n}Is divided into two components, a high frequency component and a low frequency component, from the half value of the sampling frequency, and the selector 32E selects the output of these filters according to the decoded selection information (FIG. 65) or multiplies the weighting factor. The weighted average is obtained by multiplying the respective filter outputs by the weight coefficients w11, w12, w21, w22 represented by the input codes in the units 32G11, 32G12, 32G21, 32G22 and adding them by the adder 32K (FIG. 66). Can be configured.
[0215]
FIG. 67 shows the signal S of the lower sampling frequency._{m, n-1}Signal S with lower frequency components and lower quantization accuracy than the cutoff frequency of_{m-1, n}An example of simply synthesizing the high-frequency component of is shown. Signal S of the lower sampling frequency shown in FIG. 67A_{m, n-1}The Nth (N = 0, 1, 2,...) Sample amplitude values are arranged as they are at even-numbered 2N sample positions of the double sampling frequency shown in FIG. 67B. Next, the lower quantization accuracy signal S in FIG. 67C._{m-1, n}The sample positions are aligned and the odd-numbered samples are arranged at the corresponding positions.
Or even-numbered samples are arranged in the same way as above, and odd-numbered samples are signals S of lower sample frequencies._{m, n-1}It is possible to arrange a sample of a signal obtained by selecting a weighted addition of a signal obtained by up-sampling and a signal having a lower quantization accuracy or one of them.
Sixteenth embodiment
In the fifteenth embodiment, encoding and decoding using the two-dimensional hierarchization of the quantization accuracy and sampling frequency shown in FIGS. 33 and 34 have been described. In the sixteenth embodiment, FIGS. 2 is used to encode the error signal in the frequency domain. This embodiment will be described with reference to FIG.
[0216]
As shown in FIG. 68, the encoding apparatus according to the sixteenth embodiment has a sound source 60 similar to that shown in FIG. 44 based on the signal hierarchical structure of FIGS._1,1~ 60_3,3have. In this embodiment, the sound source 60 has sampling frequencies of 96 kHz and 192 kHz._1,2~ 60_3,3Is output from the orthogonal transform unit 19 for each predetermined number of samples (transformation length) corresponding to each sampling frequency._1,2~ 19_3,3Is converted to the same number of samples in the frequency domain, and the corresponding subtractor 63_1,2~ 63_3,3Given to.
Subtraction unit 63_1,3, 63_2,3, 63_3,3The sound source 60 has a lower sampling frequency of 96 kHz._1,2, 60_2,2, 60_3,2The digital signal from the orthogonal transform unit 19_1,2, 19_2,2, 19_3,2The frequency domain signals converted and output by the_1,3, 16_2,3, 16_3,3The orthogonal transform unit 19_1,3, 19_2,3, 19_3,3The difference from the frequency domain signal from_1,3, Δ_2,3, Δ_3,3They are generated as_1,3, 61_2,3, 61_3,3Is compressed and encoded and output as codes C, K, and M. Signal S with sampling frequency 48kHz_1,1, S_2,1It is natural to convert the accuracy of quantization accuracy to the time domain._1,1, 60_2,116-bit and 20-bit digital signal S from_1,1, S_2,1Is the precision converter 62_1,1, 62_2,1Given to.
[0217]
The lowest digital signal S_1,1Is the orthogonal transform unit 19_1,1And an orthogonal transform unit 19_1,1The frequency domain signal converted by the_1,1Is compressed and encoded and output as code A.
Precision converter 62_1,1Is the given digital signal S_1,1The quantization precision is converted from 16 bits to 20 bits by adding “0” of 4 bits at the lower position to the least significant bit of each of the samples._2,1To give. Subtraction unit 63_2,1Is the precision converted signal and the sound source 60_2,1Digital signal S from_2,1Is generated as an error signal, and the orthogonal transformation unit 19_2,1To give. Orthogonal transformation unit 19_2,1Is the error signal Δ_2,1Convert to compression unit 61_2,1Compression unit 61_2,1Is the given error signal Δ_2,1Is compressed and output as a code D. Similarly, sound source 60_3,1Digital signal S from_3,1The precision converter 62_2,1The difference from the signal whose quantization accuracy is converted from 20 bits to 24 bits is the subtractor 63._3,1The error signal generated and generated by the orthogonal transform unit 19_3,1In the frequency domain error signal Δ_3,1The compression unit 61_3,1Is compressed and encoded and output as a code G.
[0218]
Signal S with a sampling frequency of 96kHz and quantization accuracy of 16 bits_1,2As is clear from FIG. 42, the signal S includes the signal components A and B and has a quantization accuracy of 20 bits._2,2Includes signal components A, D, and J, and a signal S3,2 having a quantization accuracy of 24 bits includes signal components A, D, G, and L. Accordingly, the subtracting unit 63 so that the respective signal components of the codes B, J, and L are obtained._1,2, 63_2,2, 63_3,2The frequency domain difference calculation is performed at. That is, the signal S with a quantization accuracy of 16 bits_1,1Orthogonal transformation unit 19 for_1,1The conversion signal by the correction unit 16_1,2Through subtractor 63_1,2And a signal S with a sampling frequency of 96 kHz_1,2To the frequency domain signal, and the difference is the frequency domain error signal Δ_1,2As compression part 61_1,2Is compressed and encoded and output as code B.
[0219]
Similarly, the digital signal S_2,2Is orthogonally transformed and subtracted 63_2,2And an orthogonal transform unit 19_1,1, 19_2,1The frequency domain signal from the subtracting unit 63 via the correction unit_2,2Signal S_2,2The frequency domain error signal Δ is subtracted from the frequency domain component of_2,2Is generated and its error signal Δ_2,2Compression part 61_2,2Is compressed and encoded and output as a code J. Further, the subtraction unit 63_3,2In the digital signal S_3,2Frequency domain signal to digital signal S_1,1Frequency domain component and frequency domain error signal Δ_{2, 1}, Δ_3,1By subtracting the error signal Δ_3,2The compression unit 61_3,2Is compressed and encoded and output as a code L.
[0220]
Orthogonal transformation unit 19_1,2, 19_2,2, 19_3,2The frequency domain signals from_1,3, 16_2,3, 16_3,3Through subtractor 63_1,3, 63_2,3, 63_3,3And an orthogonal transform unit 19_1,3, 19_2,3, 19_3,3Is subtracted from the frequency domain signal from the error signal Δ_1,3, Δ_2,3, Δ_3,3Is obtained. These error signals are compression encoded by the compression unit and output as codes C, K, and M.
Each orthogonal transform unit 19_1,1~ 19_3,3In order to perform reproduction without distortion, for example, DCT (Discrete Cosine Transform) or MDCT (Modified Discrete Cosine Transform) of an integer coefficient can be used. Further, by determining the conversion lengths according to the sampling frequency, an error signal between different sampling frequencies can be reduced. For example, the conversion lengths for digital signals having sampling frequencies of 48 kHz, 96 kHz, and 192 kHz are N points, 2N points, and 4N points, respectively, in terms of the number of samples. Of the 2N point signals in the frequency domain obtained by converting the 2N point samples of the sampling frequency 96 kHz signal, the lower N points are the N in the frequency domain obtained by converting the N point samples of the sampling frequency 48 kHz signal. It is similar to the point signal, and if the difference is taken as it is, the error signal can be reduced. The relationship between the sampling frequency 192 kHz signal and the sampling frequency 96 kHz signal is the same.
[0221]
Thus, what is characteristic of this embodiment is that an error signal can be generated without up-sampling between signals of different sampling frequencies in order to obtain an error signal in the frequency domain. Each correction unit 16_1,2, 16_2,2, 16_3,2, 16_1,3, 16_2,3, 16_3,3For example, as described with reference to FIG. 52, the gain of the frequency domain signal is adjusted so that the power (spectrum power) of the error signal is minimized, and a code representing the gain is output as auxiliary information. This gain adjustment may be performed by giving a weighting factor to each sample in the frequency domain.
An example of a decoding apparatus corresponding to the encoding apparatus of FIG. 68 is shown in FIG. Input codes A, D, G, B, J, L, C, K, M are expansion units 80, respectively._1,1~ 80_3,3The decompression decoding is performed and the lowest signal in the frequency domain and the error signal Δ_2,1~ Δ_3,3Is generated. Lowermost extension 80_1,1The decoded signal from the orthogonal inverse transform unit 39_1,1Is converted to a time domain signal by the lowest digital signal S_1,1Is played. Error signal Δ in the frequency domain_2,1Is the orthogonal inverse transform unit 39_2,1Is added to the adder 82 after being converted into a time domain error signal at_2,1Precision conversion unit 81_1,1Is added to the signal whose quantization accuracy is increased to 20 bits and added to the digital signal S._2,1Is played. This playback signal S_2,1Is the precision converter 81_2,1As a result, the quantization accuracy is increased to 24 bits and the adder 82_3,1Given to. Error signal Δ_3,1Is the orthogonal inverse transform unit 39_3,1Converted into a time domain error signal by the adder 82._3,1The digital signal S is added to the signal whose quantization accuracy has been increased by_3,1Is played. The orthogonal inverse transform unit 39_1,1~ 39_3,3Is the orthogonal transform unit 19 in FIG._1,1~ 19_3,3A process reverse to the above process is performed to convert the frequency domain signal into a time domain signal.
[0222]
Frequency domain error signal Δ_1,2Is the extension 80_1,2Decoded by the correction unit 36._1,2The frequency domain signal corrected by_1,2The orthogonal inverse transform unit 39_1,2Is converted to a time domain signal and digital signal S_1,2Is played. Similarly, the frequency domain error signal Δ_2,2The extension 80_1,1, 80_2,1The signal from the correction unit 36_2,282 after each correction_2,2And the addition result is an orthogonal inverse transform unit S._2,2Converted to a time domain signal by the digital signal S_2,2Is played. In addition, the frequency domain error signal Δ_3,2The extension 80_1,1, 80_2,1, 80_3,1The signal from the correction unit 36_3,282 after each correction_3,2And the addition result is the orthogonal inverse transform unit 39._3,2Converted to a time domain signal by the digital signal S_3,2Is played. Frequency domain error signal Δ_1,3, Δ_2,3, Δ_3,3Is the adder 82_1,2, 82_2,2, 82_3,2Frequency domain signals from the_1,3, 36_2,3, 36_3,3Are respectively corrected by the adder 82._1,3, 82_2,3, 82_3,3And the addition result is an orthogonal inverse transform unit 39._1,3, 39_2,3, 39_3,3Converted to a time domain signal by the digital signal S_1,3, S_2,3, S_3,3Are each played. Each correction unit 36_1,2, 36_2,2, 36_3,2, 36_1,3, 36_2,3, 36_3,368 using the parameters indicated by the input auxiliary information._1,2, 16_2,2, 16_3,2, 16_1,3, 16_2,3, 16_3,3For example, gain correction is performed.
[0223]
In the embodiment of FIG. 68, the digital signal S having the lowest sampling frequency of 48 kHz is used._2,1, S_3,1In the modified example shown in FIG. 70, the digital signal S having the lowest sampling frequency of 48 kHz is shown._2,1, S_3,1An error signal with respect to is also obtained in the frequency domain. Other configurations are the same as those in FIG.
In this case, the accuracy converter 62_1,1, 62_2,1Is a digital signal S having a quantization accuracy of 16 bits and 20 bits._1,1, S_2,1The orthogonal transform unit 19_1,1, 19_2,1The frequency domain signal converted by the above is given, and a 4-bit "0" is added to the least significant part of each frequency domain sample to improve the quantization accuracy by one layer, thereby subtracting unit 63 as 20 bits and 24 bits._2,1, 63_3,1To give. Subtraction unit 63_2,1, 63_3,1Is the digital signal S_2,1, S_3,1Is the orthogonal transform unit 19_2,1, 19_3,1Converted into a frequency domain signal by the accuracy conversion unit 62_1,1, 62_2,1Are subtracted from the precision-converted signal from the error signal Δ_2,1, Δ_3,1Is generated.
[0224]
The digital signal S with a sampling frequency of 48 kHz_1,1, S_2,1, S_3,1Is converted into a frequency domain signal, and the correction unit 16_1,2, 16_2,2, 16_3,2Through subtractor 63_1,2, 63_2,2, 63_3,2And a digital signal S with a sampling frequency of 96 kHz._1,2, S_2,2, S_3,2Is the orthogonal transform unit 19_1,2, 19_2,2, 19_3,2Is subtracted from the frequency domain signal transformed by the error signal Δ_1,2, Δ_2,2, Δ_3,2Is generated. Other configurations and operations are the same as those in FIG.
FIG. 71 shows a decoding apparatus corresponding to the encoding apparatus according to the modified embodiment of FIG. Also in this embodiment, the accuracy conversion for the decoded signal at the lowest sampling frequency is performed in the frequency domain. That is, the expansion part 80_1,1Thus, the input code A is decompressed and decoded to obtain a frequency domain signal._1,1And an orthogonal inverse transform unit 39_1,1Is converted to a time domain signal and the digital signal S_1,1Is played. Other configurations are the same as those in FIG.
[0225]
Extension part 80_2,1, 80_3,1, 80_1,2, 80_2,2, 80_3,2, 80_1,3, 80_2,3, 80_3,3Is the decompression decoding of the input codes D, G, B, J, L, C, K, M, and the frequency domain error signal Δ_{2, 1}, Δ_3,1, Δ_1,2, Δ_2,2, Δ_3,2, Δ_1,3, Δ_2,3, Δ_3,3Are generated, and the adder 82_2,1, 82_3,1, 82_1,2, 82_2,2, 82_3,2, 82_1,3, 82_2,3, 82_3,3Given to. Accuracy converter 81_1,1The signal whose quantization accuracy is converted from 16 bits to 20 bits is converted into an error signal Δ by the adder._2,1And the addition result is given to the accuracy conversion unit and the orthogonal inverse conversion unit 39_2,1Is converted to a time domain signal and digital signal S_2,1Is played. Accuracy converter 81_2,1Converts the given frequency domain signal with a quantization accuracy of 20 bits into a quantization accuracy of 24 bits, and adds the adder 82._3,1Error signal Δ_3,1And the addition result is the orthogonal inverse transform unit 39._3,1Is converted to a time domain signal and digital signal S_3,1Is played.
[0226]
Orthogonal inverse transform unit 39_1,1, 39_2,1, 39_3,1Each input signal to the correction unit 36_1,2, 36_2,2, 36_3,2Through adder 82_1,2, 82_2,2, 82_3,2And the frequency domain error signal Δ_1,2, Δ_2,2, Δ_3,2Is added. These addition results are respectively converted into orthogonal inverse transform units 39._1,2, 39_2,2, 39_3,2Is converted to a time domain signal and the digital signal S_1,2, S_2,2, S_3,2Is played. Similarly, the orthogonal transform unit 39_1,2, 39_2,2, 39_3,2Each input signal to the correction unit 36_1,3, 36_2,3, 36_3,3Through adder 82_1,3, 82_2,3, 82_3,3And the frequency domain error signal Δ_1,3, Δ_2,3, Δ_3,3Is added. These addition results are respectively converted into orthogonal inverse transform units 39._1,3, 39_2,3, 39_3,3Is converted to a time domain signal and the digital signal S_1,3, S_2,3, S_3,3Is played.
[0227]
In the embodiment of FIG. 68, the correction unit 16_1,2, 16_2,2, 16_3,2, 16_1,3, 16_2,3 16_3,3Although correction by is performed in the frequency domain, it may be performed in the time domain. The correction in the time domain is also for example the signal S so that the power of the error signal is minimized._3,2Adjust the gain. For example, the correction unit 16_3,3As shown by a broken line in FIG._3,2Time domain digital signal S_3,2The correction part 16 '_3,3And correct the correction result using the orthogonal transform unit 19 ′._3,2The subtracting unit 63 as a frequency domain signal by orthogonal transformation at_3,3To give. The same applies to other correction units. In this case, in the decoding apparatus, as shown by the broken line in FIG._3,2Time-domain playback digital signal S obtained from_3,2The correction part 36 '_3,3And correct the correction result using the orthogonal transform unit 39 ′._3,2To convert to a frequency domain signal and adder 82_3,3In the frequency domain error signal Δ_3,3Can be added to. The same applies to other correction units. Alternatively, if the correction process is a reversible process, as shown in FIG._3,2The correction part 16 "_3,3Correct by the orthogonal transform unit 19_3,2Orthogonal transformation unit 19_3,2The output of the subtractor 63_3,3You may give to. In this case, in the decoding apparatus, as indicated by a broken line in FIG._3,2The output of the adder 82_3,3And the corresponding orthogonal inverse transform unit 39_3,236 "for output time domain signal_3,3It is only necessary to make corrections. In the latter modification, it is not necessary to increase the number of orthogonal transform units in both the encoding device and the decoding device.
Seventeenth embodiment
The plurality of original sound signals handled in the present invention described above may have different signal attributes such as sampling frequency, quantization accuracy, number of channels, and the like. Overall compression efficiency can be improved by hierarchical coding of sequences. Here, a method for designating various hierarchical structures of such a plurality of signals will be described.
[0228]
As described above, the sampling frequency, the quantization accuracy, and the number of channels can be hierarchized, and the encoding of the higher layer signal can include the encoding of the lower layer signal. As a result, the original sound signal can be reproduced at the designated sampling frequency, quantization accuracy, and number of channels, and encoding of a plurality of types of conditions can be integrated. Here, a description method that secures a degree of freedom is provided particularly for each input signal.
FIG. 72 shows an example of the structure of a compression code string in which hierarchical association is designated. In this embodiment, assuming that a sampling frequency (frequency direction) and quantization accuracy are hierarchized, and further a hierarchical structure of the number of channels is assumed, an error signal between hierarchies is a code string that is compression-encoded. Here, four compression code strings M, L, G, and A are shown. Each compression code string has a series of codes obtained by compressing and encoding the original sound signal of the same layer in the data area (field x9 described later). Therefore, the same hierarchy as the original sound signal is also defined for the code string. . Each code string is added with fields x1 to x7 describing its attributes (hierarchical information).
[0229]
A field x1 represents a sequence number of a code string, and here, sequence numbers 0, 1, 2, and 3 respectively given to a plurality of code strings M, L, G, and A are written. Field x2 represents the channel structure of the corresponding original sound signal, field x3 represents the sampling rate of the original sound signal, field x4 represents the quantization accuracy of the original sound signal, and field x5 represents the number of subsequences of the corresponding sound source signal. , Field x6 represents the sequence number of the lower code string, field x7 represents an extension flag indicating whether or not auxiliary information is present by “1” or “0”, and field x9 is data (obtained by compression coding). Represents a code string). Only when the extension flag of the field x7 is “1”, the field x8 representing the auxiliary information is inserted as indicated by the code string G. For example, when there are two code sequences L and G as subsequences for the code sequence as in the code sequence M, the number x5 of subsequences is 2, and the field number x6 has sequence numbers 2, 3 of the two subsequences. Is written. The lowest series A has no lower series.
[0230]
If the extension flag x7 is “1”, the encoding auxiliary information of the field x8 is added, and if the extension flag is “0”, the data string of the field x9 is started. The code string G shows an example in which the extension flag x7 is 1, and the auxiliary information field x8 is inserted. Each sequence is generally transmitted in correspondence with packets in units of frames, and packet management may be performed according to, for example, an existing Internet protocol. When only storing without transmitting, it is common to manage the head position of each code sequence separately from the code sequence.
FIG. 73 shows an original sound signal S with a quantization accuracy of 24 bits and sampling frequencies of 192 kHz and 96 kHz._1,1, S_1,2And the sampling frequency is 48kHz, quantization accuracy is 24bit and 16bit original sound signal S_2,1, S_2,2An example of hierarchical encoding is shown.
[0231]
Signal source 10_2,2Original sound signal S from_2,2Is the lower signal S_2,1Subtracter 13 subtracts the signal obtained by upsampling the sampling frequency from 96 kHz to 192 kHz with upsampling unit 13A1._2,2The error signal Δ_2,2The compression encoding unit 11_2,2To perform lossless compression encoding and output a code string M. Also signal source 10_2,1Original sound signal S from_2,1Is the lower signal S_1,2Subtracter 13 subtracts the signal obtained by upsampling the sampling frequency from 48 kHz to 96 kHz with upsampling unit 13A2._2,1The error signal Δ_2,1The compression encoding unit 11_2,1To obtain a code string L by lossless encoding. In addition, signal source 10_1,2Original sound signal S from_1,2Is the lower signal S_1,1The subtractor 13 subtracts the signal converted from the 16-bit to 24-bit quantization accuracy by the accuracy converter 13C1._1,2The error signal Δ_1,2The compression encoding unit 11_1,2To obtain a code string G by lossless encoding. Signal source 10_1,1The lowest signal S from_1,1Is the compression encoding unit 11_1,1To be a code string A.
[0232]
Therefore, the code string M is related to the lower code string L, the code string L is related to the lower code string G, and the code string G is related to the lower code string A.
FIG. 74 shows the association between a code string and code strings in which information fields x1 to x7 defining a hierarchical structure are added to the code strings M, L, G, and A generated by the encoding process of FIG. In the code sequences M, L, G, and A, sequence numbers 0, 1, 2, and 3 are written in the field x1, respectively. In the field x2, the channel configuration (number of channels) 2, 2, 2, 2 of the original sound signal corresponding to each code sequence is written. In the field x3, the corresponding original sound signal sampling rates 192, 96, and 48 (kHz) are written. In field x4, the quantization accuracy 24, 24, 24, 16 (bit) of the corresponding original sound signal is written. Since each of the sound source signals S22, S21, S12 uses one subordinate sound source signal and the sound source signal S22 does not take a difference, the number of subsequences of the code sequences M, L, and G is “ 1 "is written in the field x5, and each lower code sequence number is written in the field x6. 0 is written in the fields x5 and x6 of the code sequence A. Since the code sequences M, L, G, and A do not use auxiliary information, the field x7 is all set to “0”.
[0233]
Fig. 75 shows a structure that encodes nine types of layered original sound signals by combining three types of sampling frequencies 192kHz, 96kHz, 48kHz and three types of quantization accuracy 24bit, 20bit, 16bit. FIG. 76 shows a code sequence having such fields. In the encoding of FIG. 75, auxiliary information is not used, so that all the extension flags of field x7 are set to “0”. Also, the lowest signal S_1,1All other signals except S_3,3, S_2,3, S_1,3, S_3,2, S_2,2, S_1,2, S_3,1, S_2,1Since each of them is different from only one lower order signal, “1” is written in the field of the number of lower series.
[0234]
FIG. 77 is a description of the hierarchical structure for the code strings I, F, C, H, E, B, G, D, and A generated by encoding the hierarchical original sound signal shown in FIG. Similarly to the case of FIG. 75, nine types of hierarchized original sound signals are compression-coded. Since auxiliary information is used for this encoding, the extension flag x7 of all other code sequences except the code sequence A is set to "1", and the encoded information field x8 is inserted immediately after that. Has been.
FIG. 78 shows a description of the hierarchical structure corresponding to the code sequence by the multi-channel hierarchical encoding shown in FIG. In the embodiments so far, the encoding apparatus is premised on subtraction from the lower sequence, or the decoding apparatus is addition from the lower sequence. In FIG. 78, the code sequences of sequence numbers 7 and 8 designated in the field x6 of the code sequences of 5 and 6 represent the conversion of the difference signal and the sum signal into the code sequence in FIG. When interpreted as a device, the compressed code data of field x9 is not added to the fifth and sixth code sequences, and the auxiliary information of sequence number 5 is summed from the code numbers of sequence numbers 7 and 8 to the decoding side. The auxiliary information of the sequence number 6 specifies that the difference signal is generated from the code sequences of the sequence numbers 7 and 8. For this reason, code sequences No. 5 and No. 6 do not have subsequent compressed code data.
[0235]
In the case of the encoding that performs the inter-channel orthogonal transform described in FIG. 55, for example, information indicating that the orthogonal transform has been performed is written in the auxiliary information field x8 of the code sequence that has been subjected to the inter-channel orthogonal transform in FIG. If necessary, the grammar is determined in advance so that further details can be added.
FIG. 79 shows a basic processing procedure by the encoding apparatus of each embodiment described above. In the present invention, a plurality of original sound signals having hierarchized attributes are to be encoded. The attributes are, for example, the type of sampling frequency and the type of quantization accuracy of the samples in the first to sixteenth embodiments, and the five-channel signal and the stereo signal (two-channel signal) in the twelfth to fourteenth embodiments. ) In a signal system of a plurality of groups having different numbers of channels such as a monaural signal (one channel signal), the number of channels of the group to which the signal belongs is also an attribute of the signal. The direction in which the number of channels decreases is the lower direction. In the fifteenth embodiment, the attributes are a plurality of predetermined sampling frequencies and a plurality of predetermined amplitude resolutions of the signal. Based on the above definition, the encoding process is performed as follows.
[0236]
Step S1: The original sound signal to be encoded is searched for an original sound signal having a lower attribute.
Step S2: If there is a lower original sound signal, an error signal between the original sound signal to be encoded and the lower original sound signal or its modified signal is generated. That is, when there are two lower original sound signals, a modified signal synthesized from the lower two signals is obtained, and an error signal between the modified signal and the encoding target signal is obtained.
Step S3: The error signal is losslessly encoded.
Step S4: It is determined whether encoding has been completed for all the original sound signals. If not completed, the process returns to step S1.
[0237]
Step S5: When there is no original sound signal lower than the original sound signal to be encoded in Step S1, the original sound signal is losslessly encoded.
FIG. 80 shows a basic processing procedure performed by the decoding apparatus according to the above-described embodiment.
Step S1: A plurality of input codes are decoded to obtain an error signal and an original sound signal.
Step S2: A decoded original sound signal having a lower attribute than the error signal or a modified signal thereof and the error signal are synthesized to generate a decoded original sound signal.
Step S3: It is determined whether or not the processing has been completed for all the input codes.
[0238]
Such an encoding process procedure and a decoding process procedure can be described as a computer-executable program. The computer in which such a program is installed can execute signal encoding and decoding according to the present invention.
FIG. 81 shows the configuration of a computer that executes the encoding method and decoding method according to the present invention described in a program. The computer 100 includes a random access memory (RAM) 110, a central processing unit (CPU) 120, a hard disk (HD) 130, an input / output interface 140, and a transmission / reception unit 150 connected to a common data bus 160. Yes. The program describing the procedure of the encoding process and the decoding process described with reference to FIGS. 79 and 80 is installed in advance from a recording medium inserted in a media drive (for example, a CD drive) not shown in the hard disk 130. Alternatively, a program downloaded from the network NW may be installed.
[0239]
When executing the encoding process or the decoding process, the program is read from the hard disk 130 into the RAM 110, and the program is executed under the control of the CPU 120. For example, when performing an encoding process, the multichannel signal from the multichannel input device 220 connected to the input / output interface 140 is encoded and temporarily stored in the hard disk 130 or transmitted from the transmission / reception unit 150 to the network NW. Also good. When performing the decoding process, for example, the code of the multi-channel music program received from the network NW can be decoded and output from the input / output interface 140 to the multi-channel playback device 210 for playback.
[0240]
【The invention's effect】
As described above, according to the present invention, an error signal is generated from a signal having a layered attribute and a signal having a lower attribute than that signal or a modified signal thereof, and the error signal is losslessly encoded. Thus, efficient encoding can be realized and reversible encoding can be realized.
[Brief description of the drawings]
FIG. 1 is a diagram showing an example functional configuration of an encoding device and a decoding device according to a first embodiment of the present invention.
FIG. 2 is a diagram showing a functional configuration example of an encoding device and a decoding device according to a second embodiment of the present invention.
FIG. 3 is a diagram showing a functional configuration example of an encoding device and a decoding device according to a third embodiment of the present invention.
4 is a diagram illustrating an example of a functional configuration of an array conversion / encoding unit 18. FIG.
FIG. 5A is a diagram for explaining bit array conversion of a sample string expressed by polarity and absolute value, B is a diagram for explaining bit array conversion of a sample string expressed by two's complement, and C is It is a figure which shows the example of a format of a packet.
FIG. 6 is a diagram showing an example functional configuration of a decoding / array inverse transform unit 45 and a loss correction unit 58;
7 is a flowchart showing an example of a procedure of missing information correction processing in FIG. 6;
8 is a diagram showing a specific functional configuration example of a missing information correction unit 58B in FIG. 6;
FIG. 9 is a diagram illustrating a functional configuration example of an encoding device and a decoding device according to a third embodiment of the present invention.
10A is a diagram illustrating a specific functional configuration example of the prediction error generation unit 31 in FIG. 9; FIG. 10B is a diagram illustrating another configuration example of the prediction error generation unit 31;
11A is a diagram illustrating a specific functional configuration example of the prediction synthesis unit 56 in FIG. 9; FIG. 11B is a diagram illustrating another configuration example of the prediction synthesis unit 56;
12A is a diagram illustrating an example of a conceptual spectral characteristic of an error signal, and FIG. 12B is a diagram illustrating a spectral characteristic obtained by inverting the frequency axis of the spectral characteristic of A. FIG.
FIG. 13 is a diagram showing a functional configuration example of an encoding device and a decoding device according to a fourth embodiment of the present invention.
14A is a diagram showing an example of hierarchical code division according to the present invention, and FIG. 14B is a diagram showing an example of the relationship between amplitude resolution and amplitude word length.
15 is a diagram showing the relationship between the hierarchically divided code combinations shown in FIG. 14A and various sampling frequencies and various amplitude resolutions.
FIG. 16 is a diagram showing a functional configuration of a fifth embodiment of the encoding apparatus according to the present invention;
17A is a diagram showing interpolation in upsampling, and B is a diagram showing an example of an interpolation filter. FIG.
18A is a functional configuration diagram showing an example of a lossless compression encoder applicable to the embodiment of the present invention, and B corresponds to the lossless compression encoder of FIG. 18A applicable to the embodiment of the present invention. The functional block diagram of a decoder.
19A is a functional configuration diagram showing an example of a lossless encoder applicable to the embodiment of the present invention, and FIG. 19B is a diagram showing a functional configuration showing an example of a lossless decoder applicable to the embodiment of the present invention.
20A is a diagram illustrating a correspondence example between auxiliary codes and the number of taps, B is a diagram illustrating a correspondence example between auxiliary codes and gains, C is a diagram illustrating a correspondence example between auxiliary codes and sample point movement, and D is an auxiliary code. FIG.
FIG. 21 is a diagram showing a functional configuration of an embodiment of a decoding device according to the present invention;
FIG. 22 is a diagram showing a functional configuration of another embodiment of the encoding apparatus according to the present invention.
FIG. 23 is a diagram showing a functional configuration of another embodiment of the decoding apparatus according to the present invention;
FIG. 24 is a diagram showing a musical sound distribution system for explaining the effect of the present invention.
FIG. 25 is a diagram showing an example of hierarchical code division according to the seventh embodiment of the present invention;
FIG. 26 is a diagram illustrating a relationship between a hierarchically divided code combination, various sampling frequencies, and various amplitude resolutions.
FIG. 27 is a diagram showing a functional configuration of an embodiment of an encoding apparatus according to a seventh embodiment of the present invention;
FIG. 28 is a diagram showing a functional configuration of a predictive encoder applicable to the embodiment of the present invention.
FIG. 29 is a diagram showing a functional configuration of another embodiment of the encoding apparatus according to the seventh embodiment of the present invention;
FIG. 30 is a diagram showing a functional configuration of a decoding apparatus according to a seventh embodiment of the present invention.
FIG. 31 is a diagram showing a functional configuration of an encoding apparatus according to an eighth embodiment of the present invention.
FIG. 32 is a diagram showing a functional configuration of a decoding apparatus according to an eighth embodiment of the present invention.
FIG. 33 is a diagram showing an example of hierarchical code division according to the ninth embodiment of the present invention;
FIG. 34 is a diagram showing a relationship among a sampling frequency, an amplitude word length, and a combination of codes used in the ninth embodiment of the present invention.
FIG. 35 is a block diagram showing the functional arrangement of encoding apparatuses according to ninth and tenth embodiments of the present invention;
36 is a diagram showing a functional configuration example of a selection unit 76 in FIG. 35. FIG.
FIG. 37 is a diagram showing a functional configuration of the decoding apparatuses according to the ninth and tenth embodiments according to the present invention;
38 is a diagram showing another functional configuration example of the selection unit 76 in FIG. 35;
FIG. 39 is a diagram illustrating a functional configuration example of a selection unit 87 applied to the decoding device according to the ninth embodiment;
FIG. 40 is a diagram illustrating another example of the encoding apparatus according to the ninth and tenth embodiments.
FIG. 41 is a diagram showing still another example of the encoding apparatus according to the ninth and tenth embodiments.
FIG. 42 is a diagram showing an example of hierarchical code division according to the eleventh embodiment of the present invention;
43 is a diagram showing the relationship between the hierarchically divided code combinations shown in FIG. 42, various sampling frequencies, and various amplitude resolutions.
FIG. 44 is a diagram showing a functional configuration of an encoding apparatus according to an eleventh embodiment of the present invention.
FIG. 45 is a diagram showing a functional configuration of a decoding apparatus according to an eleventh embodiment of the present invention.
FIG. 46 is a diagram conceptually illustrating a first embodiment of the encoding method according to the present invention.
FIG. 47 is a block diagram showing a specific configuration example of an encoding apparatus according to a twelfth embodiment.
FIG. 48 is a block diagram illustrating a specific configuration example of a decoding device according to a twelfth embodiment.
FIG. 49 is a diagram for conceptually explaining a thirteenth embodiment of the encoding method according to the present invention;
FIG. 50 is a block diagram illustrating a specific configuration example of an encoding apparatus according to a thirteenth embodiment.
FIG. 51 is a block diagram showing a specific configuration example of a decoding apparatus according to a thirteenth embodiment.
FIG. 52 is a block diagram showing a configuration of a correction unit in the encoding apparatus according to the twelfth and thirteenth embodiments.
FIG. 53 is a block diagram showing a configuration of a correction unit in the decoding devices according to the twelfth and thirteenth embodiments.
FIG. 54 is a diagram for conceptually explaining a fourteenth embodiment of an encoding method according to the present invention.
FIG. 55 is a block diagram illustrating a specific configuration example of an encoding device according to a fourteenth embodiment.
FIG. 56 is a block diagram showing a specific configuration example of a decoding apparatus according to the fourteenth embodiment.
FIG. 57 is a block diagram showing a configuration of a fifteenth embodiment of an encoding apparatus according to the present invention.
FIG. 58 is a block diagram showing a configuration example of a difference module in the fifteenth embodiment.
FIG. 59 is a block diagram showing another configuration example of the difference module.
FIG. 60 is a block diagram showing a configuration of a decoding apparatus according to a fifteenth embodiment.
61 is a block diagram showing a configuration of an addition module in FIG. 60. FIG.
FIG. 62 is a block diagram showing another configuration example of the addition module.
63 is a block diagram showing another configuration example of the difference module in FIG. 57. FIG.
64 is a block diagram showing still another configuration example of the difference module in FIG. 57. FIG.
65 is a block diagram showing another configuration example of the addition module in FIG. 60. FIG.
66 is a block diagram showing still another configuration example of the addition module in FIG. 60. FIG.
FIG. 67 is a diagram for explaining a procedure for synthesizing signals with different sampling frequencies and quantization accuracy;
FIG. 68 is a block diagram showing a configuration of a coding apparatus according to a sixteenth embodiment.
69 is a block diagram showing a configuration of a decoding apparatus corresponding to the encoding apparatus in FIG. 68. FIG.
FIG. 70 is a block diagram showing a modified example of the encoding apparatus in FIG. 68;
71 is a block diagram showing a decoding apparatus corresponding to the encoding apparatus in FIG. 70. FIG.
FIG. 72 is a diagram showing an example of hierarchical information added to a code string.
Fig. 73 is a diagram showing a four-layer encoding configuration;
74 is a diagram showing hierarchical information added to a code sequence in the case of the coding configuration in FIG. 73. FIG.
Fig. 75 is a diagram showing an encoding configuration of nine layers.
76 is a diagram showing hierarchical information added to a code sequence in the case of the coding configuration in FIG. 75. FIG.
77 is a diagram showing hierarchical information added to a code sequence in the case of the encoding configuration in FIG. 57. FIG.
78 is a diagram showing hierarchical information added to the code sequence in the case of the encoding configuration in FIG. 50. FIG.
FIG. 79 is a flowchart showing the processing procedure of the encoding method according to the present invention.
FIG. 80 is a flowchart showing a processing procedure of a decoding method according to the present invention.
FIG. 81 is a block diagram showing the configuration of a computer for executing coding and decoding according to the present invention by a program.

Claims (60)

A digital signal encoding method that encodes a predetermined digital signal at least one of the stepwise hierarchy of sampling frequency and the stepwise hierarchy of quantization accuracy, which are attributes of the digital signal, with the higher frequency being the higher order If the lower one is lower, the higher quantization accuracy is higher, and the lower is lower,
comprising the steps of: (a) hierarchy of attributes than the signal to be encoded is generated by encoding the signal or variations signals of the lower,
(b) step of the hierarchy of the signal and the attributes to be encoded reversibly encodes the error signal between signals or variations signals of the lower,
And a digital signal encoding method. The digital signal encoding method according to claim 1, wherein
The step (a) converts a digital signal having a first sampling frequency into a digital signal having a second sampling frequency lower than the first sampling frequency for each frame;
The digital signal of the second sampling frequency is compression encoded and the main code is output,
Step (b) converts the local signal corresponding to the main code into a local signal of the first sampling frequency,
Calculating an error signal between the local signal of the first sampling frequency and the digital signal of the first sampling frequency as the error signal;
Generate a prediction error signal of the error signal,
The prediction error signal is characterized in that at least one bit position representing the amplitude of each sample is losslessly encoded with a coordinate bit string straddling the sample and output as an error code.   2. The digital signal encoding method according to claim 1, wherein the step (b) performs lossless encoding of a prediction error signal for an error signal obtained by inverting the frequency axis of the error signal. The digital signal encoding method according to claim 2,
The step (b) includes converting the error signal into an error signal having a sampling frequency lower than the first sampling frequency;
Generating a prediction signal for the converted error signal and converting the prediction signal to a prediction signal of a first sampling frequency;
Obtaining the prediction error signal from the converted prediction signal and the error signal of the first sampling frequency. The digital signal encoding method according to claim 2,
The step (b) performs a linear prediction analysis of the error signal, processes the error signal with the prediction coefficient, and generates a prediction signal;
Calculating a difference between the prediction signal and the error signal, generating the prediction error signal, encoding the prediction coefficient, and outputting a coefficient code. 2. The digital signal encoding method according to claim 1, wherein the step (a) includes the m-th and n-th digital signals having the m-th quantization accuracy and the n-th sampling frequency for the set of m = 1 and n = 1. Is a step of compressing and outputting the mth and nth codes,
Step (b) above is
For a set of m, n in the range of m = 1, 1 ≦ n ≦ N−1, the m, n digital signal is upsampled to an (n + 1) th sampling frequency higher than the nth sampling frequency, and m, Generate n + 1 upsampling signal,
The m, n + 1 error signal, which is an error signal between the m, n + 1 digital signal sampled with the mth quantization accuracy and the n + 1 sampling frequency and the m, n + 1 upsampling signal, Compressed and encoded as the error signal to output the m, n + 1 code,
For m and n pairs in the range of 1 ≦ m ≦ M-1 and 1 ≦ n ≦ N, the above m, n digital signals are converted to m + 1 quantization accuracy higher than the mth quantization accuracy. To generate the (m + 1) th and nth precision conversion signal,
The m + 1, n error signal, which is the error signal between the (m + 1, n) digital signal sampled at the (m + 1) th quantization accuracy and the nth sampling frequency, and the (m + 1, n) precision conversion signal is compressed. It encodes and outputs the m + 1, n code.   7. The digital signal encoding method according to claim 6, wherein the step (b) includes the power of the m, n + 1 error signal with the m, n + 1 upsampling signal adjusted based on the adjustment parameter. Encoding the m, n + 1 auxiliary information representing the adjustment parameter that minimizes and outputting the m, n + 1 auxiliary code.   7. The digital signal encoding method according to claim 6, wherein the step (b) is such that the power of the mth, n error signal with the m + 1, n accuracy conversion signal adjusted based on the adjustment parameter is minimum. And encoding the m + 1, n auxiliary information representing the adjustment parameter to output the m + 1, n auxiliary code. The digital signal encoding method according to claim 1, wherein the step (a) compresses and encodes the mth and nth error signals for a set of m = 1 and n = 1 to generate an mth and nth code.
Step (b) generates m-1 and n codes by compressing and coding the m-1 and n digital signals for m, n pairs in the range of 2 ≤ m ≤ M and 1 ≤ n ≤ N. And
For m and n in the range of 2 ≦ m ≦ M and 1 ≦ n ≦ N−1, the (m−1) th digital signal, the (m−1) th quantization accuracy, and the (n + 1) th sampling frequency higher than the nth sampling frequency Generating an m−1, n + 1 error signal that is an error from the m−1, n + 1 digital signal of
The m−1, n + 1 error signal is compression encoded to generate the m−1, n + 1 code. 2. The digital signal encoding method according to claim 1, wherein the step (a) compresses the mth and nth digital signals having the mth quantization accuracy and the nth sampling frequency for a set of m = 1 and n = 1. Encoding,
Step (b) above is for m, n pairs in the range 2 ≦ m ≦ M, 1 ≦ n ≦ N−1.
The mth, n + 1th digital signals having the mth quantization accuracy m and the (n + 1) th sampling frequency, and the respective errors between the mth, nth digital signals and the m-1, n + 1th digital signals. m, n error signal and m-1, n + 1 error signal are generated as the error signal,
An error signal having a small distortion between the mth and nth error signals and the m−1 and n + 1 error signals is selected and losslessly encoded to generate the mth and n + 1 codes. The m, n + 1 auxiliary code indicating whether or not it has been generated is generated. 2. The digital signal encoding method according to claim 1, wherein the step (a) compresses the mth and nth digital signals having the mth quantization accuracy and the nth sampling frequency for a set of m = 1 and n = 1. Encoding,
Step (b) above is for m, n pairs in the range 2 ≦ m ≦ M, 1 ≦ n ≦ N−1.
The m, n + 1 digital signal is generated by weighted addition of the m, n digital signal and the m-1, n + 1 digital signal, and the difference from the m, n + 1 digital signal is calculated as described above. As an error signal,
The error signal is lossless compression encoded to generate an m, n + 1 code. 2. The digital signal encoding method according to claim 1, wherein the step (a) includes the m-th and n-th digital signals having the m-th quantization accuracy and the n-th sampling frequency for a set of m = 1 and n = 1. Compressed and output the mth and nth codes,
In the step (b), for the m, n pairs in the range of 1 ≦ m ≦ M and 1 ≦ n ≦ N-1, the mth and nth digital signals are sampled at the (n + 1) th sampling higher than the nth sampling frequency. Upsampling to frequency to generate m, n + 1th upsampling signal,
The m, n + 1 error signal, which is an error signal between the m, n + 1 digital signal of the mth quantization accuracy and the (n + 1) th sampling frequency and the m, n + 1 upsampling signal, is the error signal. And outputs the mth, n + 1th code after compression coding as
For m, n pairs in the range of m = 1, 1 ≦ n ≦ N−1, the mth and nth digital signals are converted to the (m + 1) th quantization accuracy higher than the mth quantization accuracy. Generate m + 1, n precision conversion signal,
The m + 1, n error signal, which is an error signal, is used as the error signal between the m + 1, n digital signal of the (m + 1) th quantization accuracy and the nth sampling frequency and the (m + 1, n) th accuracy conversion signal. The m + 1, n code is output after compression encoding. 13. The digital signal encoding method according to claim 12, wherein the step (b) includes calculating the m, n + 1 error signal with the m, n + 1 upsampling signal adjusted based on an adjustment parameter. Encoding an adjustment parameter that minimizes power and outputting an m, n + 1 auxiliary code, or
The first m + 1, n auxiliary code the first m + 1, n error signal that is adjusted based on the adjustment parameter the first m + 1, n precision conversion signal encodes the minimum Do that adjustment parameters Outputting. A digital signal encoding apparatus for encoding a digital signal, at least one of predetermined stepwise hierarchy and the quantization precision stage hierarchy of the sampling frequency is an attribute of the digital signal, the higher the higher of the frequency If the lower one is lower, the higher quantization accuracy is higher, and the lower is lower,
Hierarchical attributes than the signal to be encoded to generate a signal or a variant signal lower, main code generating means for encoding,
An error signal encoding means for losslessly encoding an error signal between the signal to be encoded and a signal having a lower level of the attribute or a modified signal thereof;
And a digital signal encoding device. The digital signal encoding device according to claim 14, wherein
The main code generating means converts a digital signal having a first sampling frequency into a digital signal having a second sampling frequency lower than the first sampling frequency for each frame;
An encoding unit that compresses and encodes the digital signal having the second sampling frequency and outputs a main code;
The error signal encoding means includes an upsampling unit that converts a local signal corresponding to the main code into a local signal having a first sampling frequency;
An error calculator that calculates an error signal between the local signal at the first sampling frequency and the digital signal at the first sampling frequency as the error signal;
A prediction error generation unit that generates a prediction error signal of the error signal,
And an array conversion unit for losslessly encoding a coordinate bit string straddling the sample for each bit position in the bit position representing the amplitude of each sample of the prediction error signal and outputting the result as an error code. . 15. The digital signal encoding device according to claim 14, wherein the main code generation means compresses and encodes the m-th and n-th digital signals for m, n sets of m = 1, n = 1. It consists of mth and nth encoding units that output n code,
The error signal encoding unit, m = 1, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the upper Symbol first m, higher than the n sampling frequency n digital signals (n + 1) -th sampling the m, m-th to generate the n + 1 up-sampling signal by up-sampling frequency, and (n + 1) up-sampling unit,
For m, n pairs in the range of m = 1, 1 ≦ n ≦ N−1, the m, n + 1 upsampling signal and the m, n + 1 digital signal are error signals m, an m, n + 1 encoding unit that compresses and encodes an n + 1 error signal as the error signal and outputs an m, n + 1 code;
For m, n pairs in the range of 1 ≦ m ≦ M-1, 1 ≦ n ≦ N, the mth and nth digital signals are converted to m + 1th quantization accuracy higher than the mth quantization accuracy. The m + 1, n accuracy conversion unit for generating the m + 1, n accuracy conversion signal. 15. The digital signal encoding apparatus according to claim 14, wherein the main code generating means outputs the mth and nth digital signals having the mth quantization accuracy and the nth sampling frequency to the (m-1) th lower than the mth quantization accuracy. Division to divide into m−1, n digital signal of quantization accuracy, nth sampling frequency and m−1, n error signal which is an error between m−1, n digital signal and m−nth digital signal And
For the set of m = 1, n = 1, the m-th, n-th compression unit that generates the m-th, n-th code by lossless compression coding the m-th, n-th error signal,
The above m-1, n digital signals or the input m-1, n digital signals are compressed and encoded for m, n pairs in the range of 2 ≦ m ≦ M and 1 ≦ n ≦ N−1. The m-1, n compression unit for generating m-1, n code,
The error signal encoding means includes the (m−1) th digital signal used to generate the (m−1, n) code, the (n + 1) th sampling higher than the (m−1) th quantization accuracy and the nth sampling frequency. An m-1, n + 1 error generation unit that generates an m-1, n + 1 error signal that is an error from the m-1, n + 1 digital signal of frequency;
An m−1, n + 1 compression unit for generating a m−1, n + 1 code by lossless compression encoding the m −1 , n + 1 error signal;
Consists of. 15. The digital signal encoding apparatus according to claim 14, wherein the main code generation means compresses the mth and nth digital signals having the mth quantization accuracy and the nth sampling frequency for a set of m = 1 and n = 1. The mth and nth encoding means for encoding,
The error signal encoding means has an m-1 quantization accuracy lower than the mth quantization accuracy and an nth sampling frequency for a set of m and n in the range of 1 ≦ m ≦ M and 1 ≦ n ≦ N−1. M-1, n + 1 encoding means for compressing and encoding an m-1, n + 1 digital signal having a high n + 1 sampling frequency;
The mth, n + 1th digital signal having the mth quantization accuracy and the (n + 1) th sampling frequency, and the mth error which is an error between the mth, nth digital signal and the m-1, n + 1 digital signal. , n error signal and error signal generating means for generating the (m-1, n + 1) th error signal,
An m, n + 1th code is generated by selecting an error signal having a small distortion between the mth, nth error signal and the (m-1, n + 1) th error signal and performing lossless compression coding. A compression section;
An m, n + 1 auxiliary encoding unit for generating an m, n + 1 auxiliary code indicating which of the above is selected;
Consists of. 15. The digital signal encoding apparatus according to claim 14, wherein the main code generation means compresses the mth and nth digital signals having the mth quantization accuracy and the nth sampling frequency for a set of m = 1 and n = 1. The mth and nth encoding means for encoding,
The error signal encoding means includes the m, n digital signal and the m-1, n + 1 digital signal for a set of m, n in the range of 2 ≦ m ≦ M and 1 ≦ n ≦ N−1. To generate an m, n + 1 added signal by weighting, and an m, n + 1 mixing unit for generating a difference from the m, n + 1 digital signal as an error signal;
The m, n + 1 compression unit that generates the m, n + 1 code by lossless compression encoding the error signal,
Consists of. 15. The digital signal encoding device according to claim 14, wherein the main code generating means is an m-th, n-th digital signal sampled at an m-th quantization accuracy and an n-th sampling frequency for a set of m = 1, n = 1. Comprising an mth and nth compression unit that compresses and encodes and outputs an mth and nth code,
The error signal encoding means converts the mth and nth digital signals to the (n + 1) th higher than the nth sampling frequency for a set of m and n in the range of 1 ≦ m ≦ M and 1 ≦ n ≦ N−1. An m, n + 1 upsampling unit that upsamples to a sampling frequency to generate an m, n + 1 upsampling signal;
The m, n + 1 error signal, which is an error signal between the m, n + 1 digital signal of the mth quantization accuracy and the (n + 1) th sampling frequency and the m, n + 1 upsampling signal, is the error signal. The m, n + 1 compression unit that compresses and outputs the m, n + 1 code as
For m, n pairs in the range of 1 ≦ m ≦ M-1, 1 ≦ n ≦ N, convert the mth and nth digital signals to m + 1th quantization accuracy higher than the mth quantization accuracy. M + 1, n precision conversion unit for generating m + 1, n precision conversion signal,
The m + 1, n error signal is compression-coded as the error signal between the m + 1, n digital signal having the (m + 1) th quantization accuracy and the nth sampling frequency and the m + 1, n accuracy conversion signal. M + 1, n compression unit for outputting the m + 1, n code,
Consists of. A decoding method for decoding a code of digital signals, at least one of predetermined stepwise hierarchy and the quantization precision stage hierarchy of the sampling frequency is an attribute of the digital signal, the higher the higher of the frequency If the lower one is lower, the higher quantization accuracy is higher, and the lower is lower,
(a) decoding an input code to generate an error signal;
(b) the error signals and the step of the error signal from the attribute hierarchy as decoded signal generated by decoding the main code to generate decrypt signals or synthesized and decoded signal and the modified signal of the lower,
The decoding method characterized by including these. The decoding method according to claim 21,
In the step (a), an input error code is decoded as the input code, and a prediction error signal having a first sampling frequency composed of a bit string at the same bit position straddling a sample at every bit position among the bit positions is reproduced. ,
The step (b) combines the prediction error signal to reproduce the error signal, converts the decoded signal obtained by decoding the main code into a decoded signal having the first sampling frequency higher than the sampling frequency, and converts the converted signal. A reproduced digital signal is obtained by adding the decoded signal and the error signal.   The decoding method according to claim 21, wherein the step (b) includes adding the error signal to the converted decoded signal after inverting the frequency axis. The decoding method according to claim 22, wherein the step (b) comprises:
Converting the prediction error signal into a prediction error signal having a second sampling frequency lower than the first sampling frequency;
Converting a prediction signal for the prediction error signal of the second sampling frequency into a prediction signal of the first sampling frequency;
The error signal is generated by adding the prediction signal of the first sampling frequency and the prediction error signal of the first sampling frequency. The decoding method according to claim 22,
The step (b) generates a prediction signal by linearly predicting the prediction error signal using a linear prediction coefficient obtained by decoding the input coefficient code,
The prediction signal and the prediction error signal are added to obtain the error signal. The digital signal decoding method according to claim 21, wherein said step (a) comprises:
m = 1, 1 ≦ n ≦ N-1 in the range of m, for a set of n m-th quantization precision as signal hierarchy subordinate of the attributes, the m n-th sampling frequency, the n digital signals a up-sampling to the (n + 1) th sampling frequency higher than the n-sampling frequency to generate the (m, n + 1) -th upsampling signal,
For a set of m, n in the range of 1 ≦ m ≦ M, 1 ≦ n ≦ N−1, the mth, n + 1 code is decoded as the input code, the mth quantization accuracy, and the n + 1 sampling frequency The m, n + 1 error signal is generated, and the m, n + 1 error signal and the m, n + 1 upsampling signal are added to generate the m, n + 1 reproduction signal. Procedure and
1 ≦ m ≦ M-1, 1 ≦ n ≦ N range m of, for a set of n, the higher than the a signal hierarchy subordinate of the attributes the m, the n digital signal the m quantization precision m + 1, n accuracy conversion signal is generated by converting the accuracy to m + 1 quantization accuracy, and the m + 1, n code is decoded as the input code to obtain the (m + 1) th quantization accuracy, the nth sampling. A second m + 1, n error signal is generated, and the m + 1, n error signal and the m + 1, n precision conversion signal are added to generate an m + 1, n digital signal. procedure,
And at least one of the steps
The step (b) is characterized in that the mth and nth digital signals are generated by decoding the mth and nth codes for the set of m = 1 and n = 1. The decoding method according to claim 26, wherein said step (a) comprises:
For m, n pairs in the range of 1 ≦ m ≦ M and 1 ≦ n ≦ N−1, the m, n + 1 auxiliary information is decoded to generate adjustment parameters for the m, n + 1 upsampling signal. And steps to
The m, n + 1 reproduction signal is generated by adding the m, n + 1 error signal and the m, n + 1 upsampling signal adjusted based on the adjustment parameter. The decoding method according to claim 26, wherein said step (a) comprises:
For m, n pairs in the range of 1 ≦ m ≦ M-1, 1 ≦ n ≦ N, the m + 1, n auxiliary code is decoded to generate adjustment parameters for the m + 1, n precision conversion signal. ,
The m + 1, n digital signal is generated by adding the m + 1, n error signal and the m + 1, n accuracy conversion signal adjusted based on the adjustment parameter. The decoding method according to claim 21, wherein said step (a) comprises:
1 ≦ m ≦ M, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the m quantizing reversibly expansion decoding hierarchy first m as signal backward, the n + 1 code of the attribute Generate the m, n + 1 error signal of accuracy, n + 1 sampling frequency,
2 ≦ m ≦ M, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the m hierarchy of the attributes as a signal of a lower, n digital signal and the m-1, n + 1 digital signal The m, n + 1 digital signal is reproduced by adding one of the signals indicated by the selection information obtained by decoding the m, n + 1 auxiliary code and the m, n + 1 error signal,
The step (b) is characterized in that the mth and nth digital signals are generated by decoding the mth and nth codes for the set of m = 1 and n = 1. The decoding method according to claim 21, wherein said step (a) comprises:
For m, n pairs in the range of 1 ≦ m ≦ M, 1 ≦ n ≦ N−1 excluding the set of m = 1 and n = 1, the mth and n + 1 codes are losslessly decompressed and the mth quantum Generating the m, n + 1 error signal of the (n + 1) th sampling frequency,
2 ≦ m ≦ M, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the m, n digital signal and the m-1, n + 1 digital signal hierarchy of the attributes as a signal of lower Are weighted and added with the information obtained by decoding the m, n + 1 auxiliary code to generate the m, n + 1 added signal having the mth quantization accuracy and the (n + 1) th sampling frequency.
The m, n + 1 added signal and the m, n + 1 error signal are added to reproduce the m, n + 1 digital signal,
The step (b) is characterized in that the mth and nth digital signals are generated by decoding the mth and nth codes for the set of m = 1 and n = 1. 23. The decoding method according to claim 21, wherein the step (a) includes a combination of m and n in the range of 1 ≦ m ≦ M and 1 ≦ n ≦ N−1, and the attribute hierarchy is lower . the m, m-th to n digital signals by up-sampling to the n + 1 sampling frequency higher than the n sampling frequency to generate n + 1 up-sampling signal as a signal,
Decoding the m, n + 1 code as the input code to generate the m, n + 1 error signal of the mth quantization accuracy and the n + 1 sampling frequency;
In the step (a), the m + n digital signals are converted to m + th higher than the mth quantization accuracy for m, n pairs in the range of 1 ≦ m ≦ M−1, 1 ≦ n ≦ N. 1) Convert the precision to quantization precision to generate m + 1, n precision conversion signal,
1 ≦ m ≦ M, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the first m, n + 1 error signal and the second m hierarchy of the attributes as a deformation signal of the lower, n + A first procedure of adding upsampling signals to generate an m, n + 1 digital signal;
1 ≦ m ≦ M-1, 1 ≦ n ≦ N range m of, for a set of n, the (m + 1) th, n error signal and the (m + 1) -th hierarchy of the attributes as a deformation signal of lower, a second procedure for adding an n-precision conversion signal to generate an m + 1, n digital signal;
And generate the decoded signal by any of the procedures
The step (b) is characterized in that the mth and nth digital signals are generated by decoding the mth and nth codes for the set of m = 1 and n = 1.   The decoding method according to claim 31, wherein the first procedure adjusts the m, n + 1 upsampling signal to be added based on an adjustment parameter obtained by decoding the m, n + 1 auxiliary code, The second procedure is characterized in that the m + 1, n auxiliary code is decoded based on the generated adjustment parameter to adjust the m + 1, n accuracy conversion signal. At least one of the stepped hierarchy of sampling frequency and the stepped hierarchy of quantization accuracy, which are attributes of the digital signal, is a predetermined digital signal decoding device, with the higher frequency being higher and the lower being lower If the higher quantization accuracy is higher and the lower quantization is lower,
Error signal generation means for decoding an input code to generate an error signal;
The error signal and a signal synthesizing means for hierarchical attributes from the error signal to produce a lower decrypted signal or synthesized and decoded signal and the modified signal,
And a digital signal decoding apparatus. 34. The decoding apparatus according to claim 33, wherein the error signal generating means is
An array conversion unit that decodes an input error code to obtain a bit string, extracts a bit at the same bit position in each bit array direction from one frame of the obtained bit string, and reproduces a prediction error signal of the first sampling frequency When,
A prediction synthesis unit that predictively synthesizes the prediction error signal and reproduces the error signal;
The signal synthesizing means is
A decoding unit that decodes an input main code to obtain a decoded signal;
An upsampling unit for converting the decoded signal into a decoded signal having the first sampling frequency higher than the sampling frequency;
And an adder that adds the converted decoded signal and the error signal to obtain a reproduced digital signal. In the decoding apparatus according to claim 33, said signal combining means, m = 1, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the m as signal hierarchy subordinate of the attribute , an upsampling unit that upsamples the n digital signal to an n + 1th sampling frequency higher than the nth sampling frequency to generate an m, n + 1 upsampling signal;
For m, n pairs in the range of 1 ≦ m ≦ M, 1 ≦ n ≦ N−1, the m, n + 1 code is decoded to obtain the mth quantization accuracy, the m + 1th sampling frequency, an m, n + 1 decoding unit for generating an n + 1 error signal;
An m, n + 1 reproducing means comprising an adding unit for adding the m, n + 1 error signal and the m, n + 1 upsampling signal to generate an m, n + 1 reproduced signal;
1 ≦ m ≦ M-1, 1 ≦ n ≦ N range m of, for a set of n, the higher than the a signal hierarchy subordinate of the attributes the m, the n digital signal the m quantization precision an accuracy conversion unit for converting the accuracy to m + 1 quantization accuracy and generating an m + 1, n-th accuracy conversion signal;
An m + 1, n decoding unit that decodes the m + 1, n code to generate an m + 1, n error signal having an (m + 1) th quantization accuracy and an nth sampling frequency;
Any of the (m + 1, n) reproducing means including an adding unit for adding the (m + 1, n) error signal and the (m + 1, n) accuracy conversion signal to generate the (m + 1, n) digital signal. With some replay means,
The signal synthesizing unit includes an m, n decoding unit that decodes the m, n code and generates the m, n digital signal for a set of m = 1, n = 1. 34. The decoding apparatus according to claim 33, wherein the error signal generating means is
For m, n pairs in the range of 2 ≦ m ≦ M, 1 ≦ n ≦ N−1 excluding the pair of m = 1, n = 1, the input multiple codes are decoded to obtain the mth quantization accuracy, The mth and n + 1th digital signals of the nth sampling frequency, the (m-1) th quantization accuracy lower than the mth quantization accuracy, and the (m + 1) th and n + 1th digital signals of the (n + 1) th sampling frequency higher than the nth sampling frequency. Playback means for playing
An m, n + 1 decompression unit that generates a mth, n + 1 error signal having an mth quantization accuracy and an (n + 1) th sampling frequency by performing lossless decompression decoding of the m, n + 1 code;
2 ≦ m ≦ M, 1 ≦ n ≦ N-1 in the range of m, for a set of n, as signal hierarchy subordinate of the attributes, the m, n digital signal and the m-1, n + 1 digital The mth, n + 1th digital signal is reproduced by adding one of the signals instructed by the selection information obtained by decoding the mth, n + 1th auxiliary code and the mth, n + 1th error signal. m, n + 1 adder,
And consist of
The signal synthesizing unit includes an m, n decoding unit that decodes the m, n code and generates the m, n digital signal for a set of m = 1, n = 1. 34. The decoding apparatus according to claim 33, wherein the error signal generating means is
For m, n pairs in the range of 1 ≦ m ≦ M, 1 ≦ n ≦ N−1 excluding the set of m = 1, n = 1, the mth and n + 1 codes are losslessly decoded and the mth quantum M, n + 1 decompression unit for generating m, n + 1 error signal of (n + 1) th sampling frequency,
The m, n + 1 auxiliary decoding unit for decoding the m, n + 1 auxiliary code and obtaining auxiliary information specifying the addition method;
2 ≦ m ≦ M, 1 ≦ n ≦ N-1 in the range of m, for a set of n, the m as signal hierarchy subordinate of the attributes based on the auxiliary information, n digital signal and the m- 1, n + 1 m-th digital signal by adding the weighted the hierarchy of the attributes to generate a first m, n + 1 added signal as a deformation signal of the lower, and the (n + 1) mixing section,
The m, n + 1 added signal and the m, n + 1 error signal are added to reproduce the m, n + 1 digital signal having the mth quantization accuracy and the (n + 1) th sampling frequency. , n + 1 adder,
It consists of. 34. The decoding device according to claim 33, wherein the error signal generation and synthesis means is a signal having a lower hierarchical level of the attribute for a set of m and n in a range of 1 ≦ m ≦ M and 1 ≦ n ≦ N−1. the m, n digital signal hierarchies higher (n + 1) -th sampling frequency up-sampled and the attributes than the n sampling frequency to generate a first m, n + 1 up-sampling signal as a deformation signal of lower as No. The m, n + 1 upsampling unit;
An m, n + 1 decompression unit that decodes the m, n + 1 code as the input code to generate an m, n + 1 error signal having an m + 1 quantization frequency and an n + 1 sampling frequency;
1 ≦ m ≦ M-1, 1 ≦ n ≦ N range m of, for a set of n, the first m as the m-th, n + 1 error signal and the deformation signal hierarchy subordinate the attribute, n An m, n + 1 reproducing means comprising an adding unit for adding an +1 upsampling signal to generate an m, n + 1 digital signal;
For the set of m, n in the range of 1 ≦ m ≦ M−1 and n = 1, the mth and nth digital signals are converted to m + 1th quantization accuracy higher than the mth quantization accuracy and an m + 1, n precision conversion unit for generating an m + 1, n precision conversion signal;
An m + 1, n decompression unit that decodes the m + 1, n code to generate an m + 1, n error signal with an (m + 1) th quantization accuracy and an Nth sampling frequency;
An m + 1, n reproducing means comprising an adding unit for adding the m + 1, n error signal and the m + 1, n accuracy conversion signal to generate an m + 1, n digital signal;
At least one of the m, n + 1 playback means, the m + 1, n playback means, the m + 1, n playback means, and the m + 1, n + 1 playback means. Become
The signal synthesizing means includes an m-th and n-th decompression unit that decodes the m-th and n-th codes and generates the m-th and n-th digital signals for the set of m = 1 and n = 1. 2. The digital signal encoding method according to claim 1, wherein the signal to be encoded is a digital signal of one channel in a first group consisting of a plurality of channels.
Hierarchy signals or variations signal that the lower of the attribute is a linear combination of the first consisting of smaller number of channels than the group the second 1-channel digital signal or a digital signal of multiple channels of the group It is characterized by that. 40. The digital signal encoding method according to claim 39, wherein the second group of digital signals includes a monaural signal having a first quantization accuracy and a first sampling frequency, and a second higher-order attribute layer than the monaural signal . A plurality of channel signals having quantization accuracy and a second sampling frequency, wherein the first group of digital signals have the second quantization accuracy and the second sampling frequency, and the second group of channel signals. Contains channel signals equal to or greater than the number,
Step (a) includes encoding the monaural signal;
Step (b) above is
(b-1) generating a transformed signal obtained by upgrading the attribute level to the second quantization accuracy and the second sampling frequency for the monaural signal;
(b-2) generating and encoding a difference between the converted signal and the second group of channel signals as a second group error signal;
(b-3) generating and encoding an error signal between each of the second group of channel signals and the first group of channel signals;
It consists of. 41. The encoding method according to claim 40, wherein the second group includes a left channel signal and a right channel signal, and the step (b-2) uses a difference signal between the left channel signal and the right channel signal as the second channel signal. Generating and encoding as one of the error signals of the group;
Generating a sum signal of the left channel signal and the right channel signal, and generating and encoding a difference signal between the converted signal and the sum signal as another one of the error signals. 15. The encoding device according to claim 14, wherein the signal to be encoded is a digital signal of one channel among a first group consisting of a plurality of channels.
Hierarchy signals or variations signal that the lower of the attribute is a linear combination of the first consisting of smaller number of channels than the group the second 1-channel digital signal or more of a plurality of channels of digital signals of the group It is characterized by that. In the coding apparatus of claim 42, the digital signal of the second group, a mono signal having a first quantization accuracy and a first sampling frequency, second quantum of hierarchical attributes from the monophonic signal is higher A plurality of channel signals having a quantization accuracy and a second sampling frequency, wherein the first group of digital signals has the second quantization accuracy and a second sampling frequency, and the number of the second group of channel signals. A first group of channel signals equal to or greater than
The main code generating means is means for compressing and encoding the monaural signal,
The error signal encoding means includes:
Upgrading means for generating a converted signal obtained by upgrading the monaural signal to a signal having the second quantization accuracy and the second sampling frequency;
A plurality of second group subtracting sections for obtaining an error between the converted signal and the second group of channel signals and outputting a plurality of first error signals;
A compression encoder for losslessly encoding the second group of error signals,
A plurality of first group subtracting units for generating a plurality of first group error signals between the second group of channel signals and the first group of channel signals;
A plurality of first group compression encoding units for lossless encoding the plurality of first group error signals;
Including. 44. The encoding device according to claim 43, wherein the second group of channel signals comprises a left channel signal and a right channel signal, and the first group of channel signals comprises two or more multi-channel signals,
The second group subtracting unit that generates the error signal of the second group includes:
A subtractor for generating a difference signal between the left and right channel signals as one of the error signals of the second group;
An adder for generating a sum signal between the left and right channel signals, and a subtractor for generating a difference between the sum signal and the converted signal as an error signal of the second group;
It consists of. The decoding method according to claim 21,
The error signal is a digital error signal of one channel in the first group consisting of a plurality of channels.
Hierarchy decrypt signals or decoded signals thereof the lower of the attributes, the linear of the first 1 channel decoded digital signal of the second group of fewer number of channels group or 2 or more of a plurality of channels of digital signals It is a combination. 46. The decoding method according to claim 45, wherein the digital signals of the second group include a monaural signal having a first quantization accuracy and a first sampling frequency, and a second higher-order attribute layer than the monaural signal . A plurality of channel signals having a quantization accuracy and a second sampling frequency, wherein the first group of digital error signals has the second quantization accuracy and a second sampling frequency, and the second group of channel signals. Contains channel signals equal to or greater than
The step (a) includes the steps of decoding the error code of the second group of channel signals and the error code of the first group of channel signals, respectively, and generating a second group error signal and a first group error signal. ,
Step (b) above is
(b-1) decoding the main code to reproduce the monaural signal;
(b-2) generating a transformed signal obtained by upgrading the attribute level to the second quantization accuracy and the second sampling frequency of the monaural signal;
(b-3) adding the converted signal and the first error signal to reproduce a second group of channel signals;
(b-4) adding the reproduced second group channel signal and the first group error signal to reproduce the first group channel signal;
It consists of.   47. The decoding method according to claim 46, wherein said second group of channel signals are a left channel signal and a right channel signal, and said step (b-3) decodes said second group of error codes to generate a second signal. Generating a sum signal and a difference signal of the left and right channel signals of the group; and adding and subtracting the sum signal and the difference signal from each other to reproduce a left channel signal and a right channel signal. The decoding device according to claim 33,
The error signal is a digital error signal of one channel in the first group consisting of a plurality of channels.
Hierarchy decrypt signals or decoded signals thereof the lower of the attributes, the linear of the first 1 channel decoded digital signal of the second group of fewer number of channels group or 2 or more of a plurality of channels of digital signals It is a combination. 49. The decoding device according to claim 48, wherein the second group of digital signals includes a monaural signal having a first quantization accuracy and a first sampling frequency, and a second higher-order attribute layer than the monaural signal . A plurality of channel signals having a quantization accuracy and a second sampling frequency, wherein the first group of error signals has the second quantization accuracy and a second sampling frequency, and the second group of channel signals Consists of channel signals equal to or greater than the number,
The error signal generating means includes: a second group decoding unit that decodes a second group error code to obtain a second group error signal; and a first group error signal that decodes the first group error code to obtain a first group error signal. It consists of a group decryption unit,
The signal synthesizing means includes a monaural signal decoding unit that decodes the main code to reproduce the monaural signal, and grades the monaural signal into a signal having the same second quantization accuracy and second sampling frequency as the second group of channel signals. An upgraded part for generating an up-converted signal, a second group adder for reproducing the second group of channel signals by adding the converted signal and the error signal of the second group, and the reproduced second A first group adder for adding the group channel signal and the first group error signal to reproduce the first group channel signal;   50. The decoding apparatus according to claim 49, wherein the first group of channel signals is a left channel signal and a right channel signal, and one of the second group of decoded error signals is between the left and right channel signals. A first adder which is a difference signal, and wherein the addition unit of the second group generates the sum signal of the left and right channel signals by adding the conversion signal and another decoded error signal of the second group And a second adder and a subtractor for adding and subtracting the difference signal and the sum signal to reproduce the left channel signal and the right channel signal, respectively. 2. The digital signal encoding method according to claim 1, wherein the signal to be encoded is a digital signal of one channel in a first group consisting of a plurality of channels.
Hierarchy signals or variations signal that the lower of the attribute is a linear combination of the first consisting of smaller number of channels than the group the second 1-channel digital signal or more of a plurality of channels of digital signals of the group It is characterized by that. 52. The encoding method according to claim 51, wherein the digital signals of the second group include a monaural signal having a first quantization accuracy and a first sampling frequency, and a second higher-order attribute layer than the monaural signal . A plurality of channel signals having a quantization accuracy and a second sampling frequency, wherein the first group of digital signals has the second quantization accuracy and a second sampling frequency, and the second group of channel signals Contains channel signals equal to or greater than the number,
The step (a) includes a step of compressing and encoding a monaural signal having the first quantization accuracy and the first sampling frequency.
Step (b) above is
Generating a converted signal obtained by upgrading the attribute hierarchy of the monaural signal to a signal having the second quantization accuracy and the second sampling frequency;
Generating and encoding a difference between the converted signal and the component of the second group of channel signals as a second group of error signals;
Generating a frequency domain signal by performing inter-channel orthogonal transform on the first group of channel signals;
Generating a difference between at least one of the frequency domain signals and the transformed signal component as the error signal of the first group;
Compressing and encoding each of the first group of error signals and frequency domain signals;
It consists of. The decoding method according to claim 21,
The error signal is a digital error signal of one channel in the first group consisting of a plurality of channels.
Hierarchy decrypt signals or decoded signals thereof the lower of the attributes, the linear of the first 1 channel decoded digital signal of the second group of fewer number of channels group or 2 or more of a plurality of channels of digital signals It is a combination. 54. The decoding method according to claim 53, wherein the digital signals of the second group include a monaural signal having a first quantization accuracy and a first sampling frequency, and a second higher-order attribute layer than the monaural signal . A plurality of channel signals having a quantization accuracy and a second sampling frequency, wherein the first group of digital error signals has the second quantization accuracy and a second sampling frequency, and the second group of channel signals. A first group of channel signals equal to or greater than
Step (b) above decodes the main code to reproduce the monaural signal,
The step (a) generates a transformed signal obtained by upgrading the monaural signal to the second quantization accuracy and the second sampling frequency, decodes the second group error code, and generates a second group error signal. , Adding the error signal of the first group and the converted signal to reproduce the channel signal of the second group,
The addition result and the remaining frequency domain signal are subjected to direct inverse transformation and reproduced as a channel signal of the second group of time domain signals. 15. The digital signal encoding apparatus according to claim 14, wherein the signal to be encoded is a digital signal of one channel in a first group consisting of a plurality of channels.
Hierarchy signals or variations signal that the lower of the attribute is a linear combination of the first consisting of smaller number of channels than the group the second 1-channel digital signal or more of a plurality of channels of digital signals of the group It is characterized by that. 56. The encoding device according to claim 55, wherein the digital signal of the second group includes a monaural signal having a first quantization accuracy and a first sampling frequency, and a second higher-order attribute layer than the monaural signal . A plurality of channel signals having a quantization accuracy and a second sampling frequency, wherein the first group of digital signals has the second quantization accuracy and a second sampling frequency, and the second group of channel signals Channel signal equal to or greater than the number,
The main code generation means is means for generating a main code by compressing and encoding a monaural signal having a first quantization accuracy and a first sampling frequency.
The error signal encoding means includes:
And upgrading unit for generating a conversion signal the monophonic signal it from the attribute hierarchies upgraded to signals above SL second quantization precision and the second sampling frequency higher,
A second group subtracting unit for generating a difference between the component of the second group of channel signals and the converted signal as an error signal of the second group;
A first compression encoder that compresses and encodes the second group of error signals and outputs an error code; and an interchannel orthogonal transformer that generates a frequency domain signal by orthogonally transforming the first group of channel signals. ,
A first group subtracting unit that generates a difference between at least one of the frequency domain signals and the converted signal as a second group error signal;
And a first group subtracting unit that generates an error signal between the frequency domain signal and the second group error signal as a first group error signal. The decoding device according to claim 33,
The error signal is a digital error signal of one channel in the first group consisting of a plurality of channels.
Hierarchy decrypt signals or decoded signals thereof the lower of the attributes, the linear of the first 1 channel decoded digital signal of the second group of fewer number of channels group or 2 or more of a plurality of channels of digital signals It is a combination. 58. The decoding apparatus according to claim 57, wherein said signal synthesis means decodes a main code and reproduces said monaural signal;
A second loop decoding unit for decoding a second group of error codes to generate a second group error signal;
A first group decoding unit that decodes a first group of codes including at least one error code to generate a frequency domain signal and a first group error signal;
An upgrade unit for generating a converted signal obtained by upgrading the monaural signal to a second quantization accuracy and a second sampling frequency;
A second group adding unit for adding the converted signal and the second group error signal to reproduce a second group channel signal;
An orthogonal inverse transform unit that adds the transformed signal and the error signal of the first group, orthogonally transforms the addition result and the frequency domain to transform it into a time domain signal, and reproduces it as a channel signal of the first group;
It consists of. A computer-executable encoding program describing a processing procedure of a digital signal encoding method according to any one of claims 1 to 13, 39, 40 , 41, 51, and 52. A computer-executable decoding program describing a processing procedure of a decoding method for decoding a code of a digital signal according to any one of claims 21 to 32 , 45 , 46 , 47 , 53, 54.

Images