JP2008309875A - Speech encoding device, speech decoding device, speech encoding method, speech decoding method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To attain high-speed speech encoding and decoding suitable for high quality speech reproduction, by taking speech signal characteristics and auditory sense characteristics into consideration. <P>SOLUTION: A speech encoding and decoding device 111 for functioning as a speech encoding device converts input speech to a spectrum constituted of a modified discrete cosine transform (MDCT) coefficients by performing MDCT, and the time dependence of the MDCT coefficients is expressed by the difference or the ratio of each of prescribed middle section bands; and information, based on the difference or the ratio, is entropy encoded and transmitted to another speech encoding and decoding device 111 that function as the speech encoding device. When speech is encoded, a CPU 121 calculates the difference or the ratio at a certain time; and when the speech is decoded, the spectrum is restored on the basis of the difference or the ratio. Information, based on the MDCT coefficient at immediately prior to that becomes necessary for the processing by the CPU 121, and the information is stored in a storage section 125. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a voice encoding device, a voice decoding device, a voice encoding method, a voice decoding method, and a program, which are required when executing voice compression / decompression considering auditory characteristics.

  In voice communication performed under a situation where the communication capacity is limited, it is necessary to devise voice encoding and voice decoding so that voice quality with as high quality as possible can be restored with as little data as possible.

  One direction of such a device is to make effective use of human auditory characteristics.

  As a speech coding method considering auditory characteristics, a method is known in which a speech signal is converted into a spectrum, and then the spectrum is divided into a plurality of subbands while considering a critical band derived from the auditory characteristics. (For example, refer to Patent Document 1 and Non-Patent Document 1).

In such a method, for each subband described above, the signal value, masking amount, noise, and the like are taken into account, and after the number of bits necessary for encoding is calculated, encoding is performed.
JP 7-46137 A JIS standard number JISX4323 “1.5Mbit / s encoding of moving image signals and accompanying sound signals for digital recording media-Part 3 Sound”, p. 96 [online], [searched August 7, 2006], Internet (URL: http://www.jisc.go.jp/app/pager?id=22028)

  However, in this method, the procedure for calculating the number of bits necessary for encoding is complicated, and many calculation steps are required. This is because, for example, it is not easy to calculate the masking amount.

  Therefore, when such a method is employed, the processing load of an arithmetic device such as a CPU inside the encoding device or the like becomes large, and the processing speed may be reduced. Then, for example, in applications such as mobile phones, it becomes difficult to make a mutual call in real time.

  Therefore, there is a need for an encoding / decoding device that can perform high-speed audio encoding / decoding processing in consideration of auditory characteristics and that can perform real-time calls and the like with sound quality that is practically acceptable.

  The present invention has been made in view of the above circumstances. That is, in a situation where communication capacity is limited, in speech coding, shortening the code length by paying attention to the continuity and continuity of the speech signal, and speeding up by band-based signal processing considering auditory characteristics In speech decoding, a speech encoding device, speech decoding device, speech encoding method, speech decoding method, and program that can restore speech of high quality without any practical problem at high speed are provided. The purpose is to do.

In order to achieve the above object, a speech encoding apparatus according to the first aspect of the present invention provides:
Discrete spectrum conversion means for obtaining a value of a frequency component for each sub-band having a predetermined bandwidth for each predetermined time segment for the digital audio signal;
For each medium partition band composed of a predetermined number of consecutive sub-compartment bands set in advance according to auditory characteristics, for each predetermined time section, out of the frequency component values belonging to the medium-compartment band A maximum value search means for searching for a maximum value;
Maximum value storage means for storing the maximum value searched by the maximum value search means;
Encoding means for quantizing and entropy-encoding and outputting information generated based on the maximum value and a frequency component normalized by division by the maximum value;
With
The maximum value search means includes:
For each predetermined time segment, the current maximum value that is the maximum value searched in the predetermined time segment is stored in the maximum value storage means, and a predetermined time segment that is earlier in time than the predetermined time segment Obtaining the past maximum value, which is the maximum value stored in the maximum value storage means, from the maximum value storage means, and converting the current maximum value into a value associated with the past maximum value,
It is characterized by that.

  Because of the continuity and continuity of the audio signal, the information content can be biased by associating the current maximum value with the past maximum value. Since such biased information is entropy-encoded, it can be encoded with high efficiency.

  A medium partition identification integer is assigned to the medium partition band in order from the low range, and the medium partition band is configured such that the logarithm of the center frequency of the medium partition band linearly depends on the medium partition identification integer. It is desirable to further comprise a zone band configuration means.

  The human auditory sense is that the lower the frequency, the more sensitive to a slight difference in frequency, and the sensitivity changes logarithmically with frequency. Therefore, it is suitable for a speech coding apparatus considering auditory characteristics to further include such a medium zone band forming means.

  The maximum value search means calculates a difference that is a value obtained by subtracting the past maximum value from the current maximum value, and the encoding means quantizes the difference and the normalized frequency component to perform entropy encoding. May be output.

  Due to the continuity and continuity of the audio signal, the value appearing as the difference is biased toward a smaller value than the value appearing as the current maximum value itself. Therefore, biased information is entropy encoded, and can be encoded with high efficiency.

  Alternatively, the maximum value search means obtains a ratio that is a value obtained by dividing the current maximum value by the past maximum value, and the encoding means quantizes the ratio and the normalized frequency component to entropy. You may make it encode and output.

  Since the value appearing as the ratio is biased to the vicinity of 1, encoding can be performed with high efficiency.

  Further comprising a maximum difference determining means, wherein the maximum value searching means obtains a difference that is a value obtained by subtracting the past maximum value from the current maximum value, and the maximum difference determining means is configured such that the maximum value searching means A maximum difference that is a maximum value among the differences determined for each of the medium partition bands is obtained, and the encoding unit quantizes and entropy-encodes the maximum difference and the standardized frequency component and outputs the result. It may be.

  Since only the maximum difference is encoded as information regarding the difference, the amount of code can be reduced.

  Alternatively, a maximum ratio determining means is further provided, wherein the maximum value searching means obtains a ratio that is a value obtained by dividing the current maximum value by the past maximum value, and the maximum ratio determining means includes all the maximum value searching means. The maximum ratio, which is the maximum value among the ratios determined for each of the medium partition bands, is obtained, and the encoding unit quantizes and entropy-encodes the maximum ratio and the normalized frequency component and outputs the result. You may do it.

  Since only the maximum ratio is encoded as information relating to the ratio, the amount of code can be reduced. In addition, since the spectrum shape of the audio signal often changes with time while maintaining similarity, a decrease in accuracy during encoding is suppressed.

  The discrete spectrum conversion means uses, for example, MDCT (Modified Discrete Cosine Transform).

In order to achieve the above object, a speech decoding apparatus according to the second aspect of the present invention provides:
Entropy coding is performed on the quantized spectrum of the audio signal for each predetermined time segment, which is a result of performing deformation including normalization for each band, and the normalization value that is a value used for the normalization. Receiving means for receiving the code generated by
Decoding means for decoding the modified spectrum data and the normalization value for each of the predetermined time intervals from the code by a decoding method corresponding to the entropy encoding;
Inverse deformation means for restoring the quantized spectrum for each of the predetermined time segments using the standardized value decoded from the deformed spectrum data decoded;
Normalization value storage means for storing the normalization value;
Discrete spectrum inverse transform means for restoring the speech signal from the restored quantized spectrum;
With
The reverse deformation means includes
For each predetermined time segment, the current standardization value, which is a standardization value decoded in the predetermined time segment, is stored in the standardization value storage means, and more time-dependent than the predetermined time segment. A past standardization value that is a standardization value stored in the standardization value storage means in a predetermined past time period, and based on the current standardization value and the past standardization value Restore the quantized spectrum,
It is characterized by that.

In order to achieve the above object, a speech encoding method according to a third aspect of the present invention includes:
For a digital audio signal, for each predetermined time segment, a discrete spectrum conversion step for obtaining a value of a frequency component for each sub-compartment band having a predetermined bandwidth;
For each medium partition band composed of a predetermined number of consecutive sub-compartment bands set in advance according to auditory characteristics, for each predetermined time section, out of the frequency component values belonging to the medium-compartment band A maximum value search step for searching for a maximum value;
A maximum value storing step for storing the maximum value searched by the maximum value searching step;
An encoding step of quantizing and entropy-encoding and outputting information generated based on the maximum value and a frequency component normalized by division by the maximum value;
Consisting of
The maximum value search step includes:
For each predetermined time segment, when the current maximum value, which is the maximum value searched in the predetermined time segment, is stored by the maximum value storing step, a predetermined past in time than the predetermined time segment is stored. Obtain the past maximum value that is the maximum value stored in the past maximum value storage step in the time segment, and convert the current maximum value to a value associated with the past maximum value,
It is characterized by that.

In order to achieve the above object, a speech decoding method according to the fourth aspect of the present invention provides:
Entropy coding is performed on the quantized spectrum of the audio signal for each predetermined time segment, which is a result of performing deformation including normalization for each band, and the normalization value that is a value used for the normalization. A receiving step for receiving a code generated by
A decoding step of decoding the modified spectrum data and the normalization value for each of the predetermined time intervals from the code by a decoding method corresponding to the entropy encoding;
An inverse transformation step of restoring the quantized spectrum for each predetermined time segment using the standardized value decoded from the decoded spectral data,
A normalization value storing step for storing the normalization value;
A discrete spectrum inverse transform step of restoring the speech signal from the restored quantized spectrum;
Consisting of
The reverse deformation step includes
For each predetermined time segment, when the current standardization value, which is a standardization value decoded in the predetermined time segment, is stored by the standardization value storage step, A past standardization value that is a standardization value stored in the past standardization value storage step in a predetermined time segment in the past is acquired, and based on the current standardization value and the past standardization value To restore the quantized spectrum,
It is characterized by that.

In order to achieve the above object, a program according to the fifth aspect of the present invention provides:
On the computer,
For a digital audio signal, for each predetermined time segment, a discrete spectrum conversion step for obtaining a value of a frequency component for each sub-compartment band having a predetermined bandwidth;
For each medium partition band composed of a predetermined number of consecutive sub-compartment bands set in advance according to auditory characteristics, for each predetermined time section, out of the frequency component values belonging to the medium-compartment band A maximum value search step for searching for a maximum value;
A maximum value storing step for storing the maximum value searched by the maximum value searching step;
An encoding step of quantizing and entropy-encoding and outputting information generated based on the maximum value and a frequency component normalized by division by the maximum value;
A program for executing
The maximum value search step includes:
For each predetermined time segment, when the current maximum value, which is the maximum value searched in the predetermined time segment, is stored by the maximum value storing step, a predetermined past in time than the predetermined time segment is stored. Obtain the past maximum value that is the maximum value stored in the past maximum value storage step in the time segment, and convert the current maximum value to a value associated with the past maximum value,
It is characterized by that.

In order to achieve the above object, a program according to the sixth aspect of the present invention provides:
On the computer,
Entropy coding is performed on the quantized spectrum of the audio signal for each predetermined time segment, which is a result of performing deformation including normalization for each band, and the normalization value that is a value used for the normalization. A receiving step for receiving a code generated by
A decoding step of decoding the modified spectrum data and the normalization value for each of the predetermined time intervals from the code by a decoding method corresponding to the entropy encoding;
An inverse transformation step of restoring the quantized spectrum for each predetermined time segment using the standardized value decoded from the decoded spectral data,
A normalization value storing step for storing the normalization value;
A discrete spectrum inverse transform step of restoring the speech signal from the restored quantized spectrum;
A program for executing
The reverse deformation step includes
For each predetermined time segment, when the current standardization value, which is a standardization value decoded in the predetermined time segment, is stored by the standardization value storage step, A past standardization value that is a standardization value stored in the past standardization value storage step in a predetermined time segment in the past is acquired, and based on the current standardization value and the past standardization value To restore the quantized spectrum,
It is characterized by that.

  According to the present invention, the audio signal is processed for each band in consideration of the characteristics of the audio signal and the auditory characteristics. Therefore, it is possible to encode and decode an audio signal at high speed and lightly while ensuring high sound quality.

  The speech encoding apparatus and speech decoding apparatus according to embodiments of the present invention will be described in detail below.

  From the viewpoint of ensuring convenience for the user, the speech encoding device and speech decoding device are integrated into a single device as a speech encoding / decoding device.

(Embodiment 1)
FIG. 1 shows a speech encoding / decoding device 111 according to this embodiment. As the device, for example, a mobile phone is assumed.

  The speech encoding / decoding device 111 includes a CPU 121, a ROM (Read Only Memory) 123, a storage unit 125, a speech processing unit 141, a wireless communication unit 161, and an operation key input content processing unit 171. These are connected to each other via a system bus 181. The system bus 181 is a transmission path for transferring commands and data.

  The ROM 123 stores an operation program for voice encoding and decoding.

  The storage unit 125 includes a RAM (Random Access Memory) 131 and a hard disk 133. The digital audio signal, the MDCT coefficient, the maximum value of the MDCT coefficient for each band, and the change in the maximum value for each predetermined time interval. Remember quantity etc. In particular, in this embodiment, the speech encoding / decoding device 111 requires information based on the speech signal of the immediately preceding time for processing at a certain time in both speech encoding and speech decoding. The unit 125 plays an important role as a delay processing buffer memory for temporarily storing such information.

  The audio encoding / decoding device 111 further includes a microphone 151, a speaker 153, an antenna 163, and operation keys 173.

  The microphone 151 collects the user's voice on the transmission side, that is, the encoding side, and delivers it to the voice processing unit 141. The speaker 153 issues the restored voice delivered from the voice processing unit 141 to the user on the receiving side, that is, the decoding side. The antenna 163 transmits the code delivered as a radio signal from the wireless communication unit 161 to the speech encoding / decoding device 111 on the reception side, that is, the decoding side, or from the speech encoding / decoding device 111 on the transmission side, that is, the encoding side. The transmitted wireless signal is received and delivered to the wireless communication unit 161. The operation key 173 is used to change an initial set value given in advance, for example, a boundary frequency of various bands for signal processing by the user's own judgment, or when the user on the transmission side, that is, the encoding side, This is used to transmit the user's intention to the device 111 when specifying the receiving-side and decoding-side devices 111 as counterparts.

  The voice processing unit 141, the wireless communication unit 161, and the operation key input content processing unit 171 are under the control of the CPU 121 via the system bus 181.

  The sound input to the microphone 151 is converted into a digital sound signal by, for example, 16 kHz sampling and 16-bit quantization by an A / D converter (not shown) inside the sound processing unit 141.

  The digital audio signal is sequentially sent to the storage unit 125 while being time-divided into frames, which are basic processing units of audio signal compression, by the audio processing unit 141.

  As will be described later, a digital audio signal of one frame is regarded as a group, and is stored in the storage unit 125, converted into a frequency domain by the CPU 121, transmitted to the wireless communication unit 161, and wirelessly transmitted by the antenna 163. It passes.

  For example, it is assumed that a signal of a certain frame existing in the storage unit 125 has been processed by the CPU 121 and has been transmitted to the wireless communication unit 161. Then, the data related to the signal of the frame is deleted from the storage unit 125 from the storage unit 125. Then, the signal of the next frame is delivered from the audio processing unit 141 to the storage unit 125.

  As described above, as long as the audio signal is continuously input, the signal processing is performed in units of frames from one to the next without generating an empty process. By adopting such a chain processing method, it is possible to perform real-time processing of an audio signal necessary for a mobile phone.

  However, the frame is a basic processing unit as described above. In this embodiment, as will be described later, in addition to the processing for each frame, processing focusing on the difference between the digital audio signals of two adjacent frames on the time axis is executed. Is the basic processing unit.

  In the following, for ease of understanding, first, it is assumed that the voice is input to the microphone 151 only over a time corresponding to one frame corresponding to a certain time t.

Assuming that one frame is composed of M signal values, an audio signal input to the microphone 151 is converted into a digital audio signal x ₀ ,..., X _M−1 by the audio processing unit 141 and stored in the storage unit 125. Suppose it was delivered. Data movement between components in the apparatus 111 is performed using the system bus 181 in accordance with instructions from the CPU 121. An instruction from the CPU 121 is issued according to an operation program stored in the ROM 123.

Digital audio signals x ₀ ,..., X _M−1 stored in the storage unit 125 are loaded into one of general-purpose registers (not shown) of the CPU 121. The digital audio signals x ₀ ,..., X _{M−1 that} are real time domain signals are converted into frequency domain signals X ₀ ,..., X _{M / 2-1} by the CPU 121 and stored in general-purpose registers. Is done. The conversion method may be any method as long as it converts a signal in the real time domain to a signal in the frequency domain, but since the imaginary part does not occur in the converted numerical value, the handling becomes easy. It is preferable to employ a modified discrete cosine transform (MDCT).

  The reason why M signal values in the real time domain correspond to M / 2 frequency conversion coefficient values in the frequency domain as described above is because MDCT is used for frequency conversion. In the case of other frequency conversion methods, the number of data in the real time domain and the number of data in the frequency domain do not always have a 2: 1 ratio, but in that case, the number attached to the final value of the frequency coefficient is The following description is applied as it is when read appropriately.

FIG. 2A schematically shows the MDCT coefficient thus generated. FIG. 2B is an enlarged view of a part thereof. Since MDCT is a kind of discrete frequency conversion, one frequency conversion coefficient is assigned to each of the M / 2 sub-compartment bands generated by dividing the frequency axis. As shown in the figure, the number _k is assigned to the (k + 1) th sub-band from the low frequency side, and the frequency conversion coefficient X _k is assigned (where 0 ≦ k ≦ M / 2-1). .) X _k is called an MDCT coefficient.

  One MDCT is performed per time section having a finite time length. Such a time segment is called an MDCT block. The number of signal samples included in one MDCT block is referred to as the MDCT order. For example, 512 is preferable as the order of MDCT.

  Basically, the time length of the MDCT block must not exceed the time length of one frame because the frame is a processing unit of audio compression. On the other hand, one frame may include a plurality of MDCT blocks. For example, it is preferable that one frame includes four MDCT blocks.

  However, here, in order to facilitate understanding by extracting only the essence of the invention, it is assumed that one frame has a one-to-one correspondence with one MDCT block. In other words, it is assumed that one frame corresponds to one MDCT block as it is. Then, in the schematic diagrams of the MDCT coefficients in FIG. 2 and subsequent figures, since M real-time signal values are included in one frame, the order of MDCT is M.

  In FIG. 2 and subsequent figures, all MDCT coefficients are drawn as if they were positive values, but this is only for easy understanding. The actual MDCT coefficient may take a negative value. In such a case, any known method such as providing a bit for representing a code may be used. As described above, the drawings relating to the MDCT coefficients in FIG. 2 and subsequent figures are schematic diagrams for explanation only.

For the MDCT coefficient X _k (0 ≦ k ≦ M / 2-1) stored in the general-purpose register, the CPU 121 changes the symbol for identifying each MDCT coefficient in order to perform subsequent processing smoothly. The replacement is performed by the CPU 121 according to the operation program read from the ROM 123. Specifically, each MDCT coefficient is re-identified with two symbols in addition to time t as follows.

First, as shown in FIG. 3 (a), the entire frequency region is divided into ω _MaxRANGE medium partition bands, and numbers such as 1, 2,..., Ω _MaxRANGE are assigned from the low frequency side. Distinguish between bands.

  One of the new symbols for identifying MDCT coefficients is this number.

  The frequency domain is divided by the medium partition band so that the logarithm of the center frequency of each medium partition band linearly depends on the number. In other words, the operation program read from the ROM 123 by the CPU 121 includes an instruction for performing such division. According to such division, the bandwidth becomes wider as the middle zone band of the high frequency region. FIG. 3A schematically shows the state.

  The reason for performing the division based on the logarithm is that the sensitivity to the frequency difference in human hearing is logarithmically lower as the high frequency component. Therefore, in order to transmit audio signals as effectively as possible with limited communication capacity, low-frequency components can be reproduced in detail to ensure playback sound quality, while high-frequency components are roughly It is appropriate to transmit only the correct information so that the total amount of information is reduced.

  For example, when the sound input to the microphone 151 is converted into a digital signal at a sampling frequency of 16 kHz in the sound processing unit 141, 11 medium partition bands are provided in the operation program stored in the ROM 123. It is preferable to set the boundary of the partition band as 187.5 Hz, 437.5 Hz, 687.5 Hz, 937.5 Hz, 1312.5 Hz, 1687.5 Hz, 2312.5 Hz, 3250 Hz, 4625 Hz, 6500 Hz.

Next, it is determined what number each MDCT coefficient is counted from the low frequency side of the medium partition band to which the MDCT coefficient belongs. It is assumed that q (ω _RANGE ) MDCT coefficients are included in the middle partition band numbered ω _RANGE (1 ≦ ω _RANGE ≦ ω _MaxRANGE ).

Then, the MDCT coefficient is specified by two symbols indicating which medium partition band it belongs to and what number coefficient is counted from the low frequency side in the medium partition band. . That is, as shown in FIG. 2B, the MDCT coefficients that have been distinguished by the numbers 1 to M / 2-1 over the entire frequency so far are newly added to the ω _RANGE middle partition band at time t. (1 ≦ ω _RANGE ≦ ω _MaxRANGE ), X (ω _RANGE , 1, t), ..., X (ω _RANGE , q (ω _RANGE ), t) Will be distinguished. This state is shown in FIG. 3 (b) in which a part of FIG. 3 (a) is enlarged.

The CPU 121 stores the MDCT coefficients X (ω _RANGE , 1, t),..., X (ω _RANGE , q (ω _RANGE ), t) (1 ≦ ω _RANGE ≦ ω _MaxRANGE ) thus re-identified. Stored in the unit 125.

Further, at time t, the maximum value of the MDCT coefficient in the medium partition band represented by ω _RANGE is set to the medium partition band maximum value X _MAX (ω _RANGE , t).

  In the following, for ease of understanding, it is assumed that the resolution in the vertical axis direction of the graphs shown in FIGS. 2 and 3, that is, the number of bits allocated for digitization, is constant in all the medium partition bands. Alternatively, a different number of bits may be determined for each band. For example, after defining a large block band that is a collection of a plurality of continuous medium block bands, the accuracy in handling the MDCT coefficient is determined in advance for each large block band, and the auditory characteristics are taken into account. Thus, the accuracy may be increased as the large frequency band on the lower frequency side. This is because hearing has a characteristic that the lower the frequency, the more sensitive the volume. In the following description, the MDCT coefficient is used as it is when calculating the difference. However, various processes may be executed after taking the logarithm of the MDCT coefficient, and the logarithm of the MDCT coefficient may be returned to the original MDCT coefficient at the final stage.

  In the present embodiment, the speech encoding / decoding device 111 uses the MDCT coefficient at a time that is a time Δt before the time when sending and receiving the MDCT coefficient at a certain time t. In order to facilitate understanding, first, referring to FIGS. 4 to 6, operations performed by the speech encoding / decoding device 111 on the speech encoding side and the speech encoding / decoding device 111 on the speech decoding side, and both Outline the information exchanged between the two. Thereafter, a more detailed processing flow will be described with reference to the flowcharts in FIG.

  The feature of this embodiment is that information based on a change in spectrum between time t-Δt and time t, that is, a difference is exchanged. Therefore, as a premise, at the start of communication from the speech encoding / decoding device 111 on the speech encoding side to the speech encoding / decoding device 111 on the speech decoding side, the MDCT coefficient necessary as an initial value is determined by any known method. It is assumed that the former device is transmitted to the latter device. In addition, when communication takes a long time, it may be impossible to ignore an error caused by difference accumulation. In order to deal with this, a refresh rate may be determined in advance, and initialization processing similar to that at the start of communication may be performed at a certain frequency. Hereinafter, only difference transmission / reception, which is a characteristic process in the present embodiment, will be described.

  4 to 6, the speech encoding / decoding device 111 on the speech encoding side is depicted on the left side, and the speech encoding / decoding device 111 on the speech decoding side is depicted on the right side. Hereinafter, each device is simply referred to as a transmitter and a receiver. 4 to 6, components other than the storage unit 125 and the antenna 163 are omitted from the components of the speech encoding / decoding device 111 illustrated in FIG.

First, as shown in FIG. 4A, the maximum value X _MAX (ω _RANGE , t-Δt) of the MDCT coefficient in the middle zone band at time t-Δt is stored in both the storage unit 125 of the transmitter and the receiver. Is stored. When the time reaches t, the CPU 121 of the transmitter calculates the MDCT coefficient at the time t and stores it in the storage unit 125 of the transmitter (see FIG. 4A).

Subsequently, the CPU 121 of the transmitter performs a search in the middle partition band, calculates a maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the middle partition band at time t, and stores it in the storage unit 125. At this time, as shown in FIG. 4 (b), the storage unit 125 of the transmitter has the maximum value of the MDCT coefficient in the middle zone band at time t-Δt and t, and the MDCT coefficient at time t. Stored. In the storage unit 125 of the receiver, the maximum value X _MAX (ω _RANGE , t−Δt) of the MDCT coefficient in the middle partition band at the time t−Δt is stored, and there is no change.

The CPU 121 of the transmitter stores the maximum value X _MAX (ω _RANGE , t-Δt) of the MDCT coefficient in the medium partition band at time t-Δt stored in the storage unit 125 of the transmitter in the storage unit 125 as well. The difference value of the maximum value at time t is obtained by subtracting it from the maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the medium partition band at time t, and stored in the storage unit 125. After this, the transmitter does not need the maximum value X _MAX (ω _RANGE , t−Δt) of the MDCT coefficient in the middle partition band at time t−Δt. Therefore, the storage capacity of the storage unit 125 of the transmitter may be deleted so as not to be compressed. The CPU 121 of the transmitter further uses the MDCT coefficient at time t stored in the storage unit 125 of the transmitter as the maximum value X _MAX ( By dividing by ω _RANGE , t), the normalized value of the MDCT coefficient at time t is obtained and stored in the storage unit 125. At this time, as shown in FIG. 5 (a), in the storage unit 125 of the transmitter, the maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the medium partition band at time t and the maximum at time t are stored. The difference value between the values and the normalized value of the MDCT coefficient at time t are stored. In the storage unit 125 of the receiver, the maximum value X _MAX (ω _RANGE , t−Δt) of the MDCT coefficient in the middle partition band at the time t−Δt is stored, and there is no change.

  The difference value of the maximum value at time t stored in the storage unit 125 of the transmitter and the normalized value of the MDCT coefficient at time t are extracted from the storage unit 125 and quantized by the CPU 121 of the transmitter, Entropy-encoded and wirelessly transmitted from the antenna 163 of the transmitter. The radio signal on which the code generated by the entropy encoding is superimposed is captured by the receiver antenna 163 in the receiver. This state is schematically shown in FIG. Note that representative entropy encoding methods include Huffman code and RangeCoder.

The code captured by the receiver antenna 163 is decoded by the CPU 121 of the receiver. The difference value of the maximum value at time t and the normalized value of the MDCT coefficient at time t, which are generated as a result of decoding, are stored in the storage unit 125 of the receiver. At this time, as shown in FIG. 6A, the storage unit 125 of the receiver stores a maximum value X _MAX (ω _RANGE , t−Δt) of MDCT coefficients in the middle partition band at time t−Δt, The difference value of the maximum value at time t and the normalized value of the MDCT coefficient at time t are stored. In the storage unit 125 of the transmitter, the maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the middle zone band at time t is left.

Similarly, the CPU 121 of the receiver stores the maximum value X _MAX (ω _RANGE , t−Δt) of the MDCT coefficient in the middle zone band at time t−Δt stored in the storage unit 125 of the receiver in the storage unit 125. The maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the medium partition band at time t is obtained by adding the difference value of the maximum value at time t and stored in the storage unit 125. Thereafter, the maximum value X _MAX (ω _RANGE , t-Δt) of the MDCT coefficient in the middle zone band at time t-Δt and the difference value between the maximum values at time t are unnecessary, so these values are not necessary. May be deleted. The CPU 121 of the receiver then continues to the MDCT coefficient standardized value at the time t stored in the storage unit 125 of the receiver, and the MDCT coefficient in the middle partition band at the time t stored in the storage unit 125. By multiplying the maximum value X _MAX (ω _RANGE , t), the MDCT coefficient at time t is obtained and stored in the storage unit 125. At this time, as shown in FIG. 6B, the storage unit 125 of the receiver stores the maximum MDCT coefficient X _MAX (ω _RANGE , t) in the medium partition band at time t and the MDCT at time t. The coefficient is stored. In the storage unit 125 of the transmitter, the maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the middle zone band at time t is left.

  In this way, the MDCT coefficient at time t initially stored in the transmitter storage unit 125 as shown in FIG. 4A is stored in the storage unit 125 of the receiver as shown in FIG. 6B. Stored. This means that information about the spectrum has been transmitted from the transmitter to the receiver. Thereafter, the receiver can restore the audio signal input to the transmitter by frequency inverse transformation or the like.

In FIG. 4A, the maximum value X _MAX (ω _RANGE , t−Δt) of the MDCT coefficient in the middle partition band at time t−Δt is stored in both the storage units 125 of the transmitter and the receiver. Correspondingly, in FIG. 6B, the maximum value X _MAX (ω _RANGE , t) of the MDCT coefficient in the middle zone band at time t is stored in both the storage units 125 of the transmitter and the receiver. Has been. Therefore, after time t + Δt, it is possible to transmit the MCDT coefficient at each time from the transmitter to the receiver by repeating the same processing as that shown in FIGS.

  The MDCT coefficient itself takes various values. On the other hand, because of the temporal continuity of the audio signal, a relatively small value appears with high frequency as the above-described maximum difference value. This tendency is more prominent in the time zone when the audio signal is in a steady state. Such biased information has high compression efficiency by entropy coding. Therefore, according to the present embodiment, it is possible to restore high-quality speech for the transmission rate as compared with the case where the MDCT coefficient itself is simply encoded.

  In order to facilitate understanding, in the above description using FIGS. 4 to 6, both the time interval for obtaining the MDCT coefficient and the time interval of the encoding process are represented by Δt. The time intervals need not be equal. For example, as long as the real-time feeling in a voice call is not impaired, a plurality of sets of MDCT coefficients calculated from voice signals in several continuous time zones are stored in the storage unit 125 of the transmitter and then quantized in a batch. Entropy encoding may be performed.

  Hereinafter, the above-described processing flow will be described with reference to flowcharts. FIG. 7 is a flowchart showing the flow of middle partition bandwidth maximum value search, middle partition bandwidth difference computation, and normalized MDCT coefficient computation performed at the transmitter at time t. It is assumed that the digital audio signal has already been subjected to MDCT, and the MDCT coefficient is stored in the storage unit 125 of the transmitter.

The CPU 121 of the transmitter initializes the band identification variable ω _RANGE to 1 (step S711), and stores MDCT coefficients X (ω _RANGE , 1, t),..., X (ω _RANGE , q (ω _RANGE ) from the storage unit 125. ), T) are loaded (step S713), and the maximum value X _MAX (ω _RANGE , t), which is the maximum value of these loaded MDCT coefficients, is obtained (step S715), and X _MAX (ω _RANGE , T) are stored in the storage unit 125 (step S717).

The reason why X _MAX (ω _RANGE , t) is stored in the storage unit 125 in step S717 is that it is necessary for processing at time t + Δt, which is the next time.

The CPU 121 loads the medium partition band maximum value X _MAX (ω _RANGE , t−Δt) immediately before from the storage unit 125 (step S719).

In step S719, the CPU 121 can load X _MAX (ω _RANGE , t−Δt) from the storage unit 125 in a step corresponding to step S717 at the immediately preceding time, in which X _MAX (ω _RANGE , t−Δt) is stored. This is because it is stored in the unit 125.

The CPU 121 calculates the middle partition band difference ΔX _MAX (ω _RANGE , t) by ΔX _MAX (ω _RANGE , t) = X _MAX (ω _RANGE , t) −X _MAX (ω _RANGE , t−Δt) (step S721), and stored in the storage unit 125 (step S723). The stored ΔX _MAX (ω _RANGE , t) is to be encoded. Subsequently, the CPU 121 calculates the normalized MDCT coefficients X _REG (ω _RANGE , 1, t),..., X _REG (ω _RANGE , q (ω _RANGE ), t), and X _REG (ω _RANGE , 1, t). = X (ω _RANGE , 1, t) / X _MAX (ω _RANGE , t), ..., X _REG (ω _RANGE , q (ω _RANGE ), t) = X (ω _RANGE , q (ω _RANGE ), t) / X _MAX (ω _RANGE , t) (step S725) and stored in the storage unit 125 (step S727). The stored X _REG (ω _RANGE , 1, t),..., X _REG (ω _RANGE , q (ω _RANGE ), t) are to be encoded. The CPU 121 further determines whether or not the processing has been completed for all the medium-compartment bandwidths (step S729). If it is determined that the processing has ended (step S729; Yes), the processing is ended and it is determined that the processing has not ended. In the case (step S729; No), ω _RANGE is incremented by 1 to process the next band (step S731), and the process returns to step S713.

The flow of the process performed at the receiver at time t corresponding to the process illustrated in the flowchart of FIG. 7 described above performed at the transmitter at time t will be described with reference to the flowchart illustrated in FIG. The receiver calculates the maximum value of the medium zone band and the MDCT coefficient. It should be noted that the medium zone band difference ΔX _MAX (ω _RANGE , t) and the normalized MDCT coefficient X _REG (ω _RANGE , 1, t),..., X that are entropy encoded by the transmitter and then transmitted to the receiver _It is assumed that _REG (ω _RANGE , q (ω _RANGE ), t) has already been decoded and stored in the storage unit 125 of the receiver.

The CPU 121 of the receiver initializes the band identification variable ω _RANGE to 1 (step S741), and loads the medium zone maximum band value X _MAX (ω _RANGE , t−Δt) immediately before from the storage unit 125 (step S743). , The medium partition band difference ΔX _MAX (ω _RANGE , t) is loaded (step S745), and the medium partition band maximum value X _MAX (ω _RANGE , t) is set to X _MAX (ω _RANGE , t) = X _MAX (ω _RANGE , t−Δt) + ΔX _MAX (ω _RANGE , t) is obtained (step S747), and X _MAX (ω _RANGE , t) is stored in the storage unit 125 (step S749).

The reason why X _MAX (ω _RANGE , t) is stored in the storage unit 125 in step S749 is that it is necessary for processing at time t + Δt, which is the next time. In addition, because the processing corresponding to step S749 was performed at t-Δt, which is the previous time, the CPU 121 stores X _MAX (ω _RANGE , t-Δt) from the storage unit 125 in the above-described step S743 at time t. Could be loaded.

The CPU 121 loads standardized MDCT coefficients X _REG (ω _RANGE , 1, t),..., X _REG (ω _RANGE , q (ω _RANGE ), t) (step S751), and MDCT coefficient X (ω _RANGE , 1, t), ..., X (ω _RANGE , q (ω _RANGE ), t), X (ω _RANGE , 1, t) = X _REG (ω _RANGE , 1, t) × X _MAX (ω _RANGE , t), ..., X (ω _RANGE , q (ω _RANGE ), t) = X _REG (ω _RANGE , q (ω _RANGE ), t) × X _MAX (ω _RANGE , t) (Step S753), X (ω _RANGE , 1, t),..., X (ω _RANGE , q (ω _RANGE ), t) are stored in the storage unit 125 (Step S755). These MDCT coefficients are subjected to well-known processing such as conversion to a real-time domain, thereby restoring the audio signal. Further, the CPU 121 determines whether or not the processing has been completed for all the medium partition bands (step S757). If it is determined that the processing has ended (step S757; Yes), the processing is ended, and it is determined that the processing has not ended. In the case (step S757; No), ω _RANGE is incremented by 1 to process the next band (step S759), and the process returns to step S743.

(Modification of Embodiment 1)
Hereinafter, a speech encoding / decoding device according to a modification of the first embodiment of the present invention will be described. The outline of the apparatus is the same as that of the speech encoding / decoding apparatus 111 according to the first embodiment.

  In the first embodiment, the difference is used as the amount representing the change in the maximum value of the middle zone band. On the other hand, the ratio is used in this modification. The contents of the processing in both are almost the same.

The process performed by the transmitter is a process obtained by changing a part of the flowchart shown in FIG. That is, in step S721 in FIG. 7, the medium partition band ratio RaX _MAX (ω _RANGE , t) is set to RaX _MAX (ω _RANGE , t) = X _MAX (ω _RANGE , t) / X _MAX (ω _RANGE , t−Δt). Change to calculate by In step S723, the storage unit 125 is changed to store RaX _MAX (ω _RANGE , t).

The process performed by the receiver is a process obtained by changing a part of the flowchart shown in FIG. That is, in step S745 in FIG. 8, the medium partition bandwidth ratio RaX _MAX (ω _RANGE , t) is changed to be loaded. In step S747, the maximum value X _MAX (ω _RANGE , t) of the medium partition band is obtained by X _MAX (ω _RANGE , t) = X _MAX (ω _RANGE , t−Δt) × RaX _MAX (ω _RANGE , t). Change as follows.

Since the value appearing as the middle zone bandwidth ratio RaX _MAX (ω _RANGE , t) is biased to the vicinity of 1, it can be encoded with high efficiency.

(Embodiment 2)
Hereinafter, a speech encoding / decoding device according to Embodiment 2 of the present invention will be described. The outline of the apparatus is the same as that of the speech encoding / decoding apparatus 111 according to the first embodiment. In addition, the operations performed by the transmitter and the receiver and the outline of information exchanged between them are substantially the same as those in the first embodiment described with reference to FIGS.

  In the first embodiment and its modifications, the difference values and ratios for all the middle zone bands are exchanged between the transceivers. On the other hand, in the present embodiment, only the maximum value among the difference values and ratios of the middle zone band is exchanged between the transceivers. Hereinafter, the flow of such processing will be described with reference to the flowcharts shown in FIGS.

  FIG. 9 is a flowchart showing a flow of middle partition bandwidth maximum value search, middle partition bandwidth difference calculation, and maximum difference computation performed at the transmitter at time t. It is assumed that the digital audio signal has already been subjected to MDCT, and the MDCT coefficient is stored in the storage unit 125 of the transmitter.

The CPU 121 of the transmitter initializes the maximum difference MaxΔX _MAX (t) to 0 (step S771), initializes the band identification variable ω _RANGE to 1 (step S773), and stores the MDCT coefficient X (ω _RANGE , 1 from the storage unit 125. , T),..., X (ω _RANGE , q (ω _RANGE ), t) are loaded (step S775), and the medium zone band maximum value X _MAX (the maximum value among these loaded MDCT coefficients) ω _RANGE , t) is obtained (step S777), and X _MAX (ω _RANGE , t) is stored in the storage unit 125 (step S779). The storage in step S779 is for use in processing at time t + Δt, which is the next time.

The CPU 121 loads the medium partition band maximum value X _MAX (ω _RANGE , t−Δt) immediately before from the storage unit 125 (step S781). Such loading is possible because a step corresponding to step S779 has been executed at the immediately preceding time.

The CPU 121 calculates the middle partition band difference ΔX _MAX (ω _RANGE , t) by ΔX _MAX (ω _RANGE , t) = X _MAX (ω _RANGE , t) −X _MAX (ω _RANGE , t−Δt) (step S783), it is determined whether or not ΔX _MAX (ω _RANGE , t) is equal to or greater than MaxΔX _MAX (t) (step S785). When it is determined that ΔX _MAX (ω _RANGE , t) is equal to or greater than MaxΔX _MAX (t) (step S785; Yes), MaxΔX _MAX (t) is set to MaxΔX _MAX (t) = ΔX _MAX (ω _RANGE , t) (Step S787), the process proceeds to step S789. If it is determined that ΔX _MAX (ω _RANGE , t) is not equal to or greater than MaxΔX _MAX (t) (step S785; No), the process immediately proceeds to step S789. In step S789, the CPU 121 determines whether or not the processing has been completed for all of the medium partition bands. If it is determined that the processing has ended (step S789; Yes), the process proceeds to step S793, and if it is determined that the processing has not ended. In (Step S789; No), ω _RANGE is incremented by 1 to process the next band (Step S791), and the process returns to Step S775. In step S793, the CPU 121 stores MaxΔX _MAX (t) in the storage unit 125, and thereafter ends the process. The MaxΔX _MAX (t) stored in step S793 is to be encoded.

  The CPU 121 of the transmitter, after finishing the process shown in the flowchart of FIG. 9, calculates the normalized MDCT coefficient when using the maximum difference by the process shown in the flowchart of FIG.

The CPU 121 of the transmitter loads the maximum difference MaxΔX _MAX (t) from the storage unit 125 (step S811), initializes the band identification variable ω _RANGE to 1 (step S813), and stores the MDCT coefficient X (ω _RANGE from the storage unit 125. , 1, t),..., X (ω _RANGE , q (ω _RANGE ), t) are loaded (step S 815), and the medium partition band maximum value X _MAX (ω _RANGE , t−Δt) immediately before is loaded. Load (step S817). Such loading is possible because a step corresponding to step S779 in FIG. 9 has been executed at the immediately preceding time. Subsequently, the CPU 121 calculates the normalized MDCT coefficients X _REG (ω _RANGE , 1, t),..., X _REG (ω _RANGE , q (ω _RANGE ), t), and X _REG (ω _RANGE , 1, t). = X (ω _RANGE , 1, t) / {X _MAX (ω _RANGE , t-Δt) + MaxΔX _MAX (t)}, ..., X _REG (ω _RANGE , q (ω _RANGE ), t) = X (ω _RANGE , q (ω _RANGE ), t) / {X _MAX (ω _RANGE , t−Δt) + MaxΔX _MAX (t)} (step S 819) and stored in the storage unit 125 (step S 821 ). The stored X _REG (ω _RANGE , 1, t),..., X _REG (ω _RANGE , q (ω _RANGE ), t) are to be encoded. The CPU 121 further determines whether or not the processing has been completed for all the medium-compartment bandwidths (step S823). If it is determined that the processing has ended (step S823; Yes), the processing ends and it is determined that the processing has not ended. In the case (step S823; No), ω _RANGE is increased by 1 to process the next band (step S825), and the process returns to step S815.

The flow of processing performed at the receiver at time t corresponding to the processing illustrated in the flowcharts of FIGS. 9 and 10 performed at the transmitter at time t will be described with reference to the flowchart illustrated in FIG. In the present embodiment, the receiver calculates the mid-zone band maximum value and the MDCT coefficient based on the maximum difference MaxΔX _MAX (t) transmitted from the transmitter. Note that the maximum difference MaxΔX _MAX (t) and the normalized MDCT coefficients X _REG (ω _RANGE , 1, t), ..., X _REG (ω _RANGE , It is assumed that q (ω _RANGE ), t) has already been decoded and stored in the storage unit 125 of the receiver.

The CPU 121 of the receiver loads the maximum difference MaxΔX _MAX (t) from the storage unit 125 (step S 831), initializes the band identification variable ω _RANGE to 1 (step S 833), and stores the medium partition band at the previous time from the storage unit 125. The maximum value X _MAX (ω _RANGE , t−Δt) is loaded (step S 835), and the normalized MDCT coefficient X _REG (ω _RANGE , 1, t),..., X _REG (ω _RANGE , q (ω _RANGE), t) to load the (step S837), MDCT coefficient _{X (ω RANGE, 1, t} ), ···, X (ω RANGE, q (ω RANGE) a, t), X (ω _RANGE , 1, t) = X _REG (ω _RANGE , 1, t) × {X _MAX (ω _RANGE , t-Δt) + MaxΔX _MAX (t)}, ..., X (ω _RANGE , q (ω _RANGE ) , T) = X _REG (ω _RANGE , q (ω _RANGE ), t) × {X _MAX (ω _RANGE , t−Δt) + MaxΔX _MAX (t)} (step S 839), and the storage unit 125. (Step S841). These MDCT coefficients are subjected to well-known processing such as conversion to a real-time domain, thereby restoring the audio signal.

Subsequently, the CPU 121 continuously determines the maximum value X of the medium partition band X which is the maximum value of X (ω _RANGE , 1, t),..., X (ω _RANGE , q (ω _RANGE ), t) obtained in step S839. _MAX (ω _RANGE , t) is obtained (step S843) and stored in the storage unit 125 (step S845). This is because the storage in step S845 is useful for processing at t + Δt, which is the next time. Note that X _MAX (ω _RANGE , t−Δt) can be loaded in step S835 because the step corresponding to step S845 was executed at t−Δt which is the previous time. The CPU 121 further determines whether or not the processing has been completed for all of the medium partition bands (step S847). If it is determined that the processing has ended (step S847; Yes), the processing is ended and it is determined that the processing has not ended. In the case (step S847; No), ω _RANGE is incremented by 1 to process the next band (step S849), and the process returns to step S835.

If step S725 in FIG. 7 is compared with step 819 in FIG. 10, the following is clear. That is, in the first embodiment, the divisor for obtaining the normalized MDCT coefficient is the maximum value of the MDCT coefficient for each medium partition band, so that the normalized MDCT coefficient can be obtained with the highest accuracy. On the other hand, in the present embodiment, since the value based on the maximum difference is adopted as the divisor for obtaining the normalized MDCT coefficient, the accuracy of the normalized MDCT coefficient obtained in comparison with the first embodiment is low. The accuracy of the MDCT coefficient restoration by the method is lower than that of the first embodiment. In other words, as apparent from the definition of the maximum difference MaxΔX _MAX (t) shown in FIG. 9, the divisor X _MAX (ω _RANGE , t) for normalization in the first embodiment and the normalization in the present embodiment. The divisor for X _MAX (ω _RANGE , t-Δt) + MaxΔX _MAX (t) and X _MAX (ω _RANGE , t) ≤ X _MAX (ω _RANGE , t-Δt) + MaxΔX _MAX ( The relationship t) holds. That is, in this embodiment, it is considered that there are many cases where the MDCT coefficient is divided by a value larger than necessary. In such a case, as a result, the normalized MDCT coefficient becomes a smaller value than necessary as a whole. By the way, it is reasonable that the number of bits for representing the normalized MDCT coefficient is determined in advance on the assumption that the normalized MDCT coefficient takes a value of 0 or more and 1 or less because of the nature of the operation of normalization. is there. Therefore, when the normalized MDCT coefficient becomes a value smaller than necessary as described above, the bit prepared to represent a number close to 1 is wasted and at the time of quantization in bit units. The error increases. In this sense, it can be said that the present embodiment performs voice encoding and decoding with lower accuracy than the first embodiment.

  However, in the case of the first embodiment, the difference in all the medium partition bands had to be exchanged between the transmitter and the receiver, whereas in the case of this embodiment, the maximum of the differences in all the medium partition bands. Only the value needs to be exchanged. Therefore, according to the present embodiment, compared to the first embodiment, the amount of data to be encoded can be reduced, which contributes to low bit rate communication.

(Modification of Embodiment 2)
Hereinafter, a speech encoding / decoding device according to a modification of the second embodiment of the present invention will be described. The outline of the apparatus is the same as that of the speech encoding / decoding apparatus 111 according to the first embodiment.

  In the second embodiment, the difference is used as the amount representing the change of the maximum value of the middle zone band. On the other hand, the ratio is used in this modification. The contents of the processing in both are almost the same.

The process performed by the transmitter is a process obtained by changing a part of the flowcharts of FIGS. 9 and 10 described above. That is, in step S771 in FIG. 9, the maximum difference MaxΔX _MAX (t) is replaced with the maximum ratio MaxRaX _MAX (t), and in step S783, the medium partition band ratio RaX _MAX (ω _RANGE , t) is changed to RaX _MAX (ω _RANGE , T) = X _MAX (ω _RANGE , t) / X _MAX (ω _RANGE , t−Δt), and in step S785, RaX _MAX (ω _RANGE , t) ≧ MaxRaX _MAX (t). In step S787, it is changed to update to MaxRaX _MAX (t) = RaX _MAX (ω _RANGE , t). In step S793, MaxRaX _MAX (t) is stored in the storage unit 125. In step S811 of FIG. 10, the maximum ratio MaxRaX _MAX (t) is changed to be loaded. In step S819, the normalized MDCT coefficient is set to X _REG (ω _RANGE , 1, t) = X (ω _RANGE , 1, t) / {X _MAX (ω _RANGE , t-Δt) × MaxRaX _MAX (t)}, ..., X _REG (ω _RANGE , q (ω _RANGE ), t) = X (ω _RANGE , q (ω _RANGE ), t) / {X _MAX (ω _RANGE , t−Δt) × MaxRaX _MAX (t)}

The process performed by the receiver is a process obtained by changing a part of the flowchart shown in FIG. That is, in step S831 in FIG. 11, the maximum ratio MaxRaX _MAX (t) is changed to be loaded, and in step S839, the MDCT coefficient is set to X (ω _RANGE , 1, t) = X _REG (ω _RANGE , 1, t). × {X _MAX (ω _RANGE , t-Δt) × MaxRaX _MAX (t)}, ..., X (ω _RANGE , q (ω _RANGE ), t) = X _REG (ω _RANGE , q (ω _RANGE ), t) × {X _MAX (ω _RANGE , t−Δt) × MaxRaX _MAX (t)}.

With respect to the ratio, only the maximum ratio MaxRaX _MAX (t) needs to be encoded, not the ratio for all of the medium partition bands, and this modification has the same effect as that of the second embodiment. In addition, there are the following effects.

  A spectrum is assumed in which the property of each middle section band is represented by the maximum value of the MDCT coefficient included in the middle section band. Then, due to the characteristics of speech, such a spectrum is similar so that the components of each mid-zone band are proportional with time rather than changing so that the entire band is raised or lowered with time. There is a strong tendency to change. Therefore, according to the present modification example that expresses the time change of the spectrum using the ratio instead of the difference, the degree to which the encoding accuracy is reduced because the divisor for normalization is too large compared to the second embodiment. Can be reduced.

  In addition, this invention is not limited to the said embodiment, A various deformation | transformation and application are possible. The above-described hardware configuration, block configuration, and flowchart are examples, and are not limited.

  For example, a mobile phone has been described as the speech encoding / decoding device 111 shown in FIG. 1, but the PHS (Personal Handyphone System), the PDA (Personal Digital Assistants), or a general personal computer has The invention can be easily applied. That is, the said embodiment is for description and does not restrict | limit the scope of the present invention.

It is a figure which shows the structure of the audio | voice encoding / decoding apparatus which concerns on embodiment of this invention. It is a figure which shows a mode that an audio | voice spectrum is represented by the MDCT coefficient which belongs to a small division zone | band. It is a figure which shows typically the medium division zone | band and the MDCT coefficient as a function of time in Embodiment 1 of this invention. It is a figure which shows the outline of the calculation which the audio | voice coding apparatus in Embodiment 1 of this invention performs. It is a figure which shows the outline of the information transmitted to the audio | voice decoding apparatus from the audio | voice encoding apparatus in Embodiment 1 of this invention. It is a figure which shows the outline of the calculation which the audio | voice decoding apparatus in Embodiment 1 of this invention performs. It is a figure which shows the flow of calculation of medium division zone maximum value search, middle division zone difference, and calculation of a normalization MDCT coefficient in Embodiment 1 of this invention. It is a figure which shows the flow of calculation of the middle division zone | band maximum value and MDCT coefficient in Embodiment 1 of this invention. It is a figure which shows the flow of the calculation of the middle division band maximum value search, the calculation of a middle division band difference, and the calculation of a maximum difference in Embodiment 2 of this invention. It is a figure which shows the flow of calculation of the normalization MDCT coefficient at the time of the largest difference use in Embodiment 2 of this invention. It is a figure which shows the flow of calculation of the medium division zone | band maximum value and MDCT coefficient based on the maximum difference in Embodiment 2 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 111 ... Speech encoding and decoding apparatus, 121 ... CPU, 123 ... ROM, 125 ... Memory | storage part, 131 ... RAM, 133 ... Hard disk, 141 ... Voice processing part, 151 ... Microphone, 153 ... Speaker, 161 ... Wireless communication unit, 163 ... Antenna, 171 ... Operation key input content processing unit, 173 ... Operation key, 181 ... System bus

Claims (12)

Discrete spectrum conversion means for obtaining a value of a frequency component for each sub-band having a predetermined bandwidth for each predetermined time segment for the digital audio signal;
For each medium partition band composed of a predetermined number of consecutive sub-compartment bands set in advance according to auditory characteristics, for each predetermined time section, out of the frequency component values belonging to the medium-compartment band A maximum value search means for searching for a maximum value;
Maximum value storage means for storing the maximum value searched by the maximum value search means;
Encoding means for quantizing and entropy-encoding and outputting information generated based on the maximum value and a frequency component normalized by division by the maximum value;
With
The maximum value search means includes:
For each predetermined time segment, the current maximum value that is the maximum value searched in the predetermined time segment is stored in the maximum value storage means, and a predetermined time segment that is earlier in time than the predetermined time segment Obtaining the past maximum value, which is the maximum value stored in the maximum value storage means, from the maximum value storage means, and converting the current maximum value into a value associated with the past maximum value,
A speech encoding apparatus characterized by that. A medium partition identification integer is assigned to the medium partition band in order from the low range, and the medium partition band is configured such that the logarithm of the center frequency of the medium partition band linearly depends on the medium partition identification integer. Further comprising a partition band configuration means,
The speech encoding apparatus according to claim 1. The maximum value search means includes:
Find a difference that is a value obtained by subtracting the past maximum value from the current maximum value,
The encoding means includes
The difference and the normalized frequency component are quantized and entropy encoded and output.
The speech encoding apparatus according to claim 1 or 2, characterized in that The maximum value search means includes:
Obtain a ratio that is a value obtained by dividing the current maximum value by the past maximum value,
The encoding means includes
The ratio and the normalized frequency component are quantized and entropy encoded and output.
The speech encoding apparatus according to claim 1 or 2, characterized in that A maximum difference determining means;
The maximum value search means includes:
Find a difference that is a value obtained by subtracting the past maximum value from the current maximum value,
The maximum difference determining means includes
Find the maximum difference that is the maximum value of the differences obtained by the maximum value search means for every medium partition band,
The encoding means includes
The maximum difference and the normalized frequency component are quantized and entropy encoded and output.
The speech encoding apparatus according to claim 1 or 2, characterized in that A maximum ratio determining means;
The maximum value search means includes:
Obtain a ratio that is a value obtained by dividing the current maximum value by the past maximum value,
The maximum ratio determining means includes
The maximum value search means calculates a maximum ratio that is the maximum value among the ratios determined for all the medium partition bands,
The encoding means includes
The maximum ratio and the normalized frequency component are quantized and entropy encoded and output.
The speech encoding apparatus according to claim 1 or 2, characterized in that The discrete spectrum conversion means includes
Use MDCT (Modified Discrete Cosine Transform),
The speech encoding apparatus according to claim 1, wherein the speech encoding apparatus is a part of the speech encoding apparatus. Entropy coding is performed on the quantized spectrum of the audio signal for each predetermined time segment, which is a result of performing deformation including normalization for each band, and the normalization value that is a value used for the normalization. Receiving means for receiving the code generated by
Decoding means for decoding the modified spectrum data and the normalization value for each of the predetermined time intervals from the code by a decoding method corresponding to the entropy encoding;
Inverse deformation means for restoring the quantized spectrum for each of the predetermined time segments using the standardized value decoded from the deformed spectrum data decoded;
Normalization value storage means for storing the normalization value;
Discrete spectrum inverse transform means for restoring the speech signal from the restored quantized spectrum;
With
The reverse deformation means includes
For each predetermined time segment, the current standardization value, which is a standardization value decoded in the predetermined time segment, is stored in the standardization value storage means, and more time-dependent than the predetermined time segment. A past standardization value that is a standardization value stored in the standardization value storage means in a predetermined past time period, and based on the current standardization value and the past standardization value Restore the quantized spectrum,
A speech decoding apparatus characterized by that. For a digital audio signal, for each predetermined time segment, a discrete spectrum conversion step for obtaining a value of a frequency component for each sub-compartment band having a predetermined bandwidth;
For each medium partition band composed of a predetermined number of consecutive sub-compartment bands set in advance according to auditory characteristics, for each predetermined time section, out of the frequency component values belonging to the medium-compartment band A maximum value search step for searching for a maximum value;
A maximum value storing step for storing the maximum value searched by the maximum value searching step;
An encoding step of quantizing and entropy-encoding and outputting information generated based on the maximum value and a frequency component normalized by division by the maximum value;
Consisting of
The maximum value search step includes:
For each predetermined time segment, when the current maximum value, which is the maximum value searched in the predetermined time segment, is stored by the maximum value storing step, a predetermined past in time than the predetermined time segment is stored. Obtain the past maximum value that is the maximum value stored in the past maximum value storage step in the time segment, and convert the current maximum value to a value associated with the past maximum value,
A speech encoding method characterized by the above. Entropy coding is performed on the quantized spectrum of the audio signal for each predetermined time segment, which is a result of performing deformation including normalization for each band, and the normalization value that is a value used for the normalization. A receiving step for receiving a code generated by
A decoding step of decoding the modified spectrum data and the normalization value for each of the predetermined time intervals from the code by a decoding method corresponding to the entropy encoding;
An inverse transformation step of restoring the quantized spectrum for each predetermined time segment using the standardized value decoded from the decoded spectral data,
A normalization value storing step for storing the normalization value;
A discrete spectrum inverse transform step of restoring the speech signal from the restored quantized spectrum;
Consisting of
The reverse deformation step includes
For each predetermined time segment, when the current standardization value, which is a standardization value decoded in the predetermined time segment, is stored by the standardization value storage step, A past standardization value that is a standardization value stored in the past standardization value storage step in a predetermined time segment in the past is acquired, and based on the current standardization value and the past standardization value To restore the quantized spectrum,
A speech decoding method characterized by the above. On the computer,
For a digital audio signal, for each predetermined time segment, a discrete spectrum conversion step for obtaining a value of a frequency component for each sub-compartment band having a predetermined bandwidth;
For each medium partition band composed of a predetermined number of consecutive sub-compartment bands set in advance according to auditory characteristics, for each predetermined time section, out of the frequency component values belonging to the medium-compartment band A maximum value search step for searching for a maximum value;
A maximum value storing step for storing the maximum value searched by the maximum value searching step;
An encoding step of quantizing and entropy-encoding and outputting information generated based on the maximum value and a frequency component normalized by division by the maximum value;
A program for executing
The maximum value search step includes:
For each predetermined time segment, when the current maximum value, which is the maximum value searched in the predetermined time segment, is stored by the maximum value storing step, a predetermined past in time than the predetermined time segment is stored. Obtain the past maximum value that is the maximum value stored in the past maximum value storage step in the time segment, and convert the current maximum value to a value associated with the past maximum value,
A program characterized by that. On the computer,
Entropy coding is performed on the quantized spectrum of the audio signal for each predetermined time segment, which is a result of performing deformation including normalization for each band, and the normalization value that is a value used for the normalization. A receiving step for receiving a code generated by
A decoding step of decoding the modified spectrum data and the normalization value for each of the predetermined time intervals from the code by a decoding method corresponding to the entropy encoding;
An inverse transformation step of restoring the quantized spectrum for each predetermined time segment using the standardized value decoded from the decoded spectral data,
A normalization value storing step for storing the normalization value;
A discrete spectrum inverse transform step of restoring the speech signal from the restored quantized spectrum;
A program for executing
The reverse deformation step includes
For each predetermined time segment, when the current standardization value, which is a standardization value decoded in the predetermined time segment, is stored by the standardization value storage step, A past standardization value that is a standardization value stored in the past standardization value storage step in a predetermined time segment in the past is acquired, and based on the current standardization value and the past standardization value To restore the quantized spectrum,
A program characterized by that.

Images