JP2001184090A - Signal encoding device and signal decoding device, and computer-readable recording medium with recorded signal encoding program and computer-readable recording medium with recorded signal decoding program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problem that it is difficult for a conventional encoding and decoding device for an acoustic signal, etc., to meet requirements of compressibility, quality, and an arithmetic quantity at high level. SOLUTION: High compressibility and high quality are secured with a high arithmetic quantity by determining the number of allocation bits allocated to each frequency component according to the spectrum envelope of the frequency spectrum of an object signal and encoding the frequency spectrum quantized corresponding to the number of allocation bits and the spectrum envelope.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for encoding and decoding a temporally or spatially changing target signal such as an audio signal including a sound signal and a musical sound signal, or a video signal, and the like. The present invention relates to a computer-readable recording medium on which a program for performing encoding and decoding is recorded.

[0002]

2. Description of the Related Art With the rapid spread of the Internet in recent years, communication transmission lines available to general consumers are diversifying, and some communication services capable of providing a bandwidth of about several Mbps are available. It is starting to appear. However, there are currently few users who use relatively high-speed communication services, and the bandwidth available to many users is at most about 128 kbps. This bandwidth is not enough to transmit so-called multimedia information such as music and video. The most basic encoding method for digitizing the multimedia information is a linear PCM (Pulse Code Modulation) method. The linear PCM method replaces a target signal with a code by quantizing the magnitude of a temporally or spatially discretized signal in equal steps.
The bit rate of the information encoded by the linear PCM method is sampling frequency × quantization bit number × channel number, and is, for example, an audio CD (sampling rate 44.1 kHz, quantization bit number 16, stereo).
The bit rate of the corresponding information is about 1411 kbps. The bandwidth available to many users is much smaller than this bit rate. With the above-mentioned bandwidth of about 128 kbps, it takes at least about 44 minutes to obtain a tone signal of about 4 minutes. Of course, under this condition, it is impossible to reproduce the tone signal in real time. However, since the multimedia information has a relatively large number of redundant portions, the information can be reversibly and irreversibly encoded (compressed) with high efficiency. When using a relatively low-speed communication transmission line as described above, it is practically essential to perform high-efficiency coding on a target signal. For example, a typical method for performing high-efficiency encoding of a tone signal is the international standard MPEG-1 (Moving Picture Expe
rts Group) MPEG-1 audio layer3 (so-called MP
It is compatible with the decoding method specified in 3) and TwinVQ (T) developed by NTT Human Interface Laboratories.
ransform-domain Weighted INterleave Vector Quantiz
ation). These control the deterioration of sound quality by utilizing the human auditory characteristics, and
The tone signal is compressed to about 20. The high-efficiency encoding method for these tone signals has become a main cause of the spread of music distribution services using the Internet and the like, and has also made it possible to commercialize portable reproduction terminals using semiconductor memories. The portable playback terminal that adopts the MP3 and TwinVQ,
With a memory capacity of about 64 Mbytes, recording for just over one hour is possible even with sound quality equivalent to a CD, which is sufficiently practical. Performing high-efficiency coding also saves limited communication transmission path resources. For example, in the field of mobile radio communications represented by mobile phones,
In consideration of wireless bandwidth restrictions, ADPCM (Adaptive Differential
ial Pulse Code Modulation) method and various CELP (Code Ex
Audio signals have been encoded according to the cited linear prediction (cid Linear Prediction) method. When the audio signal is encoded by the linear PCM system, a bandwidth of at least 64 kbps is required. However, the bit rate of the audio signal encoded by the ADPCM system is 32 kbps, and the bit rate of the CELP system is 32 kbps. The bit rate of the encoded audio signal is approximately 4 kbps to 16 kbps.
When coding by the various CELP methods, the linear
Compared to the case of encoding by the PCM method, the number of bits is considerably reduced, and the necessary radio frequency band is saved. In addition, such an encoding method of voice communication is used not only in the field of the mobile radio communication but also in an IP (Internet Protocol).
l) VOIP (Voice Ove) that performs voice communication using a network
r IP), which is expected to provide a low-cost service for long-distance telephone communication using the Internet.

[0003]

In performing high-efficiency encoding processing on the audio signal including the tone signal and the audio signal, the compression rate, the quality, and the calculation amount are the main evaluation criteria. The compression ratio is, of course, the ratio of the bit rates before and after encoding, and the quality is a sensory evaluation when the encoded signal is decoded and reproduced. The operation amount is an operation amount required when encoding or decoding a signal, and is related to the complexity of processing. Although the various high-efficiency coding methods described above (and the decoding methods corresponding to them) satisfy the above criteria under an appropriate balance according to the application, it is necessary to obtain a certain sound quality and a high compression rate. However, the processing tends to be complicated and the amount of calculation tends to increase, and it is difficult to satisfy all of the three evaluation criteria at a high level. The problem is not only with time-varying signals like the acoustic signals,
The same occurs when high-efficiency encoding is performed on a spatially changing signal such as an image included in a still image or a moving image. In order to solve the problems in the conventional technology, the present invention improves a signal encoding device, a signal encoding program, a signal decoding device, and a signal decoding program, and reduces the number of signals while maintaining the quality. A signal encoding device capable of obtaining a high compression ratio with an operation amount, a computer-readable recording medium recording a signal encoding program, and an encoding generated using the signal encoding device and the signal encoding program An object of the present invention is to provide a signal decoding device capable of decoding a signal with a small amount of calculation and performing high-quality reproduction, and a computer-readable recording medium recording a signal decoding program.

[0004]

In order to achieve the above object, the invention according to claim 1 comprises a frequency spectrum calculating means for obtaining a frequency spectrum of a target signal which changes temporally or spatially, and the frequency spectrum calculating means. Spectrum envelope extracting means for extracting a spectrum envelope from the frequency spectrum of the target signal calculated by the means; and, for each frequency component of the frequency spectrum of the target signal, according to the spectrum envelope extracted by the spectrum envelope extracting means. Allocation bit number determination means for determining the allocation bit number to be allocated; frequency spectrum quantization means for quantizing a frequency spectrum of the target signal based on the allocation bit number determined by the allocation bit number determination means; The circumference of the target signal quantized by the spectrum quantization means And the frequency spectral information about the number spectrum, and is configured and spectrum envelope information about the spectral envelope as a signal encoding apparatus comprising; and a coded signal generating means for generating an encoded signal encoded. The coded signal generated by the signal coding apparatus according to claim 1 includes frequency spectrum information related to a frequency spectrum of a target signal that changes temporally or spatially, and spectrum envelope information related to a spectrum envelope of the target signal. Is included. The frequency spectrum information is obtained by quantizing and encoding the frequency spectrum of the target signal based on the number of allocated bits allocated according to the spectrum envelope. It is much less than when encoding using the PCM method, and is equal to or less than high-efficiency encoding methods such as MP3 and TwinVQ that use the spectral envelope. Further, since the frequency spectrum is encoded almost as it is, deterioration of the quality of the target signal can be suppressed. Moreover, the conversion operation from the time domain to the frequency domain only needs to be performed once for encoding, and the amount of operation is small, and encoding can be performed in a short time.

According to a second aspect of the present invention, a coded signal obtained by coding frequency spectrum information on a frequency spectrum of a target signal which changes temporally or spatially and spectrum envelope information on a spectrum envelope of the target signal is encoded. The spectrum envelope restoring means for acquiring the spectrum envelope information and restoring the spectrum envelope for each frequency component of the frequency spectrum of the target signal based on the acquired spectrum envelope information; and the spectrum envelope restoring means. From the spectral envelope,
An allocation bit number obtaining unit that obtains an allocation bit number allocated to each frequency component of the frequency spectrum of the target signal; and the frequency spectrum information that obtains the frequency spectrum information from the coded signal, which is obtained by the allocation bit number obtaining unit. Frequency spectrum decoding means for decoding a frequency spectrum of the target signal from the acquired frequency spectrum information based on the allocated bit number, and a frequency spectrum of the target signal decoded by the frequency spectrum decoding means And a signal restoring means for restoring the target signal based on the signal decoding. According to the second aspect of the present invention, the coded signal generated by the signal encoding apparatus according to the first aspect is decoded to restore a temporally or spatially changing target signal. A suitable signal decoding device can be provided. 3. The signal decoding apparatus according to claim 2, wherein the main operations performed for decoding are a one-time inverse transform process for transforming from the frequency domain to the time domain and a comparison required for obtaining the number of allocated bits. Since only simple calculations are performed, decoding of the target signal can be performed in a short time, and real-time reproduction can be easily performed.

According to a third aspect of the present invention, there is provided a computer which obtains a frequency spectrum of an object signal which changes spatially or temporally, extracts a spectrum envelope from the frequency spectrum of the object signal, and A procedure for determining the number of bits to be allocated to each frequency component of the frequency spectrum of the target signal, a procedure for quantizing the frequency spectrum of the target signal based on the number of allocated bits, and a frequency spectrum of the target signal. A computer-readable recording medium on which a signal encoding program for executing a procedure for generating an encoded signal obtained by encoding frequency spectrum information and spectrum envelope information defining the spectrum envelope is recorded. A computer-readable recording medium according to a third aspect provides a signal encoding program suitable for implementing the signal encoding apparatus according to the first aspect by a computer. A coded signal generated by a computer in accordance with the signal coding program according to claim 3 includes, as in the signal coding apparatus according to claim 1, a target signal that changes temporally or spatially. Frequency spectrum information on a frequency spectrum and spectrum envelope information on a spectrum envelope of the target signal are included. The frequency spectrum information is obtained by quantizing and encoding the frequency spectrum of the target signal based on the number of bits allocated according to the spectrum envelope. The amount of information required for transmission and recording is , The MP3 or Tw using spectral envelope is much less than when the target signal is coded by the linear PCM method.
Equal to or less than a high efficiency coding method such as inVQ. Further, since the frequency spectrum is encoded almost as it is, deterioration of the target signal can be suppressed. In addition, the calculation such as the FFT associated with the spectrum analysis need only be performed once in the encoding, and the amount of computation in the computer is small, and the encoding can be performed in a short time.

According to a fourth aspect of the present invention, there is provided a computer which encodes frequency spectrum information relating to a frequency spectrum of a target signal which changes temporally or spatially and spectrum envelope information relating to a spectrum envelope of the target signal. Acquiring the spectrum envelope information from the coded signal and restoring the spectrum envelope for each frequency component of the frequency spectrum of the target signal based on the spectrum envelope information. Obtaining the number of allocated bits allocated to the frequency component, obtaining the frequency spectrum information from the encoded signal, and decoding the frequency spectrum of the target signal from the frequency spectrum information based on the allocated number of bits. Procedure, frequency spectrum of the target signal A computer-readable recording medium recording a signal decoding program for executing the steps to restore the target signal based on. According to the computer-readable recording medium of the fourth aspect, the encoded signal generated by the signal encoding device of the first aspect or the signal encoding program of the third aspect is used. A signal decoding program suitable for decoding a coded signal generated by a computer and restoring a temporally or spatially changing target signal can be provided. Also in the signal decoding program according to the fourth aspect, since the inverse transform processing for transforming from the frequency domain to the time domain only needs to be performed once for decoding, the total amount of computation performed by the computer is small. The decoding of the target signal can be performed in a short time, and the real-time reproduction can be performed using a relatively low-speed computer.

[0008]

Embodiments of the present invention will be described below with reference to the accompanying drawings to provide an understanding of the present invention.
The following embodiments are specific examples of the present invention and do not limit the technical scope of the present invention. As shown in FIG. 1, a coding apparatus 1 according to an embodiment of the present invention provides a frequency for obtaining a frequency spectrum of an audio signal (an example of a target signal) including a time-varying audio signal, a musical sound signal, or both. A spectrum calculation unit 103, a spectrum envelope extraction unit 104 for extracting a spectrum envelope from the frequency spectrum of the acoustic signal calculated by the frequency spectrum calculation unit 103, and a spectrum envelope extracted by the spectrum envelope extraction unit 104. Accordingly, based on the allocated bit number determined by the allocated bit number determining unit 105, the allocated bit number determining unit 105 determines the number of allocated bits to be assigned to each frequency component of the frequency spectrum of the audio signal. A frequency spectrum quantization unit 106 for quantizing the frequency spectrum of the Comprising a frequency spectrum information about the frequency spectrum of the acoustic signal which has been quantized by spectrum quantization unit 106, and a coded signal generating unit 107 and spectral information to generate an encoded signal encoded for said spectral envelope. Further, as shown in FIG. 2, the signal decoding device 2 according to the embodiment of the present invention comprises frequency spectrum information relating to the frequency spectrum of the acoustic signal which changes with time and spectrum envelope information relating to the spectral envelope of the bi-resonant signal. A spectrum envelope restoring unit 203 for acquiring the spectrum envelope information from an encoded signal obtained by encoding the above, and restoring the spectrum envelope for each frequency component of the frequency spectrum of the audio signal based on the acquired spectrum envelope information; An allocation bit number obtaining section 204 for obtaining the number of allocated bits allocated to each frequency component of the frequency spectrum of the audio signal from the spectrum envelope restored by the spectrum envelope restoring section 203; The frequency spectrum information is acquired, and the allocated bit number acquiring section 20 is acquired. Based on the number of allocated bits obtained by,
A frequency spectrum decoding unit 205 for decoding a frequency spectrum of the audio signal from the acquired frequency spectrum information, and a frequency spectrum decoding unit 205
And a signal restoring unit 206 for restoring the audio signal based on the frequency spectrum of the audio signal decoded according to.

[0009] A signal encoding program according to an embodiment of the present invention is provided to a computer as shown in FIG.
Step S10 of obtaining a frequency spectrum of the acoustic signal
1, a step S102 of extracting a spectrum envelope from a frequency spectrum of the audio signal, a step S103 of determining an allocated bit number to be assigned to each frequency component of the frequency spectrum of the audio signal in accordance with the spectrum envelope, and based on the allocated bit number. Then, a step S104 of quantizing the frequency spectrum of the audio signal and a step S104 of generating an encoded signal obtained by encoding frequency spectrum information on the frequency spectrum of the audio signal and spectrum envelope information defining the spectrum envelope are executed. For example, a CD-ROM on which the signal encoding program is recorded is a specific example of a computer-readable recording medium on which the signal encoding program according to the embodiment of the present invention is recorded. In addition, the signal decoding program according to the embodiment of the present invention stores in a computer, as shown in FIG. 4, frequency spectrum information relating to a time-varying frequency spectrum of the acoustic signal and a spectrum relating to a spectrum envelope of the acoustic signal. Acquiring the spectrum envelope information from the encoded signal obtained by encoding the envelope information, and restoring the spectrum envelope for each frequency component of the frequency spectrum of the audio signal based on the acquired spectrum envelope information, step S201, Step S202 of acquiring the number of allocated bits assigned to each frequency component of the frequency spectrum of the audio signal from the spectrum envelope, acquiring the frequency spectrum information from the encoded signal, and obtaining the frequency based on the number of allocated bits. From the spectrum information, the frequency spectrum of the acoustic signal And a step S204 of restoring the audio signal based on the frequency spectrum of the audio signal. For example, a CD-ROM storing this signal decoding program is It is a specific example of a computer-readable recording medium recording a signal decoding program according to an embodiment of the present invention.

The signal encoding device 1 and the signal decoding device 2 are each embodied as, for example, a computer capable of executing the signal encoding program and the signal decoding program. As shown in FIG. 5, the computer includes an input device 501 such as a keyboard and a mouse, an output device 502 such as a display, and a CD-R.
OM drive 503, hard disk drive 504,
Sound input / output board 50 to which RAM 505, arithmetic unit 506, speaker 5071 and microphone 5072 are connected.
7, a standard configuration including a communication device 508 such as a LAN board and a modem. In the computer, the CD-ROM in which one or both of the signal encoding program and the signal decoding program are recorded in a compressed state is stored in the CD-ROM drive 50.
3, when the installation program in the CD-ROM is executed in accordance with an instruction given by the user using the input device 501, the signal encoding program and the signal decoding read from the CD-ROM are executed. Either or both of the computerized programs are developed in an executable state on a storage medium such as the hard disk drive 504. When a user issues an instruction to execute one or both of the signal encoding program and the signal decoding program from the user via the input device 501, the signal expanded on the hard disk drive 504 or the like. A part or the whole of one or both of the encoding program and the signal decoding program is stored in the RAM 50.
5 and the hard disk drive 504, and the arithmetic unit 506 executes one or both of the signal encoding program and the signal decoding program.

Hereinafter, the signal encoding device 1, the signal decoding device 2, the signal decoding device 2, and the signal decoding device 2 will be described by taking as an example the case where the signal encoding device 1 and the signal decoding device 2 are realized by the same computer. Details of the encoding program and the signal decoding program will be described. When the signal encoding program is executed and the computer operates as the signal encoding device 1, various settings such as designation of the audio signal to be encoded and designation of a bit rate after encoding can be performed first. A dialog is displayed on the output device 502. The designation of the bit rate after the encoding is used to determine the upper limit of the number of bits that can be used in the encoding process by the signal encoding device 1. The audio signal to be encoded may be input from a microphone at that time, or may be already digitized in some encoding format, and may be encoded in the hard disk drive 504.
Such as those stored above. When the user selects to input the acoustic signal from the microphone, the user outputs information such as the number of quantization bits at the time of capture, the sampling rate, and the distinction between stereo and monaural to the output device 502. A dialog for specifying is displayed. When the user gives necessary information at the time of capturing by using the input device 501 and further gives an instruction for capturing, the acoustic signal input from the microphone is
The encoding process is started. In the signal encoding device 1, the audio signal input from a microphone or the like via an input terminal 101 is supplied to the frequency spectrum calculation unit 103 via an A / D converter 102. In this example in which the signal encoding device 1 is realized by the computer, as a microphone connected to the input terminal 101,
For example, a microphone 5072 connected to the sound input / output board 507 is used, and an A / D converter (not shown) mounted on the sound input / output board 507 is used as the A / D converter 102. The signal output from the A / D converter mounted on the sound input / output board 507 is actually temporarily stored on the hard disk drive 504 or the like. The digitized audio signal output from the A / D converter 102 and temporarily stored on the hard disk drive 504 or the like is the linear PCM
It corresponds to the one encoded by the method. If the capture condition is equivalent to CD, the bit rate is
It is about 1411 bps. The digitized audio signal is read by the arithmetic unit 506 operating as the frequency spectrum arithmetic unit 103 according to the signal encoding program stored in the RAM 505 or the like.

[0012] The frequency spectrum calculator 103 performs FFT (Fast Fourier Tr) processing on the input acoustic signal.
ansform) operation to obtain a frequency spectrum of the acoustic signal. The sound signal supplied to the frequency spectrum calculation unit 103 has been divided in time series in advance. At the time of division, processing is performed so that a part of each frame overlaps. The width of the frame of the audio signal divided in time series is 100 ms, which is longer than other encoding methods that often employ a frame width of about 20 ms to 60 ms.
The degree is appropriate. This is because the encoding method according to the present embodiment determines the number of allocated bits of each frequency component based on the spectrum envelope extracted from the frequency spectrum, and quantizes each frequency component according to the allocated number of bits. In order to reduce the bit rate, it is necessary to satisfactorily extract the spectrum envelope. The frame width does not need to be constant for each frame, and may be dynamically changed according to the signal of the target frame. Each frame of the audio signal is multiplied by a window function such as a Hamming window before the FFT operation, if necessary. The frequency spectrum calculator 1
In step 03, the FFT operation is performed for each frame to determine a power component and a phase component of the acoustic signal in the frequency domain. When the sampling rate for the acoustic signal is 44.1 kHz, the number of sample points in the FFT operation is appropriately set at 4096. Information such as the setting for the frame width, the presence or absence of the window function, and the type of the window function may be changed by the user as appropriate by displaying a dialog on the output device 502 or the like.
The above-described processing performed by the frequency spectrum calculation unit 103 is processing corresponding to the step S101 that the signal encoding program causes the computer to execute.
The power component and the phase component of the frequency spectrum of the acoustic signal determined by the frequency spectrum calculation unit 103 are temporarily stored in a primary cache or the like such as the RAM 505 or the calculation device 506. The arithmetic unit 5
06 operates as the spectrum envelope extraction unit 104 in accordance with the signal encoding program. The spectrum envelope extraction unit 104 extracts a spectrum envelope from a power component of a frequency spectrum of the audio signal. LP
Although it is possible to use C (Linear prediction code) or the like, in the case of the signal encoding device (and the signal encoding program) according to the present embodiment, the amount of information necessary to determine the spectrum envelope is as small as possible. It is also important that the area occupied by the spectrum envelope in the two-dimensional frequency-intensity space is as small as possible.

FIG. 6 shows a specific calculation procedure of spectrum envelope extraction more suitable for the signal encoding device (and signal encoding program) according to the present embodiment. As shown in FIG. 6, first, the frequency domain of the power component is divided into a plurality of frequency bands, and the maximum value of the power component in each frequency band is extracted (S1001). Since the number of divisions of the power component in the frequency domain is related to the compression ratio, it is determined according to a bit rate specified in advance by a user or the like.
Next, a temporary spectrum envelope connecting the maximum values of the power components in the respective frequency bands is set (S100).
2). Next, whether or not there is a frequency component significantly larger than the temporary spectrum envelope among the power components is searched for in all frequency regions (S1003). Whether or not the magnitude of the frequency component is significantly greater than the temporary spectrum envelope may be determined based on, for example, whether or not the difference is greater than a preset threshold. Next, when there is a frequency component determined to be significantly larger than the temporary spectrum envelope, the frequency component and the maximum value of the power component in each of the frequency bands are remarkably compared to the temporary spectrum envelope. If there is no frequency component determined to be large, only the maximum value of the power component in each frequency band is set as the peak point of the spectrum envelope to be extracted (S1004).
Note that the size of each frequency band for obtaining the spectral envelope does not need to be the same, and may be different depending on the frequency. It is preferable that the number of the peak points is large in the low and middle frequency regions in terms of human hearing characteristics. Even if the sound in the high frequency range is not reproduced very precisely, there is little problem in terms of hearing. After the peak points of the spectrum envelope are set, the valley portion of the frequency spectrum between the two peak points is expressed in the manner of hanging the thread with the adjacent peak points as fulcrums. The function used to represent the valley portion is hereinafter referred to as a valley function. As a specific example of the valley function, a sine wave function L = A ×
sin (ω × n) or the square root function of a sine wave L = A × √
(Sin (ω × n)), power function of sine wave L = A × s
^inc (ω × n), parabolic function L = A × n ² + b, higher-order function, suspension curve L = A × cosh (ω × n), and the like. Here, n is the number of samples when counting each sample (on the frequency axis) between both peak points from one peak point, and L is the size of each sample between both peak points. For example, when the sine wave function is used as the valley function, the spectrum envelope between both peak points is represented by the following equation (1). L = a.times.n-A.times.sin (.omega..times.n) + C (1) As shown in FIG. 7, N in the above equation (1) is a value from one peak point P1 to the other peak point P2 when counting is started. The number of samples, where 0 ≦ n ≦ N and ω = π / N.
A is the slope of the temporary spectrum envelope L 'between the peak points P1 and P2, and C is the magnitude of one peak point P1. When the size of the peak point P is considerably larger than the number of samples between the two peak points P, that is, when the valley of the frequency spectrum is deeper, the sine wave function is It is better to use a power function of. The spectral envelope between both peak points P1 and P2 is determined by changing the coefficient A of the valley function included in the above equation (1). For example, among the samples located between both peak points P1 and P2, the size of any one of the samples E1 smaller than the temporary spectrum envelope L ′ and the value L of the spectrum envelope
Then, the spectrum envelope represented by the above equation (1) using the coefficient A at the time when is matched is determined as a part of the spectrum envelope to be extracted (spectrum envelope between both peak points) (S1005). By repeating the procedure S1005 for all the peak points (S100
6) For example, a spectrum envelope for each frequency component as shown in FIG. 8 is obtained. As shown in FIG. 8, the spectrum envelope extracted according to the above-described procedure does not become smaller than the magnitude of each frequency component by a predetermined value or more, and the area occupied by the spectrum envelope in the frequency-intensity two-dimensional space is small.

Incidentally, in order to obtain the above-mentioned preferable spectral envelope, it is preferable that the interval between the peak points is wide and the valley portion is deep. The reason that the frame width of about 100 ms is suitable as described above is related to this point. 20
Even when a frame width of about ms is adopted, an information amount equivalent to that of about 100 ms is required to obtain a suitable spectral envelope. However, since information on the spectrum envelope for decoding is required for each frame, the information amount of the spectrum envelope increases as the frame width decreases. On the other hand, even if the frame width is increased, the number of the peak points does not increase so much, and the depth of the valleys increases. This is because, for example, in the case of voiced speech, the fundamental frequency and its harmonics appear as peaks, but when the frame width is increased, even if the interval between the peaks is the same when viewed in frequency, it increases when viewed in order. . Also, in the case of musical sounds, the frequency components do not fold at random, but have regularity in 12 steps, and the sound of musical instruments often consists of the fundamental frequency and its harmonics. Is the same as Another reason is that the signal steadiness increases with a frame width of about 100 ms. The above-described processing performed by the spectrum envelope extraction unit 104 is the same as the procedure S in which the signal encoding program is executed by the computer.
This is processing corresponding to 102.

The spectrum envelope extracted by the spectrum envelope extraction unit 104 is temporarily stored in the RAM 505, the primary cache of the arithmetic unit 506, or the like. Next, the arithmetic unit 506 operates as the allocated bit number determining unit 105 according to the signal encoding program. The allocation bit number determination unit 105 allocates a bit number according to each value of the spectrum envelope to each frequency component. Basically, more bits are allocated to a frequency component having a larger value of the spectrum envelope. For example, each value of the spectrum envelope is logarithmized. Although logarithmization is not essential, it is most preferable in terms of the characteristics of bits. Next, the spectrum envelope is normalized so that the lower limit value provided in advance becomes zero. Next, the sum of the logarithmic values of the spectrum envelope is obtained for each of the plurality of divided frequency bands. Next, a coefficient (= total bit number / sum of each value of spectrum envelope) such that the sum of each value of the spectrum envelope in each frequency band is equal to the total number of bits of the frequency band is determined. The value of the spectrum envelope is multiplied, and the number of bits allocated to each frequency component is calculated according to the value. Distributing the total total number of bits in a predetermined all frequency range for each of a plurality of frequency bands means that when a large component is concentrated in a certain frequency range, the number of bits allocated to another frequency range becomes extremely small. This is to prevent that. This is because components in a region near a frequency region where large components are concentrated are difficult for a human to perceive due to a mask effect, but components in a distant region are relatively easy to perceive. The ratio of the bit number distribution may be determined in advance, or may be determined according to the frequency spectrum of the audio signal to be encoded.
Basically, allocating more bits to the frequency band in the low-middle region matches human hearing characteristics. When the number of bits is determined from each value of the spectrum envelope, the number of bits below the decimal point is rounded down or rounded off. Further, the number of bits may be varied depending on the number of bits, such as rounding to a small number of bits and truncation to a large number of bits. Also, if you allocate more bits than necessary,
Since the sound quality is not improved, in order to prevent bits from being wasted, for example, an upper limit value of the number of bits allocated to each frequency band is set in advance, and the number of bits corresponding to each value of the spectrum envelope sets the upper limit value. If it exceeds, the upper limit is assigned instead of the number of bits corresponding to the frequency component. Further, if necessary, the total number of bits allocated to the frequency band and the total number of bits actually allocated to the number of bits surplus due to rounding down or the like become equal or fall within a predetermined range. , Or the coefficient is adjusted. The above-described processing performed by the allocation bit number determination unit 105 is the same as the procedure S1 that causes the computer to execute the signal encoding program.
03.

The number of allocated bits determined by the allocated bit number determining unit 105 is temporarily stored in the RAM 505, the primary cache of the arithmetic unit 506, or the like.
Next, the arithmetic unit 506 operates as the frequency spectrum quantization unit 106 according to the signal encoding program. The frequency spectrum quantization unit 106 includes a power component quantization unit 1061 for quantizing the power component of the frequency spectrum, a phase component quantization unit 1062 for quantizing the phase component, and a frequency component in which the power component becomes 0. It comprises a zero component elimination unit 1063 for eliminating the phase component. The power component quantization unit 1061 is supplied with the power component from the frequency spectrum calculation unit 103, and according to the number of allocated bits supplied from the allocated bit number determination unit 105, each frequency component of the power component is It is quantized linearly or non-linearly. When performing quantization linearly, the size of each frequency component after quantization is the maximum value that can be expressed by the number of allocated bits in the ratio of the size of the frequency component to the spectrum envelope of the frequency spectrum component. Expressed by multiplied value. The decimal point generated at the time of the quantization operation is rounded such as rounding. However, when the number of allocated bits is small, quantization distortion may increase. For this reason, the quantization distortion is reduced by, for example, differentiating the rounding method according to the following equations (2a) and (2b) depending on whether the number of allocated bits is 1 bit or 2 bits or more. More preferably, _{R = I nt (Sp / En} + 0.3) (2a) R = I nt (Sp × (2 Ba -1) /En+0.5) (2b) where, R represents the size of the frequency components after quantization , Sp
Is the magnitude of the frequency component, En is the spectrum envelope of the frequency component, Ba is the number of bits allocated to the frequency component,
_Int is an integer function. Further, the phase component is supplied to the phase component quantization unit 1062 from the frequency spectrum calculation unit 103. For example, like the power component quantization unit 1061, the allocated bit number determination unit 10
According to the number of allocated bits supplied from 5, each frequency component of the phase component is quantized. However, since the frequency component in which the power component becomes 0 is not restored, the phase component quantization unit 1062 is quantized in advance by the 0 component elimination unit 1063 according to the power component output from the power component quantization unit 1061. The frequency component is excluded from the target. Since the phase component has a great effect on the clarity of the restored tone (especially the striking sound, etc.), the frequency component which is assigned only a small number of bits, such as 1 bit or 2 bits, has an additional 1 bit or more. A few bits may be added to improve the reproducibility of the phase component. Further, in the phase component, since the phase difference between components in the vicinity of the pole is particularly important, it is preferable to quantize the difference from an adjacent sample rather than quantize an absolute value. However, when quantizing the difference from the adjacent sample, a deviation from an absolute value may occur due to a calculation error or the like. For this reason, when quantizing the difference from the adjacent sample, it is better to set the difference target not to the absolute value of the adjacent sample but to the cumulative value of the quantized value. Note that the above-described processing performed by the frequency spectrum quantization unit 106 is processing corresponding to step S104 that the signal encoding program causes the computer to execute.

The power component and the phase component quantized as described above are substantially the frequency spectrum itself, although the number of quantization bits differs for each frequency component. Although the number of quantization bits changes according to the spectral envelope, the spectral envelope extracted so as not to be smaller than each frequency component by a predetermined value or more results in human auditory characteristics such as a mask effect. Match,
There is very little information lost before and after quantization in terms of quality. The power component and the phase component quantized by the frequency spectrum quantization unit 106 are temporarily stored in the RAM 505, the primary cache of the arithmetic device 506, or the like. Next, the arithmetic unit 506 operates as the coded signal generation unit 107 according to the signal coding program. In the coded signal generation unit 107, the spectrum envelope extracted by the spectrum envelope extraction unit 104,
Based on the power component quantized by the power component quantization unit 1061 and the phase component quantized by the phase component quantization unit 1062, spectrum envelope information on the spectrum envelope, power component information on the power component, An encoded signal is generated by encoding the phase component information relating to the phase component. The spectrum envelope information is information necessary for determining the spectrum envelope. When the spectrum envelope is extracted by the above-described preferred procedure, the spectrum envelope information includes the size of the peak point, the frequency of the peak point, the type of the valley function, and parameters of the valley function. The above four parameters are required in a number substantially proportional to the number of the peak points. That is, the compression ratio can be changed by changing the number of the peak points. The size of each peak point is quantized separately from the quantization of the power component. For example, a predetermined number may be used as the number of quantization bits when quantizing the size of each peak point. The number of quantization bits for each peak point may also be determined for each of a plurality of frequency bands. When quantizing the magnitude of the peak point for each of a plurality of frequency bands, it is preferable to increase the number of bits allocated to the low and middle frequency regions, similarly to the quantization of the power component. In addition, when the frequency of the peak point is represented by the difference frequency from the adjacent peak point, the required number of bits can be reduced rather than the absolute frequency. The type of the valley function is necessary only when a plurality of valley functions are used, and can be omitted when a valley function is determined in advance. The valley function has one parameter or a plurality of parameters depending on the type of the valley function. When the sine wave function or its square root function is used, the parameter may be only one of the amplitudes A. This amplitude is 2-4
It is appropriate to express it by the bit quantization bit number. The power component information is basically the power component itself quantized as described above. However, the size of the peak point and the frequency of the peak point are included in the spectrum envelope information, and can be removed from the power component. Also, the phase component information is basically the phase component itself quantized as described above. It is needless to say that the power component and the phase component quantized as described above may be subjected to some encoding processing as the power component information and the phase component information. For example, when the phase component is quantized to a difference value instead of an absolute value, a small value appears more frequently. Therefore, if necessary, Huffman encoding is performed to reduce the data amount of the encoded signal. Can be made smaller. The above-described processing performed by the coded signal generation unit 107 is processing corresponding to step S105 that the signal coding program causes the computer to execute. The encoded signal output from the encoded signal generation unit 107 is stored, for example, on a hard disk drive 504 of the computer via an output terminal 108. In this way, the storage capacity required to store the coded signal coded by the signal coding apparatus is much smaller than when the acoustic signal is coded by the linear PCM method, and is smaller than that of the MP3 or TwinVQ. Equal or less. Also, since the frequency spectrum is encoded almost as it is, there is little deterioration in sound quality. In addition, when the frequency spectrum envelope information is determined, the FFT conversion needs to be performed only once, and the MP3 or Twin using the spectrum envelope information is used.
The amount of calculation can be significantly reduced as compared with VQ.

The device suitable for decoding the encoded signal generated by the signal encoding device 1 is the signal decoding device 2. When the signal decoding program is executed and the computer operates as the signal decoding device 2, first, for example, a dialog for designating the coded signal to be decoded is displayed on the output device 5.
02 is displayed. When the user specifies the coded signal to be decoded, decoding processing is started for the coded signal. In the signal decoding device 2, the coded signal input via the input terminal 201 is temporarily stored in a buffer 202. In this example in which the signal decoding device 2 is realized by the computer, the buffer 202 uses the hard disk drive 504, the RAM 505, or the like.
The encoded signal stored on the hard disk drive 504 or the like used as the buffer 202 is read by the arithmetic unit 506 operating as the spectrum envelope restoration unit 203 and the frequency spectrum decoding unit 205. The spectrum envelope restoration unit 203 acquires the spectrum envelope information from the coded signal, and restores the spectrum envelope for each frequency component of the frequency spectrum of the audio signal based on the acquired spectrum envelope information. As described above, the spectrum envelope information includes, for example, the size of the peak point, the frequency of the peak point, the type of the valley function, and the parameter of the valley function. If the number of quantization bits used to represent the size of the peak point is included in the spectral envelope information, the size of the peak point is restored using the number of quantization bits. In order to reduce the amount of data necessary for the spectrum envelope information as much as possible, it is desirable that the number of quantization bits is previously shared between the signal encoding device 1 and the signal decoding device 2. When the magnitude of the peak point and its frequency are determined, the spectrum envelope is restored if the type of the valley function and the parameters of the valley function (for example, coefficient A) are obtained from the spectrum envelope information. The spectrum envelope restoration unit 20
3 is a process corresponding to step S201 that the signal decoding program causes the computer to execute. The spectrum envelope restored by the spectrum envelope restoration unit 203 is supplied to an allocated bit number acquisition unit 204. The allocated bit number acquisition section 204
Obtains the number of bits allocated to each frequency component of the frequency spectrum of the audio signal from the spectrum envelope. The process of acquiring the number of allocated bits from the spectrum envelope is the same as the process performed by the allocated bit number determination unit 105 of the signal encoding device 1. The above-described processing performed by the allocated bit number acquiring unit 204 is as follows.
This is processing corresponding to step S202 that the signal decoding program causes the computer to execute. The number of allocated bits for each frequency component obtained by the allocated bit number obtaining unit 204 is supplied to the frequency spectrum decoding unit 205.

The frequency spectrum decoding unit 205 comprises:
Power component decoding section 205 for decoding the power component
1, a phase component decoding unit 205 for decoding the phase component
It has a 0 component elimination unit 2053 for eliminating 2,0 components. The power component decoding unit 2051 obtains power component information of the frequency spectrum information from the coded signal stored in the buffer 202, and obtains power component information for each frequency component obtained by the allocated bit number obtaining unit 204. The power component is decoded according to the number of allocated bits. If the power component information is converted into a bit stream, the power component is decoded by sequentially extracting the assigned bits for each frequency component. In addition, the phase component decoding unit 2052
02 of the frequency spectrum information is obtained from the coded signal stored in No. 02, and the phase component is obtained in accordance with the number of bits allocated to each frequency component obtained by the allocated bit number obtaining unit 204. Decrypt.
However, since the frequency component whose power component has a magnitude of 0 is not restored, the phase component decoding unit 2052 performs decoding based on the power component decoded by the power component decoding unit 2051 based on the power component. , The zero-component elimination unit 2053 eliminates the frequency component whose power component is zero. When the power component and the phase component are quantized using the above equations (2a) and (2b) and the rounding processing is varied according to the number of allocated bits, the following equations (3a),
The decoding process is performed according to (3b). Sp = 0.85 × R × Ep (3a) Sp = R × Ep / (2 ^Ba −1) (3b) The above-described processing performed by the frequency spectrum decoding unit 205 is performed by the signal decoding program. Is a process corresponding to the procedure S203 to be executed by the computer. When the power component and the phase component are decoded by the power component decoding unit 2051 and the phase component decoding unit 2052, the arithmetic unit 506 next outputs
It operates as the signal restoration unit 206. The signal restoration unit 20
6, the power component decoded by the power component decoding unit 2051 and the phase component decoding unit 205
A real number component and an imaginary number component are derived from the phase component decoded by step 2, and an IFFT (Inverse FFT) operation is performed to convert the component into a time domain signal, thereby restoring the acoustic signal. In the signal encoding device 1, when a window function is applied to each frame, it is necessary to multiply the value after the IFFT operation by the inverse function of the corresponding window function to restore. The information on the window function is shared by the signal encoding device 1 and the signal decoding device 2 or included in the encoded signal. By overlapping the window-applied portions of the joints of the respective frames, noise caused by discontinuity of the joints can be removed. In addition, it is also effective to regenerate the joined portion by excluding a portion near the portion subjected to the window function and multiplying both ends of the remaining portion by the window function. In such a case, it is necessary to previously secure a portion to be excluded when dividing the frame. The above-described processing performed by the signal restoring unit 205 is processing corresponding to step S204 that the signal decoding program causes the computer to execute.

When an instruction to execute reproduction is given to the acoustic signal restored by the signal restoring unit 206, the acoustic signal is converted into an analog signal by a D / A converter 207 and then output to an output terminal 208. Output from a speaker or the like via the As described above, the signal decoding device according to the embodiment of the present invention is suitable for decoding the encoded signal generated by the signal encoding device and restoring a temporally changing target signal. . Also in the signal decoding device, the inverse transform process for transforming from the frequency domain to the time domain only needs to be performed once in performing the decoding, so that the entire operation amount is small and the target signal can be converted in a short time. Decoding can be performed, and real-time reproduction can be easily performed. Also, most of the processing of the signal decoding device (and the signal decoding program) is the same as that of the signal encoding device (and the signal coding program).
Or it is a contrasting process, and it is relatively easy to configure both by one device and one program.
Here, FIG. 9 shows a comparison between the power components of the source signal, the signal coded / decoded by the MP3, and the signal coded / decoded by the method according to the present embodiment. Show. The bit rates of the MP3 and the method according to the present embodiment are both 128 kbps. As shown in FIG. 9, the frequency spectrum of the method according to the present embodiment matches well from the low-frequency region to the high-frequency region.
The method based on MP3 is significantly different from that based on the source signal particularly in the low frequency region and the high frequency region. That is, in the encoding / decoding method according to the present embodiment, high quality can be ensured while maintaining the same compression ratio. Also, the amount of calculation at that time is small.

[0021]

In the above embodiment, an example has been described in which the signal encoding device 1 and the signal decoding device 2 are realized by a single computer. However, the present invention is not limited to this. The decoding device 1 and the signal decoding device 2 may be realized by separate computers. An encoded signal generated by a computer that implements the signal encoding device 1 is transmitted to a computer that implements the signal decoding device 2 via the Internet or the like using the communication device provided in each computer. What should I do? A computer that realizes the signal encoding device 1 stores, for example, the encoded signal in a packet corresponding to a protocol corresponding to the communication, and the signal encoding device 2 depackets the packet and then performs a decoding process. I do. When real-time reproduction is performed via a network whose transmission time is not guaranteed, the spectrum envelope information and the frequency spectrum information (the power component information and the phase component information) included in the coded signal are stored in the same packet. Is preferably performed. However, even if each information is stored in a separate packet,
For example, only the spectrum envelope information and the power component information, only the spectrum envelope information and the phase component information,
Alternatively, if a packet storing only the frequency spectrum information has arrived, it is possible to reproduce on the receiving side although the sound quality is deteriorated. In addition, the signal encoding device 1 and the signal decoding device 2 may be realized not by a general-purpose computer but by dedicated hardware including a DSP or the like. Also in this case, the same device may include the signal encoding device 1 and the signal decoding device 2 or separate devices include the signal decoding device 1 and the signal decoding device 2, respectively. May be. Further, in the above-described embodiment, an example in which the audio signal is encoded and decoded has been described. However, the present invention is not limited to this, and other temporally or spatially changing other images such as a still image and a moving image may be used. The present invention can be applied to a target signal. For example, if the target signal is a spatially changing still image, the conversion such as FFT is performed by 2
When performed in a dimensional manner, a spectral distribution in the spatial frequency domain is obtained. According to the spectral intensity of each spatial frequency component,
The number of bits allocated to the spatial frequency component is changed. As described above, since the frame width affects the compression ratio, a process of dynamically optimizing the frame width with respect to the target signal may be added. For example, the following techniques (1) to (4) are available as optimization techniques. (1) Change the frame width and adopt the frame width that minimizes the area of the spectral envelope. (2) By changing the frame width and compressing them at the same compression ratio, a frame width that minimizes the difference between the compressed spectrum and the uncompressed spectrum is adopted. (3) The frame width is changed, and compression is performed at a compression ratio that can obtain the same level of spectral accuracy, and a frame width that minimizes the compression ratio is adopted. (4) Any one of the above (1), (2), and (3), or a combination thereof, is used to determine in advance a frame width having the best compression ratio or spectral accuracy by a statistical method.
Use that frame width. At this time, voice (separated by gender),
It is also effective to calculate the frame width for each type of musical sound (pops, rock, enka, etc.). Further, in the above embodiment, FFT and IFFT are used in relation to the frequency spectrum.
Other conversion processes such as a cosine transform (screte Cosine Transform) and an inverse conversion process may be used.

[0022]

As described above, according to the signal encoding apparatus of the first aspect, a high compression rate and high quality can be secured with a small amount of calculation. Further, the above-mentioned claim 2
According to the invention described in (1), it is suitable for decoding the encoded signal generated by the signal encoding device according to claim 1 and restoring a temporally or spatially changing target signal. It is possible to provide a signal decoding device that requires a small amount of calculation for decoding. Further, according to the computer-readable recording medium of the third aspect, it is possible to provide a signal encoding program suitable for realizing the signal encoding device of the first aspect by a computer. According to a fourth aspect of the present invention, there is provided a computer-readable recording medium, wherein the encoded signal generated by the signal encoding apparatus according to the first aspect or the signal encoding method according to the third aspect is used. It is possible to provide a signal decoding program suitable for decoding a coded signal generated by a computer according to the program and restoring a temporally or spatially changing target signal.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of a signal encoding device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a schematic configuration of a signal decoding device according to the embodiment of the present invention.

FIG. 3 is a flowchart for explaining a signal encoding program according to the embodiment of the present invention.

FIG. 4 is a flowchart for explaining a signal decoding program according to the embodiment of the present invention.

FIG. 5 is a diagram showing a configuration example of a computer that realizes the signal encoding device and the signal decoding device.

FIG. 6 is a flowchart for explaining a spectrum envelope extraction process.

FIG. 7 is a view for explaining a method of determining a valley function in the spectral envelope extraction processing.

FIG. 8 is a diagram showing an example of an extracted spectrum envelope.

FIG. 9 is a diagram showing a source signal, the present scheme, and other encoding schemes by comparing frequency spectra.

[Explanation of symbols]

 Reference numeral 103: frequency spectrum calculation unit 104: spectrum envelope extraction unit 105: allocation bit number determination unit 106: frequency spectrum quantization unit 107: coded signal generation unit 203: spectrum envelope restoration unit 204: allocation bit number acquisition unit 205: frequency spectrum Decoding unit 206: signal restoration unit

 ──────────────────────────────────────────────────続 き Continuation of the front page (54) [Title of the Invention] A signal encoding device, a signal decoding device, a computer-readable recording medium on which a signal encoding program is recorded, and a signal decoding program are recorded. Computer readable recording medium

Claims (4)

[Claims] 1. A frequency spectrum calculating means for calculating a frequency spectrum of a target signal which changes temporally or spatially, and a spectrum envelope extracting means for extracting a spectrum envelope from the frequency spectrum of the target signal calculated by the frequency spectrum calculating means. Means, allocated bit number determining means for determining the number of allocated bits to be allocated to each frequency component of the frequency spectrum of the target signal according to the spectrum envelope extracted by the spectrum envelope extracting means, and the allocated bit number determining means. Frequency spectrum quantization means for quantizing the frequency spectrum of the target signal based on the determined number of allocated bits; frequency spectrum information on the frequency spectrum of the target signal quantized by the frequency spectrum quantization means; , The spectral envelope And a coded signal generating means for generating a coded signal obtained by coding spectrum envelope information related to the signal. 2. Obtaining the spectrum envelope information from an encoded signal obtained by encoding frequency spectrum information on a frequency spectrum of a target signal that changes temporally or spatially and spectrum envelope information on a spectrum envelope of the target signal, A spectrum envelope restoration unit for restoring the spectrum envelope for each frequency component of the frequency spectrum of the target signal based on the acquired spectrum envelope information; and the spectrum envelope restored by the spectrum envelope restoration unit. Allocation bit number obtaining means for obtaining the number of allocated bits allocated to each frequency component of the frequency spectrum of the frequency spectrum information, and obtaining the frequency spectrum information from the encoded signal, and obtaining the allocated bit number obtained by the allocated bit number obtaining means. Based on the number,
Frequency spectrum decoding means for decoding the frequency spectrum of the target signal from the acquired frequency spectrum information, based on the frequency spectrum of the target signal decoded by the frequency spectrum decoding means,
A signal decoding device comprising: a signal restoring unit that restores the target signal. 3. A computer for obtaining a frequency spectrum of a spatially or temporally changing target signal, extracting a spectrum envelope from the frequency spectrum of the target signal, and responding to the spectrum envelope according to the spectrum envelope. A procedure for determining the number of bits to be allocated to each frequency component of the frequency spectrum, a procedure for quantizing the frequency spectrum of the target signal based on the number of allocated bits, frequency spectrum information on the frequency spectrum of the target signal, and the spectrum envelope A computer-readable recording medium on which a signal encoding program for executing a procedure for generating an encoded signal obtained by encoding spectral envelope information defining the following is stored. 4. A computer according to claim 1, wherein said spectral envelope information is obtained from an encoded signal obtained by encoding frequency spectrum information relating to a frequency spectrum of a target signal which changes temporally or spatially and spectral envelope information relating to a spectrum envelope of said target signal. Acquiring and restoring the spectrum envelope for each frequency component of the frequency spectrum of the target signal based on the spectrum envelope information; and assigning bits assigned to each frequency component of the frequency spectrum of the target signal from the spectrum envelope. Obtaining the frequency spectrum information from the coded signal, decoding the frequency spectrum of the target signal from the frequency spectrum information based on the allocated bit number, and obtaining the frequency spectrum of the target signal. The target signal is restored based on A computer-readable recording medium on which a signal decoding program for executing the original procedure is recorded.