JPH10124093A - Method and device for speech compressive encoding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To decrease the arithmetic quantity for encoding (quantization) and to reduce the memory capacity for storing noise code vectors by encoding (quantizing) a secondary error signal itself for the encoding (quantization) of the secondary error signal which was carried out by a noise code vector search in a process of CELP system encoding. SOLUTION: When noise source information is extracted and encoded, a secondary error signal constitution part 301 constitutes the secondary error signal of a subframe signal, extracted spectrum envelope information, pitch information, gain information and the residue signal up to the least subframe, a DCT transformation part 302 performs the discrete cosine transformation of the extracted secondary error signal, and a coefficient conversion part 303 converts a DCT transformed coefficient sequence to specific bit length, thereby extracting and encoding noise source information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice compression encoding method and apparatus applied to an answering machine, a voice response system, a voice mail, and the like. The present invention relates to an audio compression encoding method and apparatus for compressing the amount of data by encoding the digital audio waveform by a predetermined encoding method after conversion.

[0002]

2. Description of the Related Art In recent years, there has been an increasing demand for practical low bit rate speech coding due to the expansion of channel capacity in mobile communications such as automobile telephones and the necessity of storing and transmitting enormous information in multimedia communications. ing.

As an additional function of the facsimile modem, a function having a voice encoding / decoding function for an answering machine is required, and a low bit rate voice code for the voice encoding / decoding is required. There is also a desire for the development of a method for the conversion.

At present, in a low bit rate speech coding system of 10 kbps or less, CELP (Code Exit) is used.
ed Linear Prediction codi
ngsystem) method has become mainstream. This CE
The LP method uses a speech AR (Auto-
This is an encoding method based on a regressive (autoregressive) model.

Specifically, on the encoding side, speech is divided into units called frames or subframes, and LPC (LPC (LPC)
inner Prediction Coding (linear prediction) coefficient, a pitch lag representing the pitch information thereof,
It extracts noise source information, which is sound source information, and gain, performs encoding (quantization), and stores or transmits them.

On the decoding side, the encoded information is restored, an excitation source signal is generated by adding pitch information to the noise source information, and this excitation source signal is subjected to linear prediction synthesis composed of LPC coefficients. The synthesized speech is obtained through a filter.

[0007]

However, the above-mentioned conventional CELP system has an advantage that good speech can be obtained at a low bit rate of 10 kbps, but the amount of calculation in the encoding process of each parameter is small. There was a problem that there were many.

In particular, for pitch lag encoding and noise source information encoding, it is necessary to generate a synthesized speech obtained by passing the encoded excitation source signal through a linear prediction synthesis filter, and to compare the synthesized speech with the original speech. Since many operations are required for the filter operation, it is impractical to pass all the excitation source signals through the filter.

Further, the conventional CELP system has a codebook of a secondary error signal, synthesizes a secondary error signal from each code vector belonging to the codebook and a spectral envelope, and obtains a secondary error signal obtained from an input signal. Since encoding is performed by selecting a code that minimizes the distortion as compared with the signal, the amount of computation for searching for the codebook and the amount of memory for storing the codebook also increase. Was.

As a conventional technique for reducing the amount of calculation in the CELP system, for example, a preliminary selection method of narrowing down by a parameter that can be compared with the original voice approximately, instead of performing a filter calculation and comparing. Has been proposed.

Further, a noise source generally stores a noise vector corresponding to a given number of bits, and a method of reducing the amount of calculation by devising the configuration has been proposed. Specifically, VSELP which has a noise vector by the number of bits and represents a noise source by the sum or difference thereof
(Vector Sum Excited Linea
r Prediction coding) method and a method using an overlapping type code length (Japanese Patent Laid-Open No.
-54497, "Code Excited Linear Prediction Vocoder and Method Thereof").

However, due to the demand for practical low-bit-rate speech coding, in addition to the above-mentioned methods for reducing the amount of computation in the conventional CELP method (preliminary selection method, VSELP method, etc.), methods different from those methods are used. What can reduce the amount of calculation is demanded.

The present invention has been made in view of the above, and in the process of coding in the CELP system, the coding (quantization) of the secondary error signal performed by the noise code vector search is performed by the secondary error signal. An object of the present invention is to encode (quantize) a signal itself to reduce the amount of computation for encoding (quantization) and to reduce the amount of memory for storing a noise code vector.

[0014]

According to a first aspect of the present invention, there is provided a speech compression encoding method according to the first aspect of the present invention, wherein an analog speech waveform is inputted and converted into a digital speech waveform.
And a second step of encoding the digital audio waveform by a predetermined encoding method, a third step of storing the encoded digital audio waveform, and extracting the accumulated digital audio waveform. And a fifth step of converting the decoded digital audio waveform into an analog audio waveform, wherein the second step comprises: A frame dividing step of dividing a waveform into processing units called frames; a spectrum envelope information extracting step of extracting and encoding spectrum envelope information representing a spectrum envelope from each of the divided frames; Using a subframe division step of dividing into processing units called subframes and the extracted spectral envelope information, A pitch information extracting step of extracting and encoding pitch information from each of the divided subframes; a gain information extracting step of extracting and encoding gain information from the extracted pitch information; Extracting noise source information as sound source information from subframes, pitch information and gain information, and encoding the extracted noise source information, wherein the fourth step comprises the steps of: Information, gain information and noise source information, an excitation source signal generation step of generating an excitation source signal from the restored pitch information, gain information and noise source information, and an excitation source signal and the restored And a synthesized speech generating step of generating a synthesized speech from the spectrum envelope information. A second-order error signal extracting step of extracting a second-order error signal using the short-circuit information, the subframe, the pitch information, and the gain information; a discrete cosine transform step of performing a discrete cosine transform of the extracted second-order error signal; A coefficient sequence converting step of converting the discrete cosine transform coefficient sequence obtained in the cosine transform step into a predetermined bit length, thereby extracting and encoding the noise source information.

According to a second aspect of the present invention, in the first aspect, the coefficient sequence converting step converts the discrete cosine transform coefficient sequence into a predetermined bit length. At this time, after selecting a predetermined number of coefficients from the plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, the position and coefficient value of the selected coefficient and the coefficient located before and after the selected coefficient are determined. The discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding a coefficient value of a predetermined number of coefficients into a predetermined bit length.

According to a third aspect of the present invention, in the voice compression encoding method according to the second aspect, the coefficient sequence conversion step includes the step of converting a plurality of discrete cosine transforms obtained in the discrete cosine conversion step. When a predetermined number of coefficients are selected from the coefficients, the coefficient having the largest absolute value of the coefficient value is selected.

According to a fourth aspect of the present invention, in the voice compression encoding method according to the second aspect, the coefficient sequence conversion step includes the step of converting a plurality of discrete cosine transforms obtained in the discrete cosine conversion step. When a predetermined number of coefficients are selected from the coefficients, a predetermined number of coefficients are selected from those having the largest absolute value of the coefficient value.

According to a fifth aspect of the present invention, there is provided the voice compression encoding method according to the second aspect, wherein the coefficient sequence conversion step is performed by a plurality of discrete cosine transforms obtained in the discrete cosine transform step. When a predetermined number of coefficients are selected from the coefficients, a coefficient having a coefficient value exceeding a preset threshold is selected.

According to a sixth aspect of the present invention, in the first aspect, the coefficient sequence converting step converts the discrete cosine transform coefficient sequence into a predetermined bit length. At this time, a coefficient having a coefficient value exceeding a predetermined threshold is selected from a plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, and the position and coefficient value of the selected coefficient are set to a predetermined bit length. By encoding, the discrete cosine transform coefficient sequence is converted into a predetermined bit length.

According to a seventh aspect of the present invention, in the voice compression encoding method according to the first aspect, the coefficient sequence converting step converts the discrete cosine transform coefficient sequence into a predetermined bit length. At this time, one coefficient is selected from the plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, and a predetermined number of positions and coefficient values of the selected coefficient and a predetermined number of positions before and after the selected coefficient are selected. After encoding the coefficient value of the coefficient to a predetermined bit length, one coefficient is further selected from the plurality of discrete cosine transform coefficients excluding the encoded coefficient, and the position and coefficient value of the selected coefficient and The step of encoding a predetermined number of coefficient values before and after the selected coefficient to a predetermined bit length is executed a predetermined number of times, whereby the discrete cosine transform coefficient sequence And converts it into a predetermined bit length.

According to an eighth aspect of the present invention, in the voice compression encoding method according to the seventh aspect, the coefficient sequence conversion step includes the step of converting the plurality of discrete cosine transforms obtained in the discrete cosine conversion step. When selecting one coefficient from the coefficients, the coefficient having the largest absolute value of the coefficient value is selected, and when selecting one coefficient from the plurality of discrete cosine transform coefficients excluding the coded coefficient, The coefficient whose absolute value is the largest is selected.

According to a ninth aspect of the present invention, in the audio compression encoding method according to the first aspect, the coefficient sequence converting step converts the discrete cosine transform coefficient sequence into a predetermined bit length. At this time, a predetermined number of coefficients are selected from the plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, and a coefficient value of the selected coefficient and a predetermined number of coefficients located before and after the selected coefficient are selected. After obtaining the pattern of the first ratio with the coefficient value of the coefficient of the above, from the correspondence relationship information between the plurality of patterns of the second ratio and the code prepared in advance, the pattern having the property closest to the pattern of the first ratio is obtained. By selecting a pattern of the second ratio and encoding the pattern of the first ratio using a code corresponding to the selected pattern of the second ratio, the discrete cosine transform coefficient sequence is converted to a predetermined bit. Convert to long It is intended.

According to a tenth aspect of the present invention, in the voice compression encoding method according to the ninth aspect, when the coefficient sequence conversion step selects the pattern of the second ratio, The pattern of the second ratio having the smallest Euclidean distance from the pattern of the first ratio is selected.

[0024] According to a still further aspect of the present invention, there is provided a voice compression encoding method as defined in claim 10, wherein
Correspondence information between the plurality of second ratio patterns and codes prepared in advance is managed by a multidimensional data structure, and when the coefficient sequence conversion step selects the second ratio pattern, , Using the nearest point search to select the pattern of the second ratio having the smallest Euclidean distance from the pattern of the first ratio.

According to a twelfth aspect of the present invention, in the audio compression encoding method according to the ninth aspect, when the coefficient sequence conversion step selects the pattern of the second ratio, The second ratio pattern having the smallest norm with respect to the first ratio pattern is selected.

A speech compression encoding apparatus according to a thirteenth aspect of the present invention provides an A / D conversion means for inputting an analog speech waveform and converting it into a digital speech waveform, and encoding the digital speech waveform by a predetermined encoding method. Voice encoding means, a storage means for storing the coded digital voice waveform, a voice decoding means for taking out and decoding the stored digital voice waveform, and a voice decoding means for storing the decoded digital voice waveform. And a D / A converting means for converting the digital audio waveform into an analog audio waveform. The audio encoding means comprises: a frame dividing means for dividing the digital audio waveform into processing units called frames; Spectrum envelope information extraction means for extracting and encoding spectrum envelope information representing a spectrum envelope from each of the frames, A sub-frame dividing unit that divides each of the divided frames into processing units called sub-frames, and a pitch that extracts pitch information from each of the divided sub-frames and encodes the pitch information using the extracted spectral envelope information. Information extraction means, gain information extraction means for extracting and encoding gain information from the extracted pitch information, and noise source information as sound source information from the spectrum envelope information, subframe, pitch information and gain information. Source information extraction means for encoding
Wherein the speech decoding means restores the encoded spectrum envelope information, pitch information, gain information and noise source information, and an excitation source based on the restored pitch information, gain information and noise source information. An excitation source signal generation unit for generating a signal; and a synthesized speech generation unit for generating a synthesized speech from the excitation source signal and the restored spectrum envelope information. A second-order error signal extracting means for extracting a second-order error signal using the extracted spectral envelope information, subframe, pitch information, and gain information; a discrete cosine transform means for performing a discrete cosine transform of the extracted second-order error signal; A coefficient sequence converting means for converting a discrete cosine transform coefficient sequence obtained by the discrete cosine transform means into a predetermined bit length, It is intended to encode by extracting the sound source information.

According to a fourteenth aspect of the present invention, in the voice compression encoding apparatus according to the thirteenth aspect,
When the coefficient sequence converting means converts the discrete cosine transform coefficient sequence into a predetermined bit length, after selecting a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means The discrete cosine transform coefficient sequence is encoded by encoding the position and coefficient value of the selected coefficient and the coefficient values of a predetermined number of coefficients before and after the selected coefficient into a predetermined bit length. Is converted to the bit length.

According to a fifteenth aspect of the present invention, in the voice compression encoding apparatus according to the fourteenth aspect,
The coefficient sequence converting means selects a coefficient having a maximum absolute value of a coefficient value when selecting a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means. .

According to a sixteenth aspect of the present invention, in the audio compression encoding apparatus according to the fourteenth aspect,
When the coefficient sequence conversion means selects a predetermined number of coefficients from the plurality of discrete cosine transform coefficients obtained by the discrete cosine conversion means, the coefficient sequence conversion means selects a predetermined number of coefficients from the largest absolute value of the coefficient value. This is for selecting the coefficient.

According to a seventeenth aspect of the present invention, in the voice compression encoding apparatus according to the fourteenth aspect,
The coefficient sequence conversion means, when selecting a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained by the discrete cosine conversion means, selects a coefficient having a coefficient value exceeding a preset threshold value It is.

[0031] The speech compression encoding apparatus according to claim 18 is the speech compression encoding apparatus according to claim 13,
When the coefficient sequence converting means converts the discrete cosine transform coefficient sequence into a predetermined bit length, the coefficient sequence converting means has a coefficient value exceeding a predetermined threshold from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform coefficient. After selecting a coefficient, the discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding the position and coefficient value of the selected coefficient to a predetermined bit length.

[0032] The speech compression encoding apparatus according to claim 19 is the speech compression encoding apparatus according to claim 13,
When the coefficient sequence converting means converts the discrete cosine transform coefficient sequence into a predetermined bit length, the coefficient sequence converting means selects one coefficient from the plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means, and After encoding the position and coefficient value of a coefficient and the coefficient values of a predetermined number of coefficients positioned before and after the selected coefficient to a predetermined bit length, the plurality of discrete values excluding the encoded coefficient are further encoded. Selecting one coefficient from the cosine transform coefficients and encoding the position and coefficient value of the selected coefficient and the coefficient values of a predetermined number of coefficients positioned before and after the selected coefficient into a predetermined bit length; Is executed a predetermined number of times to convert the discrete cosine transform coefficient sequence into a predetermined bit length.

According to a twentieth aspect of the present invention, in the voice compression encoding apparatus according to the nineteenth aspect,
When the coefficient sequence converting means selects one coefficient from the plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means, the coefficient sequence converting means selects the coefficient having the largest absolute value of the coefficient value, and selects the coded coefficient. When selecting one coefficient from the plurality of discrete cosine transform coefficients except for the coefficient, the coefficient having the largest absolute value of the coefficient value is selected.

According to a twenty-first aspect of the present invention, in the audio compression encoding apparatus according to the thirteenth aspect,
When the coefficient sequence converting means converts the discrete cosine transform coefficient sequence into a predetermined bit length, a predetermined number of coefficients are selected from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means, After obtaining a first ratio pattern between a coefficient value of the selected coefficient and coefficient values of a predetermined number of coefficients positioned before and after the selected coefficient, a plurality of second ratio patterns prepared in advance And selecting the second ratio pattern having the property closest to the first ratio pattern from the correspondence information between the first ratio pattern and the first ratio pattern using the code corresponding to the selected second ratio pattern. The discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding a pattern having a ratio of.

The speech compression encoding apparatus according to claim 22 is the speech compression encoding apparatus according to claim 21,
The coefficient sequence converting means selects the second ratio pattern having the shortest Euclidean distance from the first ratio pattern when selecting the second ratio pattern.

According to a twenty-third aspect of the present invention, in the voice compression encoding apparatus according to the twenty-second aspect,
Correspondence information between the plurality of second ratio patterns and codes prepared in advance is managed by a multidimensional data structure, and when the coefficient sequence conversion means selects the second ratio pattern, , Using the nearest point search to select the pattern of the second ratio having the smallest Euclidean distance from the pattern of the first ratio.

Further, in the speech compression encoding apparatus according to claim 24, in the speech compression encoding apparatus according to claim 21, when the coefficient sequence converting means selects the pattern of the second ratio, The second ratio pattern having the smallest norm with respect to the first ratio pattern is selected.

[0038]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A speech compression encoding method and apparatus according to the present invention will be described below with reference to [Embodiment 1], [Embodiment 2], [Embodiment 3], [Embodiment 4], [ Fifth Embodiment, Sixth Embodiment, Seventh Embodiment,
[Embodiment 8], [Embodiment 9], [Embodiment 1]
0], [Embodiment 11], and [Embodiment 12] will be described in detail with reference to the drawings.

[Embodiment 1] FIG. 1 is a schematic block diagram of a speech compression encoding apparatus 100 according to Embodiment 1. The audio compression encoding apparatus 100 receives an analog signal (analog audio waveform) and inputs a digital signal (digital audio waveform).
A / D conversion unit 101 as A / D conversion means for converting to A / D
And a digital signal input from the A / D conversion unit 101, and stores an audio encoding unit 102 as audio encoding means for performing compression encoding, and stores a compression-encoded compression-encoded signal (quantized signal). A storage unit 103 serving as storage means, a voice decoding unit 104 serving as voice decoding means for expanding and decoding the compressed coded signal (quantized signal), and a D / A converting the decoded digital signal into an analog signal. And a D / A conversion unit 105 as conversion means.

FIG. 2 is a block diagram showing the arrangement of the speech encoding unit 102. The speech encoding unit 102 divides the input digital signal into frames each having a predetermined number of samples, and outputs a frame signal.
And a spectrum envelope extraction unit 202 for extracting and encoding spectrum envelope information representing a spectrum envelope in frame units from the frame (frame signal) divided by the frame configuration unit 201, and further dividing the frame divided by the frame configuration unit 201. The sub-frame forming unit 203 divides the data into sub-frame units of a predetermined number of samples and outputs a sub-frame signal, and the spectral envelope information extracted by the spectral envelope extracting unit 202.
03, a pitch information extracting unit 204 for extracting and encoding pitch information from the subframe, a gain extracting unit 205 for extracting and encoding gain information from the pitch information, a spectrum envelope information, a subframe, and pitch information. And a noise source extracting unit 206 that extracts noise source information, which is sound source information, from the gain information and encodes it.

FIG. 3 shows a block diagram of the noise source extraction unit 206. The noise extraction unit 206 extracts the spectrum envelope information extracted by the spectrum envelope extraction unit 202, the subframe signal output from the subframe construction unit 203, the pitch information extracted by the pitch information extraction unit 204, and the gain extraction unit 205. Using the gain information obtained,
A secondary error signal composing unit 301 for extracting a secondary error signal;
DCT (Discc) for performing a discrete cosine transform of the secondary error signal
The DCT transform unit 302 includes a transform unit 302 and a coefficient transform unit 303 that transforms a DCT transform coefficient sequence obtained by the DCT transform by the DCT transform unit 302 into a predetermined bit length.

FIG. 4 is a block diagram of the speech decoding unit 104. The audio decoding unit 104 stores
3, a spectrum envelope decoding unit 401 for restoring encoded spectrum envelope information, a pitch information decoding unit 402 for restoring encoded pitch information, and encoded noise. Noise source decoding section 403 for restoring noise source information from source information, gain decoding section 404 for restoring encoded gain information, restored spectrum envelope information and restored pitch information, gain information, and noise source information And a speech synthesis unit 405 that generates a synthesized speech from the excitation source signal generated from the speech.

The operation of the above configuration will be described. In FIG. 1, an analog voice input device (not shown)
The analog signal (analog sound waveform) input from is A
The signal is converted into a digital signal by the / D converter 101. Here, as an analog voice input device, for example,
Examples include a microphone, a CD player, and a cassette deck.

Subsequently, the digital signal is sent to the speech encoding unit 102, and as shown in FIG.
01 (for example, 25
(Six samples).
This frame is output as a frame signal to spectrum envelope extracting section 202 and subframe forming section 203.

The spectrum envelope extraction unit 202 extracts and encodes (quantizes) the spectrum envelope information from the frame signal, and extracts the pitch information extraction unit 204 and the noise source extraction unit 2.
06. As the spectrum envelope information, for example, a linear prediction coefficient based on linear prediction analysis, PARCOR
Coefficients, LSP coefficients, and the like. The encoding (quantization) of the spectral envelope information includes vector quantization, scalar quantization, split vector quantization, multi-stage vector quantization, or a combination of these multiple quantizations.

On the other hand, the sub-frame forming section 203 receives a frame signal from the frame forming section 201, divides the frame signal into a predetermined number of samples (for example, 64 samples), and outputs it as a sub-frame signal.

Each sub-frame is provided with a pitch information extracting unit 20.
In 4, the pitch information is extracted and encoded using the spectrum envelope information extracted by the spectrum envelope extraction unit 202. CELP for pitch information extraction
An adaptive codebook search used in the scheme or a method of obtaining from spectral information such as Fourier transform coefficients and wavelet transform coefficients can be applied. In addition, the adaptive codebook search may use an auditory weighting filter. The auditory weighting filter can be composed of the above-described linear prediction coefficients.

FIG. 2 shows an apparatus configuration for extracting pitch information by adaptive codebook search. The pitch information extracted by the pitch information
05, gain information (gain component) is extracted and encoded.

In noise source extracting section 206, as shown in FIG. 3, secondary error signal forming section 301 forms a secondary error signal using the subframe signal, spectrum envelope information, pitch information and gain information. At this time, a pitch component residual signal is formed by the residual signal up to the previous subframe, the pitch information extracted by the pitch information extracting unit 204, and the gain information extracted by the gain extracting unit 205.
Further, a second-order error signal can be obtained by constructing a pitch component signal from the composed pitch component residual signal and the spectrum envelope information and then subtracting the pitch component signal from the subframe signal. In order to obtain a pitch component signal from the pitch component residual signal and the spectrum envelope information, a method of passing the residual signal through a synthesis filter obtained from the spectrum envelope information can be used. The secondary error signal thus configured is supplied to the DCT converter 3
02 is output.

The DCT transform unit 302 receives the secondary error signal, performs DCT transform, and outputs a plurality of DCT transform coefficients obtained by the DCT transform to the coefficient transform unit 303.

The coefficient transforming unit 303 receives a plurality of DCT transform coefficients and converts the DCT transform coefficient sequence into a predetermined bit length. That is, the DCT transform coefficient sequence is encoded and output as noise source information.

As described above, the speech encoding unit 102 extracts and encodes spectral envelope information, pitch information, gain information, and noise source information from the digital signal, and outputs these as a quantized signal. These quantized signals are accumulated by the accumulation unit 103 as compression-encoded signals.

The quantized signal stored in the storage unit 103 is
The audio data is read out and decoded (restored) by the audio decoding unit 104 as necessary. In the speech decoding unit 104, as shown in FIG. 4, the spectrum envelope information is restored by the spectrum envelope decoding unit 401, and the pitch information decoding unit 40
2, the pitch information is restored, the noise source information is restored by the noise source decoding unit 403, and the gain information is restored by the gain decoding unit 404. Here, the restored pitch information, noise source information and gain information constitute a residual signal (excitation source signal). The speech synthesizer 405 generates a decoded speech (synthesized speech) as a digital signal from the restored spectrum envelope information and the residual signal, and outputs the decoded speech to the D / A converter 105.

The digital signal output from the voice synthesizer 405 is converted to an analog signal (analog voice waveform) by the D / A converter 105 as shown in FIG.

The speech coding method used in the present invention is a coding method belonging to CELP speech coding. The conventional CELP method has a codebook for the secondary error signal,
The secondary error signal is synthesized from each code vector belonging to the codebook and the spectral envelope information, compared with the secondary error signal obtained from the input signal, and the code having the minimum distortion is selected to perform the encoding. Is going. Incidentally, an auditory weighting filter can be used in this search.

However, although the CELP system is a high-quality speech compression coding technique at a low bit rate, the CELP system has a problem of a large amount of calculation for searching a codebook and a large amount of memory for storing a codebook. I have. On the other hand, according to the audio compression encoding apparatus of the first embodiment, since the secondary error signal itself is encoded (quantized), the amount of calculation can be reduced,
Also, since there is no need to store a codebook, a low memory C
An ELP method can be provided.

[Second Embodiment] The voice compression encoding apparatus according to the second embodiment is the same as the voice compression encoding apparatus according to the first embodiment, except that the coefficient conversion unit 303 converts the DCT-transformed coefficient sequence into a predetermined bit length. In this case, after selecting a predetermined number of coefficients from a plurality of DCT transform coefficients, the position and the coefficient value of the selected coefficient and the coefficient values of the predetermined number of coefficients located before and after the selected coefficient are determined. The DCT transform coefficient sequence is converted into a predetermined bit length by encoding to a predetermined bit length.

Speech compression encoding apparatus 100 of the first embodiment
In, the larger the number of DCT transform coefficients to be selected / encoded, the more the sound quality of the reproduced sound can be improved. For that purpose, it is necessary to perform a process of selecting / encoding a large number of coefficients with a smaller number of bits. Therefore, the coefficient conversion unit 303 of the second embodiment selects a predetermined number of coefficients, and determines the position and coefficient value of the selected coefficient and the coefficient values of the predetermined number of coefficients located before and after the selected coefficient. Is encoded into a predetermined bit length. As a result, it is not necessary to encode the position information for the coefficients located before and after the selected coefficient, and it is possible to encode more coefficients. .

FIG. 5 shows a coefficient conversion unit 303 according to the second embodiment.
6 is a flowchart showing the operation procedure of FIG. The basic configuration and operation of the audio compression encoding apparatus according to the second embodiment are the same as those of the audio compression encoding apparatus according to the first embodiment described with reference to FIGS. 1 to 4, and description thereof is omitted here. I do.

In FIG. 5, the coefficient conversion unit 303
, Input a plurality of DCT transform coefficients obtained by the DCT transform from the DCT transform unit 302, and select predetermined M coefficients (Ci (i = 0,..., M−1)) (S501). ).

Subsequently, the coefficient conversion unit 303 determines in step S
One of the coefficients selected in 501, for example, Ci
The coefficient of (i = 0) is selected (S502). Then, in step S503, the position and coefficient value of the coefficient and the coefficient values of the predetermined N coefficients positioned before and after the selected coefficient are encoded into a predetermined bit length.

After the encoding process is performed on the coefficients (for example, Ci (i = 0)), in step S504, all the coefficients (Ci (i = 0,...) Selected in step S501 are selected.
.., M-1)), it is determined whether or not the process of step S503 has been executed. If the encoding process has been performed for all the selected coefficients, the process ends. On the other hand, if the encoding has not been completed for all the coefficients, step S
Proceeding to step 505, for each of the remaining coefficients, step S
503 is executed. That is, the coefficient conversion unit 303
Converts the DCT-transformed coefficient sequence into a predetermined bit length by executing the processing of step S503 for all selected coefficients and encoding the selected coefficients, as shown in FIG.

As described above, according to the audio compression encoding apparatus of the second embodiment, it is not necessary to encode position information for coefficients located before and after the selected coefficient, and it is possible to encode more coefficients. DCT after conversion
The sound quality of the reproduced sound can be improved without increasing the number of bits of the conversion coefficient sequence.

For example, if the subframe length is 64 samples,
When 5 bits including +/- sign are added to the coefficient value encoding, 11 bits are required to encode one DCT transform coefficient. Therefore, generally, 22 bits are required to encode two predetermined coefficients. On the other hand, when the audio compression encoding apparatus according to Embodiment 2 is used, 21
Bits can encode three coefficients.

[Third Embodiment] The speech compression encoding apparatus according to the third embodiment is the same as the speech compression encoding apparatus according to the second embodiment, except that the coefficient transforming section 303 includes a plurality of DCTs obtained by the DCT transforming section 302. When a predetermined number of coefficients are selected from the conversion coefficients, the coefficient having the largest absolute value of the coefficient value is selected.

FIG. 6 shows a coefficient conversion unit 303 according to the third embodiment.
6 is a flowchart showing the operation procedure of FIG. The basic operation is the same as that of the audio compression encoding apparatus according to the second embodiment shown in FIG. 5, and only different parts will be described here.

In FIG. 6, the coefficient conversion unit 303
A plurality of DCT transform coefficients obtained by the DCT transform are input from the DCT transform unit 302 shown in (1), and the coefficient Ci having the largest absolute value of the coefficient value is selected (S601).

Subsequently, the coefficient conversion unit 303 performs the processing according to the second embodiment.
As described above, by coding the position and coefficient value of the selected coefficient and the coefficient values of predetermined N coefficients before and after the selected coefficient to a predetermined bit length,
The CT conversion coefficient sequence is converted into a predetermined bit length (S50).
3)

As described above, according to the audio compression encoding apparatus of the third embodiment, since the coefficient having the largest absolute value of the coefficient value is selected from the plurality of DCT transform coefficients, the coefficient to be encoded is easily selected. be able to.

[Fourth Embodiment] The speech compression encoding apparatus according to the fourth embodiment is the same as the speech compression encoding apparatus according to the second embodiment, except that the coefficient transforming section 303 includes a plurality of DCTs obtained by the DCT transforming section 302. When a predetermined number of coefficients are selected from the conversion coefficients, a predetermined number of coefficients are selected from those having the largest absolute value of the coefficient value.

FIG. 7 shows a coefficient conversion unit 303 according to the fourth embodiment.
6 is a flowchart showing the operation procedure of FIG. The basic operation is the same as that of the audio compression encoding apparatus according to the second embodiment shown in FIG. 5, and only different parts will be described here.

In FIG. 7, the coefficient conversion unit 303
, A plurality of DCT transform coefficients obtained by the DCT transform from the DCT transform unit 302, and a predetermined number of M coefficients (Ci (i = 0,
.., M-1)) are selected (S701).

Thereafter, the coefficient conversion unit 303 determines in step S
For each of the coefficients selected in step 701, the DCT transform coefficient sequence is obtained by encoding the position and coefficient value of the coefficient and the coefficient values of predetermined N coefficients located before and after the selected coefficient into a predetermined bit length. Is converted to a predetermined bit length (S502 to S505).

As described above, according to the audio compression encoding apparatus of the fifth embodiment, a predetermined number of coefficients are selected from a plurality of DCT transform coefficients from those having the largest absolute value of the coefficient value. The coefficient to be converted can be easily selected. In addition, the DCT transform coefficient generally has a property that both the coefficient located near the coefficient having a high absolute value and the absolute value of the coefficient value are high. , The DCT transform coefficient sequence can be more efficiently converted into a predetermined bit length.

[Fifth Embodiment] The voice compression encoding apparatus according to the fifth embodiment is the same as the voice compression encoding apparatus according to the second embodiment, except that the coefficient transforming section 303 includes a plurality of DCTs obtained by the DCT transforming section 302. When a predetermined number of coefficients are selected from the conversion coefficients, a coefficient having a coefficient value exceeding a preset threshold is selected.

FIG. 8 is a diagram showing a coefficient conversion unit 303 according to the fifth embodiment.
6 is a flowchart showing the operation procedure of FIG. The basic operation is the same as that of the audio compression encoding apparatus according to the second embodiment shown in FIG. 5, and only different parts will be described here.

In FIG. 5, the coefficient conversion unit 303
, A plurality of DCT transform coefficients obtained by the DCT transform from the DCT transform unit 302, and a coefficient (Ci (i = 0,...) Having a coefficient value exceeding a preset threshold Th.
., M-1)) is selected (S801).

Thereafter, the coefficient conversion unit 303 determines in step S
For each of the coefficients selected in 801, the DCT transform coefficient sequence is obtained by encoding the position and coefficient value of the coefficient and the coefficient values of predetermined N coefficients located before and after the selected coefficient into a predetermined bit length. Is converted to a predetermined bit length (S502 to S505).

As described above, according to the speech compression encoding apparatus of the fifth embodiment, since a coefficient having a coefficient value exceeding a preset threshold value is selected from a plurality of DCT transform coefficients, the coefficient to be encoded can be easily determined. You can choose.

[Embodiment 6] The speech compression encoding apparatus according to the sixth embodiment is the same as the speech compression encoding apparatus according to the first embodiment, except that the coefficient transforming unit 303 transforms the DCT transform coefficient sequence into a predetermined bit length. In this case, a coefficient having a coefficient value exceeding a predetermined threshold value is selected from a plurality of DCT transform coefficients, and a position and a coefficient value of the selected coefficient are encoded into a predetermined bit length, thereby obtaining a DCT transform coefficient sequence. Is converted into a predetermined bit length.

FIG. 9 shows a coefficient conversion unit 303 according to the sixth embodiment.
6 is a flowchart showing the operation procedure of FIG. The basic configuration and operation of the speech compression encoding apparatus according to the sixth embodiment are the same as those of the speech compression encoding apparatus according to the first embodiment described with reference to FIGS. 1 to 4, and description thereof is omitted here. I do.

In FIG. 9, the coefficient conversion unit 303
, A plurality of DCT transform coefficients obtained by the DCT transform from the DCT transform unit 302, and a coefficient (Ci (i = 0,...) Having a coefficient value exceeding a preset threshold Th.
., M-1)) is selected (S901).

Subsequently, the coefficient conversion unit 303 determines in step S
One of the coefficients selected in 901, for example, Ci
The coefficient of (i = 0) is selected (S902). Then, in step S903, the position and coefficient value of the selected coefficient are encoded into a predetermined bit length.

In step S903, the coefficient (for example, Ci
(I = 0)), and then, step S9
In 04, it is determined whether or not the processing of step S903 has been performed for each coefficient selected in step S901. If the coding has been performed on all the selected coefficients, the process ends. On the other hand, if the coding has not been completed on all the coefficients, the process proceeds to step S905, and the process of step S903 is performed on the remaining coefficients. In other words, as shown in FIG. 9, the code conversion unit 303 executes the process of step S903 for all the selected coefficients to perform coding, thereby converting the DCT-transformed coefficient sequence into a predetermined bit length.

As described above, according to the speech compression encoding apparatus of the sixth embodiment, when converting a DCT transform coefficient sequence into a predetermined bit length, a coefficient value exceeding a predetermined threshold value is obtained from a plurality of DCT transform coefficients. Is selected without increasing the number of bits in the transformed DCT coefficient sequence.
The sound quality of the reproduced sound can be improved, and the coefficient to be encoded can be easily selected.

[Seventh Embodiment] The speech compression encoding apparatus according to the seventh embodiment is the same as the speech compression encoding apparatus according to the first embodiment, except that the coefficient transforming unit 303 transforms the DCT transform coefficient sequence into a predetermined bit length. In this case, one coefficient is selected from a plurality of DCT transform coefficients, and the position and coefficient value of the selected coefficient and the coefficient values of a predetermined number of coefficients before and after the selected coefficient are set to a predetermined bit length. After encoding, one coefficient is selected from a plurality of DCT transform coefficients excluding the encoded coefficient, and the position and coefficient value of the selected coefficient and a predetermined number of coefficients before and after the selected coefficient are selected. The step of encoding the coefficient value into a predetermined bit length is performed a predetermined number of times, thereby converting the DCT transform coefficient sequence into a predetermined bit length.

FIG. 10 is a diagram showing a coefficient conversion unit 30 according to the seventh embodiment.
6 is a flowchart showing an operation procedure of No. 3; The basic configuration and operation of the speech compression encoding apparatus according to the seventh embodiment are the same as those of the speech compression encoding apparatus according to the first embodiment described with reference to FIGS. 1 to 4, and description thereof is omitted here. I do.

In the coefficient transforming section 303, the number of coefficients to be selected and encoded from a plurality of DCT transform coefficients is set in advance (for example, M). In FIG. 10, the coefficient conversion unit 3
03 inputs a plurality of DCT transform coefficients obtained by the DCT transform from the DCT transform unit 302 shown in FIG. 2 and confirms that the process enters the first coefficient process (S1001).
One coefficient is selected from the DCT transform coefficients (S100
2).

Subsequently, the coefficient conversion unit 303 determines in step S
The position and coefficient value of the coefficient selected in 1002 and the coefficient values of the predetermined N coefficients positioned before and after the selected coefficient are encoded into a predetermined bit length (S1003).

After coding the coefficients selected in step S1002, it is determined in step S1004 whether or not the processing of all of the M previously set coefficients has been completed. If encoding has been performed for all coefficients to be processed, the process ends.

On the other hand, if the coding has not been completed for all the coefficients, the flow advances to step S1005 to add 1 to the number of coefficients to be coded. Then, step S1002
Then, one coefficient is selected from the DCT transform coefficients excluding the already coded coefficients, and encoding is performed in step S1003. Then, in step S1004, the selection and coding processing of the repetition coefficient is executed until it is confirmed that the above processing has been performed on the M coefficients.

As described above, according to the audio compression encoding apparatus of the seventh embodiment, it is not necessary to encode position information for coefficients located before and after the selected coefficient, and encoding is performed for more coefficients. Enabled, the converted DC
The sound quality of the reproduced sound can be improved without increasing the number of bits of the T conversion coefficient sequence. In addition, since the already encoded DCT transform coefficients are not re-encoded, the encoding process can be executed more efficiently.

[Eighth Embodiment] An audio compression encoding apparatus according to the eighth embodiment is the same as the audio compression encoding apparatus according to the seventh embodiment, except that the coefficient conversion unit 303 selects one coefficient from a plurality of DCT conversion coefficients. When the coefficient is selected, the coefficient with the largest absolute value is selected, and when selecting one coefficient from a plurality of DCT transform coefficients excluding the encoded coefficient, the coefficient with the largest absolute value is selected. Is what you do.

FIG. 11 shows the coefficient conversion unit 30 according to the eighth embodiment.
6 is a flowchart showing an operation procedure of No. 3; The basic operation is the same as that of the speech compression encoding apparatus according to the seventh embodiment shown in FIG. 10, and only different parts will be described here.

In the coefficient transforming section 303, the number of coefficients to be selected and coded from the DCT transform coefficients is set in advance (for example, M). In FIG. 11, a coefficient conversion unit 303
Inputs a plurality of DCT transform coefficients obtained by the DCT transform from the DCT transform unit 302 shown in FIG. 2 and confirms that the process enters the first coefficient process (S1001).
The coefficient having the largest absolute value of the coefficient value is selected from the T-transform coefficients (S1101).

Subsequently, the coefficient conversion unit 303 determines in step S
After the position and coefficient value of the coefficient selected in 1002 and the coefficient values of the predetermined N coefficients positioned before and after the selected coefficient are coded to a predetermined bit length (S1003), the coded data are further coded. A coefficient having the largest absolute value of the coefficient value is selected from a plurality of DCT transform coefficients excluding the coefficient (S110).
1), encoding is performed in step S1003. Then, in step S1004, the selection and coding processing of the repetition coefficient is executed until it is confirmed that the above processing has been performed on the M coefficients.

As described above, according to the audio compression encoding apparatus of the eighth embodiment, when one coefficient is selected from a plurality of DCT transform coefficients, the coefficient having the largest absolute value of the coefficient value is selected, When one coefficient is selected from the plurality of DCT transform coefficients excluding the converted coefficients, the coefficient having the largest absolute value of the coefficient value is selected, so that the coefficient to be coded can be easily selected. Also, DCT transform coefficients generally have a property that both the coefficient located near a coefficient having a high absolute value and the absolute value of the coefficient value are high. By making a selection, the DCT transform coefficient sequence can be more efficiently converted to a predetermined bit length.

[Embodiment 9] The speech compression encoding apparatus according to the ninth embodiment is the same as the speech compression encoding apparatus according to the first embodiment, except that the coefficient transform unit 303 transforms the DCT transform coefficient sequence into a predetermined bit length. In this case, a predetermined number of coefficients are selected from a plurality of DCT transform coefficients, and a pattern of the ratio between the coefficient value of the selected coefficient and the coefficient value of a predetermined number of coefficients located before and after the selected coefficient (see FIG. After obtaining the first ratio pattern), the closest relationship to the first ratio pattern is obtained from the correspondence information (pattern database) between a plurality of ratio patterns (second ratio patterns) and codes prepared in advance. By selecting the pattern of the second ratio having the property and encoding the pattern of the first ratio using the code given to the selected pattern of the second ratio, the DCT transform coefficient sequence has a predetermined bit length. Is converted to

The coefficient conversion unit 303 corresponds to a ratio pattern (first ratio pattern) between the coefficient value of the selected coefficient and the coefficient value of a predetermined number of coefficients located before and after the selected coefficient. A pattern database (not shown) storing a plurality of ratio patterns (second ratio patterns) is provided in advance. Each of the patterns of the second ratio stored in the pattern database is given a sign, and the coefficient conversion unit 303 determines the plurality of patterns of the second ratio stored in the pattern database and the obtained first ratio. Are compared with each other, and the second ratio pattern having the property closest to the first ratio pattern is selected, whereby encoding is performed using the code given to the selected second ratio pattern. Do.

FIG. 12 is a diagram showing a coefficient conversion unit 30 according to the ninth embodiment.
6 is a flowchart showing an operation procedure of No. 3; The basic configuration and operation of the speech compression encoding apparatus according to the ninth embodiment are the same as those of the speech compression encoding apparatus according to the first embodiment described with reference to FIGS. 1 to 4, and description thereof is omitted here. I do.

In FIG. 12, a coefficient transforming unit 303 receives a plurality of DCT transform coefficients obtained by DCT transform from the DCT transforming unit 302 shown in FIG. 2, and sets M predetermined coefficients (Ci (i = 0) ,..., M-1)) are selected (S1201).

Subsequently, the coefficient conversion unit 303 determines in step S
One coefficient (for example, C
i (i = 0)), and a ratio pattern (first ratio pattern) between the coefficient value of the coefficient and the coefficient value of a predetermined number of coefficients before and after the selected coefficient is obtained ( S12
02). Then, the coefficient conversion unit 303 determines that the J ratio patterns (second patterns) stored in the pattern database described above.
A second ratio pattern (for example, J (j = 0)) is read from the ratio pattern: J (j = 0,..., J−1)) (S1203).

In step S1204, the coefficient conversion unit 303 compares the pattern of the first ratio obtained in step S1202 with the pattern of the second ratio read from the pattern database in step S1203, and evaluates as a comparison result. Find the value Dj.

After obtaining the evaluation value Dj, the coefficient conversion unit 303
Compares the magnitude relationship between the obtained evaluation values Dj and Dmin (S1205). For example, if it is assumed that the pattern of the first ratio relating to the coefficient Ci (i = 0) is being compared with the pattern of the second ratio stored in the pattern database, the coefficient Ci ( For i = 0), the best evaluation value Dj at present is set. However, when comparing for the first time a pattern of the first ratio with respect to the coefficient Ci (i = 0), a preset initial value is used as Dmin.

As a result of comparing Dj and Dmin in step S1205, if Dj is smaller by Dmim, the flow advances to step S1206 to set this evaluation value Dj to Dmim, and to determine the current pattern in the first ratio pattern thus obtained. The evaluation value D is defined as a pattern MIN of a second ratio having similar properties.
A second ratio pattern J from which j is obtained is set. On the other hand, if Dj is greater than Dmin, step S120
Go to 7. In addition, one coefficient (for example, Ci (i = 0))
As a result of completing the comparison of the first ratio pattern with respect to all the second ratio patterns stored in the pattern database, the second ratio pattern set in MIN is the first ratio obtained. And the code assigned to the second ratio pattern is selected as the code of the first ratio pattern for the coefficient (Ci (i = 0)).

In step S1207, it is determined whether the comparison between the pattern of the first ratio and all the patterns of the second ratio stored in the pattern database has been completed. If the comparison with all the second ratio patterns has not been completed, the process proceeds to step S1209, and the first ratio pattern and the next second ratio pattern (for example, J = 1) are compared. .

On the other hand, if the comparison between the pattern of the first ratio and all the patterns of the second ratio stored in the pattern database is completed, the process proceeds to step S1208, where all the coefficients Ci ( i = 0, ...
.., M-1), it is determined whether the comparison with the second ratio pattern stored in the pattern database has been completed.

When the comparison has been completed for all the selected coefficients Ci, the encoding process is completed. On the other hand, when the comparison has not been completed for all the selected coefficients Ci, the next coefficient (for example, i = 1) is compared with the second ratio pattern stored in the pattern database.

As described above, according to the speech compression encoding apparatus of the ninth embodiment, it is not necessary to encode position information for coefficients located before and after the selected coefficient, and encoding is performed for more coefficients. Enabled, the converted DC
The sound quality of the reproduced sound can be improved without increasing the number of bits of the T conversion coefficient sequence. It should be noted that when matching the pattern of the ratio prepared in advance and the pattern of the obtained ratio is performed, it looks like it is equivalent to performing a codebook search. However, in the case of the ninth embodiment, the noise code in CELP is used. Since the number of dimensions is much smaller than book search, the amount of computation is much smaller.

[Embodiment 10] The speech compression encoding apparatus according to the tenth embodiment differs from the speech compression encoding apparatus according to the ninth embodiment in that the coefficient conversion unit 303 selects a pattern having the second ratio. , The second ratio pattern having the smallest Euclidean distance (square distance: the square root of the difference in each coordinate axis and taking the square root thereof) with the first ratio pattern is selected.

FIG. 13 shows the coefficient conversion unit 3 according to the tenth embodiment.
11 is a flowchart showing an operation procedure of the third embodiment. Note that the basic operation is the same as that of the speech compression encoding apparatus according to the ninth embodiment shown in FIG. 12, and only different parts will be described here.

The coefficient conversion unit 303 according to the tenth embodiment compares the pattern of the first ratio obtained in step S1202 with the pattern of the second ratio read from the pattern database in step S1203, and generates an evaluation value Dj. The Euclidean distance is obtained (S1301). That is, in searching for the ratio pattern stored in the pattern database, the ratio pattern is considered as a vector in a multidimensional space, and the Euclidean distance between the obtained first ratio pattern and the stored second ratio pattern is calculated. Ask. And the first
As a result of comparing the pattern of the second ratio with the patterns of all the second ratios stored in the pattern database, the pattern of the second ratio having the smallest Euclidean distance is selected from the pattern database, and the selected second ratio is selected. An encoding process is performed using the code given to the pattern.

As described above, according to the speech compression encoding apparatus of the tenth embodiment, when selecting the pattern of the second ratio, the second ratio having the smallest Euclidean distance from the pattern of the first ratio is selected. In this case, the pattern of the second ratio having the property closest to the pattern of the first ratio can be easily selected.

[Eleventh Embodiment] The speech compression encoding apparatus according to the eleventh embodiment is the same as the speech compression encoding apparatus according to the tenth embodiment, except that the correspondence between a plurality of second ratio patterns prepared in advance and the code. Information (pattern database) is managed by a multi-dimensional data structure, and the coefficient conversion unit 30
3 selects the second ratio pattern having the shortest Euclidean distance from the first ratio pattern by using the nearest point search when selecting the second ratio pattern.

The pattern database according to the eleventh embodiment has a multidimensional data structure such as a B-tree, kd-tree, and quadtree. In step S1301 in FIG. 13, the coefficient conversion unit 303 performs a nearest point search using the pattern database having the multidimensional data structure.
An encoding process is executed by selecting, from the pattern database, a code corresponding to the pattern having the ratio that minimizes the Euclidean distance from the obtained pattern of the ratio.

As described above, according to the speech compression encoding apparatus of the eleventh embodiment, since a pattern database having a multidimensional data structure such as a B-tree, kd-tree, or quadtree is used, the stored ratio Can be efficiently managed, and the search time can be reduced. Further, when selecting the pattern of the second ratio, the closest point search is used to select the pattern of the second ratio having the smallest Euclidean distance from the pattern of the first ratio. The pattern of the second ratio having the property closest to the above can be easily selected.

[Twelfth Embodiment] The speech compression encoding apparatus according to the twelfth embodiment is different from the speech compression encoding apparatus according to the ninth embodiment in that the coefficient conversion unit 303 selects a pattern having the second ratio. , The second ratio pattern having the smallest norm (absolute value distance: adding the absolute value of the difference in each coordinate axis) with the first ratio pattern is selected.

FIG. 14 is a diagram showing a coefficient conversion unit 3 according to the twelfth embodiment.
11 is a flowchart showing an operation procedure of the third embodiment. Note that the basic operation is the same as that of the speech compression encoding apparatus according to the ninth embodiment shown in FIG. 12, and only different parts will be described here.

The coefficient conversion unit 303 according to the twelfth embodiment compares the pattern of the first ratio obtained in step S1202 with the pattern of the second ratio read from the pattern database in step S1203, and generates an evaluation value Dj. A norm is obtained (S1401). That is, in the search for the ratio pattern stored in the pattern database, the ratio pattern is considered as a vector in a multidimensional space, and the norm between the obtained first ratio pattern and the stored second ratio pattern is calculated. . Then, as a result of comparing the pattern of the first ratio with all the patterns of the second ratio stored in the pattern database, the pattern of the second ratio having the smallest norm is selected from the pattern database, and the selected second pattern is selected. The encoding process is performed using the code given to the pattern of the ratio of.

As described above, according to the speech compression encoding apparatus of the twelfth embodiment, when the second ratio pattern is selected, the second ratio pattern having the smallest norm with the first ratio pattern is selected. In order to select a pattern, it is possible to easily select a pattern having a second ratio having properties closest to the pattern having the first ratio.

[0121]

As described above, according to the speech compression / encoding method of the present invention (claim 1), the noise source extracting step includes:
A second-order error signal extracting step of extracting a second-order error signal using the extracted spectral envelope information, subframe, pitch information, and gain information; a discrete cosine transform step of performing a discrete cosine transform of the extracted second-order error signal; And a coefficient sequence converting step of converting the discrete cosine transform coefficient sequence obtained in the discrete cosine transform step into a predetermined bit length, thereby extracting and encoding noise source information. In the process, the amount of calculation can be reduced and the amount of memory can be reduced. That is, according to the speech compression encoding method of the present invention, in the quantization of the secondary error signal, the codebook search is not performed by using the filter calculation without using the codebook. A high-quality and low-bit-rate voice compression encoding method to be realized can be obtained.

Further, according to the audio compression encoding method of the present invention (claim 2), in the audio compression encoding method according to claim 1, the coefficient sequence conversion step includes the step of converting the discrete cosine transform coefficient sequence to a predetermined bit length. In the conversion to a discrete cosine transform step, after selecting a predetermined number of coefficients from the plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, the positions of the selected coefficients and the coefficient values and the positions before and after the selected coefficients are determined. By coding the coefficient value of the specified number of coefficients and the predetermined bit length,
Since the discrete cosine transform coefficient sequence is converted to a predetermined bit length, it is not necessary to encode position information for coefficients located before and after the selected coefficient, and more coefficients can be encoded. Therefore, the transformed DCT
The sound quality of the reproduced sound can be improved without increasing the number of bits of the conversion coefficient sequence.

Further, according to the audio compression encoding method of the present invention (claim 3), in the audio compression encoding method according to claim 2, the coefficient sequence conversion step includes a plurality of steps obtained by the discrete cosine conversion step. When a predetermined number of coefficients are selected from the discrete cosine transform coefficients, the coefficient having the maximum absolute value of the coefficient value is selected, so that the coefficient to be coded can be easily selected.

Further, according to the voice compression encoding method of the present invention (claim 4), in the voice compression encoding method according to claim 2, the coefficient sequence conversion step includes a plurality of steps obtained by the discrete cosine conversion step. When selecting a predetermined number of coefficients from the discrete cosine transform coefficients, the coefficient to be coded can be easily selected because a predetermined number of coefficients having the largest absolute value of the coefficient value are selected. . In addition, the DCT transform coefficient generally has a property that both the coefficient located near the coefficient having a high absolute value and the absolute value of the coefficient value are high. , The DCT transform coefficient sequence can be more efficiently converted into a predetermined bit length.

Further, according to the audio compression encoding method of the present invention (claim 5), in the audio compression encoding method according to claim 2, the coefficient sequence conversion step includes a plurality of steps obtained by the discrete cosine conversion step. When a predetermined number of coefficients are selected from the discrete cosine transform coefficients, a coefficient having a coefficient value exceeding a preset threshold is selected, so that a coefficient to be coded can be easily selected.

According to the voice compression encoding method of the present invention (claim 6), in the voice compression encoding method according to claim 1, the coefficient sequence conversion step includes the step of converting the discrete cosine transform coefficient sequence to a predetermined bit length. When transforming to a coefficient, a coefficient having a coefficient value exceeding a predetermined threshold is selected from a plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, and the position of the selected coefficient and the coefficient value are determined by a predetermined bit length. , The discrete cosine transform coefficient sequence is converted to a predetermined bit length, so that the sound quality of the reproduced sound can be improved without increasing the number of bits of the converted DCT transform coefficient sequence. The coefficients to be coded can be easily selected.

According to the audio compression encoding method of the present invention (claim 7), in the audio compression encoding method according to claim 1, the coefficient sequence conversion step includes the step of converting the discrete cosine transform coefficient sequence to a predetermined bit length. When converting to a discrete cosine transform, one coefficient is selected from the plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, and the position of the selected coefficient and a predetermined number of coefficients located before and after the selected coefficient are selected. After the coefficient value of the coefficient is encoded to a predetermined bit length, one coefficient is selected from a plurality of discrete cosine transform coefficients excluding the encoded coefficient, and the position and coefficient value of the selected coefficient and the selected coefficient are selected. The discrete cosine transform coefficient sequence is converted to a predetermined bit length by executing a process of encoding a predetermined number of coefficient values of a predetermined number of coefficients positioned before and after the predetermined number of bits into a predetermined bit length. For coefficients located before and after the selected coefficients eliminates the need to encode the position information, it is possible to perform encoding for more coefficients. Therefore, the sound quality of the reproduced sound can be improved without increasing the number of bits of the converted DCT coefficient sequence.
Further, since the already encoded DCT transform coefficients are not re-encoded, the encoding process can be executed more efficiently.

Further, according to the audio compression encoding method of the present invention (claim 8), in the audio compression encoding method according to claim 7, the coefficient sequence conversion step comprises a plurality of steps obtained by the discrete cosine conversion step. When selecting one coefficient from the discrete cosine transform coefficients, the coefficient having the largest absolute value of the coefficient value is selected, and when selecting one coefficient from a plurality of discrete cosine transform coefficients excluding the coded coefficients, Since the coefficient having the largest absolute value of the coefficient value is selected, the coefficient to be coded can be easily selected. Also, DCT transform coefficients generally have a property that both the coefficient located near a coefficient having a high absolute value and the absolute value of the coefficient value are high. By making a selection, the DCT transform coefficient sequence can be more efficiently converted to a predetermined bit length.

Further, according to the audio compression encoding method of the present invention (claim 9), in the audio compression encoding method according to claim 1, the coefficient sequence conversion step includes converting the discrete cosine transform coefficient sequence to a predetermined bit length. When transforming to a discrete cosine transform step, a predetermined number of coefficients are selected from a plurality of discrete cosine transform coefficients obtained in the discrete cosine transform step, and a coefficient value of the selected coefficient and a predetermined number of coefficients located before and after the selected coefficient are selected. After obtaining the pattern of the first ratio with the coefficient value of the coefficient of
The second ratio pattern having the property closest to the first ratio pattern is selected from the correspondence information between the ratio pattern and the code, and the first pattern is selected using the code corresponding to the selected second ratio pattern. In order to convert a discrete cosine transform coefficient sequence into a predetermined bit length by encoding a pattern having a ratio of, it is not necessary to encode position information for coefficients located before and after the selected coefficient. Coding can be performed on the coefficients. Therefore, the sound quality of the reproduced sound can be improved without increasing the number of bits of the converted DCT coefficient sequence.

According to the audio compression encoding method of the present invention (claim 10), in the audio compression encoding method according to claim 9, the coefficient sequence conversion step is performed when a pattern having the second ratio is selected. First, in order to select the second ratio pattern having the smallest Euclidean distance from the first ratio pattern, the first ratio pattern is selected.
The pattern of the second ratio having the property closest to the pattern of the ratio can be easily selected.

According to the audio compression encoding method of the present invention (claim 11), in the audio compression encoding method according to claim 10, the correspondence between a plurality of second ratio patterns and codes prepared in advance is provided. The relation information is managed by a multi-dimensional data structure, and when the coefficient sequence conversion step selects a pattern of the second ratio, the Euclidean distance from the pattern of the first ratio is determined using nearest point search. Since the pattern with the smallest second ratio is selected, the stored ratio pattern can be efficiently managed, and the search time can be reduced. Further, when selecting the pattern of the second ratio, since the closest point search is used to select the pattern of the second ratio having the smallest Euclidean distance from the pattern of the first ratio,
The pattern of the second ratio having the property closest to the pattern of the first ratio can be easily selected.

Further, according to the audio compression encoding method of the present invention, the coefficient sequence conversion step is performed when the second ratio pattern is selected. Since the second ratio pattern having the smallest norm with respect to the first ratio pattern is selected, the second ratio pattern having properties closest to the first ratio pattern can be easily selected. .

Further, according to the speech compression encoding apparatus of the present invention (claim 13), the noise source extracting means uses the extracted spectral envelope information, subframe, pitch information and gain information to obtain a second error signal. , A discrete cosine transform unit for performing a discrete cosine transform of the extracted secondary error signal, and a coefficient for converting a discrete cosine transform coefficient sequence obtained by the discrete cosine transform unit into a predetermined bit length By including the column conversion means, the noise source information is extracted and encoded, so that the amount of calculation and the amount of memory can be reduced in the CELP encoding process. That is, according to the speech compression encoding apparatus of the present invention, in the quantization of the second-order error signal, the codebook is not provided and the codebook search is not performed by using the filter calculation. A high-quality and low-bit-rate audio compression / encoding device that can be realized can be obtained.

According to the audio compression coding apparatus of the present invention, the coefficient sequence conversion means converts the discrete cosine transform coefficient sequence to a predetermined bit length. In the conversion into a plurality of discrete cosine transform coefficients, a predetermined number of coefficients are selected from the plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means. In order to convert a discrete cosine transform coefficient sequence to a predetermined bit length by encoding a predetermined number of coefficient values and a coefficient value to a predetermined bit length, position information is encoded for coefficients located before and after the selected coefficient. Therefore, it is possible to encode more coefficients. Therefore, the converted D
The sound quality of the reproduced sound can be improved without increasing the number of bits of the CT conversion coefficient sequence.

Further, according to the audio compression coding apparatus of the present invention, the coefficient sequence conversion means may include a plurality of coefficient sequences obtained by the discrete cosine conversion means. When a predetermined number of coefficients are selected from the discrete cosine transform coefficients, the coefficient having the maximum absolute value of the coefficient value is selected, so that the coefficient to be coded can be easily selected.

Further, according to the speech compression encoding apparatus of the present invention, the coefficient sequence conversion means may include a plurality of coefficient sequences obtained by the discrete cosine transformation means. When selecting a predetermined number of coefficients from the discrete cosine transform coefficients, the coefficient to be coded can be easily selected because a predetermined number of coefficients having the largest absolute value of the coefficient value are selected. . In addition, the DCT transform coefficient generally has a property that both the coefficient located near the coefficient having a high absolute value and the absolute value of the coefficient value are high. , The DCT transform coefficient sequence can be more efficiently converted into a predetermined bit length.

Further, according to the speech compression encoding apparatus of the present invention, the coefficient sequence conversion means may include a plurality of coefficient sequences obtained by the discrete cosine transformation means. When a predetermined number of coefficients are selected from the discrete cosine transform coefficients, a coefficient having a coefficient value exceeding a preset threshold is selected, so that a coefficient to be coded can be easily selected.

Further, according to the audio compression encoding apparatus of the present invention, the coefficient sequence conversion means converts the discrete cosine transform coefficient sequence to a predetermined bit length. In the conversion to a coefficient, after selecting a coefficient having a coefficient value exceeding a predetermined threshold from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means, the position of the selected coefficient and the coefficient value are converted into predetermined bits. Since the discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding to a long length, the sound quality of the reproduced sound can be improved without increasing the number of bits of the converted DCT transform coefficient sequence. , Coefficients to be encoded can be easily selected.

According to the audio compression encoding apparatus of the present invention, the coefficient sequence converting means converts the discrete cosine transform coefficient sequence to a predetermined bit length. When transforming to a coefficient, one coefficient is selected from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means, and the position of the selected coefficient and a predetermined number of coefficients located before and after the coefficient are selected. After the coefficient value of the coefficient is encoded to a predetermined bit length, one coefficient is selected from a plurality of discrete cosine transform coefficients excluding the encoded coefficient, and the position and coefficient value of the selected coefficient and the selected coefficient are selected. The discrete cosine transform coefficient sequence is converted into a predetermined bit length by executing a process of coding a predetermined number of coefficient values of a predetermined number of coefficients positioned before and after the predetermined number of times into a predetermined bit length. Therefore, the coefficients located before and after the selected coefficients eliminates the need to encode the position information,
Encoding can be performed for more coefficients. Therefore, the sound quality of the reproduced sound can be improved without increasing the number of bits of the converted DCT coefficient sequence. Further, since the already encoded DCT transform coefficients are not re-encoded, the encoding process can be executed more efficiently.

Further, according to the audio compression encoding apparatus of the present invention, the coefficient sequence converting means may include a plurality of coefficient sequences obtained by the discrete cosine converting means. When selecting one coefficient from the discrete cosine transform coefficients, the coefficient having the largest absolute value of the coefficient value is selected, and when selecting one coefficient from a plurality of discrete cosine transform coefficients excluding the coded coefficients, Since the coefficient having the largest absolute value of the coefficient value is selected, the coefficient to be coded can be easily selected. Further, DCT transform coefficients generally have a property that both the coefficient located in the vicinity of the coefficient having a high absolute value of the coefficient value and the absolute value of the coefficient value are high.
By taking advantage of this property and selecting from the coefficient having the largest absolute value, the DCT transform coefficient sequence can be more efficiently converted to a predetermined bit length.

According to the audio compression encoding apparatus of the present invention, the coefficient sequence conversion means converts the discrete cosine transform coefficient sequence to a predetermined bit length. When transforming to a discrete cosine transform means, a predetermined number of coefficients are selected from a plurality of discrete cosine transform coefficients obtained by the discrete cosine transform means, and a coefficient value of the selected coefficient and a predetermined number of coefficients located before and after the selected coefficient are selected. After obtaining the pattern of the first ratio with the coefficient value of the coefficient of the second ratio, from the correspondence information between the plurality of patterns of the second ratio and the code prepared in advance, the second pattern having the property closest to the pattern of the first ratio To convert a discrete cosine transform coefficient sequence into a predetermined bit length by selecting a pattern having a ratio of the first ratio and encoding the pattern having the first ratio using a code corresponding to the selected second ratio pattern. ,
It is not necessary to encode position information for coefficients located before and after the selected coefficient, and it is possible to encode more coefficients. Therefore, the transformed DCT
The sound quality of the reproduced sound can be improved without increasing the number of bits of the conversion coefficient sequence.

According to the speech compression encoding apparatus of the present invention, the coefficient sequence conversion means may select a pattern having the second ratio. In addition, since the second ratio pattern having the smallest Euclidean distance from the first ratio pattern is selected, it is possible to easily select the second ratio pattern having the property closest to the first ratio pattern. it can.

According to the speech compression encoding apparatus of the present invention (claim 23), in the speech compression encoding apparatus according to claim 22, correspondence between a plurality of second ratio patterns and codes prepared in advance. The relation information is managed by a multi-dimensional data structure, and when the coefficient sequence conversion means selects the pattern of the second ratio, the nearest Euclidean distance from the pattern of the first ratio is determined using nearest point search. Since the pattern having the small second ratio is selected, the stored ratio pattern can be efficiently managed, and the search time can be reduced. Further, when selecting the pattern of the second ratio, since the closest point search is used to select the pattern of the second ratio having the smallest Euclidean distance from the pattern of the first ratio, the first ratio is selected.
The pattern of the second ratio having the property closest to the pattern of the ratio can be easily selected.

Further, according to the audio compression encoding apparatus of the present invention, in the audio compression encoding apparatus according to the present invention, when the coefficient sequence converting means selects the pattern of the second ratio. Since the second ratio pattern having the smallest norm with respect to the first ratio pattern is selected, the second ratio pattern having properties closest to the first ratio pattern can be easily selected. .

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of an audio compression encoding device according to a first embodiment.

FIG. 2 is a block diagram of a speech encoding unit according to the first embodiment.

FIG. 3 is a schematic block diagram of a noise source extraction unit according to the first embodiment.

FIG. 4 is a block diagram illustrating a partial configuration of a speech decoding unit according to the first embodiment.

FIG. 5 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the second embodiment.

FIG. 6 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the third embodiment.

FIG. 7 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the fourth embodiment.

FIG. 8 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the fifth embodiment.

FIG. 9 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the sixth embodiment.

FIG. 10 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the seventh embodiment.

FIG. 11 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the eighth embodiment.

FIG. 12 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the ninth embodiment;

FIG. 13 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the tenth embodiment.

FIG. 14 is a flowchart illustrating an operation procedure of a coefficient conversion unit according to the twelfth embodiment.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 audio compression encoding apparatus 101 A / D conversion section 102 audio encoding section 103 storage section 104 audio decoding section 105 D / A conversion section 201 frame configuration section 202 spectrum envelope extraction section 203 subframe configuration section 204 pitch information extraction section 205 Gain extraction unit 206 Noise source extraction unit 301 Secondary error signal configuration unit 302 DCT conversion unit 303 Coefficient conversion unit 401 Spectrum envelope decoding unit 402 Pitch information decoding unit 403 Noise source decoding unit 404 Gain decoding unit 405 Voice synthesis unit

Claims (24)

[Claims] A first step of inputting an analog audio waveform and converting it into a digital audio waveform; a second step of encoding the digital audio waveform by a predetermined encoding method; A third step of storing an audio waveform, a fourth step of extracting and decoding the stored digital audio waveform, and a fifth step of converting the decoded digital audio waveform to an analog audio waveform. ,
Wherein the second step comprises: a frame dividing step of dividing the digital speech waveform into processing units called frames; and extracting spectrum envelope information representing a spectrum envelope from each of the divided frames. Spectral envelope information extracting step of performing coding and encoding, subframe dividing step of dividing each of the divided frames into processing units called subframes, and using the extracted spectral envelope information, A pitch information extracting step of extracting and encoding pitch information from each of them, a gain information extracting step of extracting and encoding gain information from the extracted pitch information, and the spectrum envelope information, subframe, pitch information and gain. Extracts and encodes noise source information, which is sound source information, from information A sound source information extracting step, wherein the fourth step is a restoring step of restoring the encoded spectrum envelope information, pitch information, gain information and noise source information, and the restored pitch information, gain information and An excitation source signal generation step of generating an excitation source signal from noise source information; and a synthesized speech generation step of generating a synthesized speech from the excitation source signal and the restored spectrum envelope information. An extraction step of extracting a secondary error signal using the extracted spectral envelope information, subframe, pitch information and gain information, and a discrete cosine transform of the extracted secondary error signal A discrete cosine transform step, and a coefficient sequence transforming step of transforming the discrete cosine transform coefficient sequence obtained in the discrete cosine transform step into a predetermined bit length. By,
A speech compression encoding method, wherein the noise source information is extracted and encoded. 2. The voice compression encoding method according to claim 1, wherein the coefficient sequence conversion step is performed by converting the discrete cosine transform coefficient sequence into a predetermined bit length. After selecting a predetermined number of coefficients from the plurality of discrete cosine transform coefficients, the position and the coefficient value of the selected coefficient and the coefficient values of the predetermined number of coefficients located before and after the selected coefficient are determined by a predetermined number. A speech compression encoding method characterized in that said discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding to a bit length. 3. The audio compression encoding method according to claim 2, wherein said coefficient sequence converting step selects a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained in said discrete cosine transform step. And selecting a coefficient having the largest absolute value of the coefficient value. 4. The voice compression encoding method according to claim 2, wherein said coefficient sequence converting step selects a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained in said discrete cosine transform step. And a method of selecting a predetermined number of coefficients from those having the largest absolute value of the coefficient values. 5. A speech compression encoding method according to claim 2, wherein said coefficient sequence converting step selects a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained in said discrete cosine transform step. And selecting a coefficient having a coefficient value exceeding a preset threshold value. 6. The audio compression encoding method according to claim 1, wherein the coefficient sequence conversion step is performed by converting the discrete cosine transform coefficient sequence into a predetermined bit length. Selecting a coefficient having a coefficient value exceeding a predetermined threshold from a plurality of discrete cosine transform coefficients;
A speech compression encoding method characterized in that the discrete cosine transform coefficient sequence is converted into a predetermined bit length by coding the selected coefficient position and the coefficient value into a predetermined bit length. 7. The audio compression encoding method according to claim 1, wherein the coefficient sequence conversion step is performed by converting the discrete cosine transform coefficient sequence into a predetermined bit length. One coefficient is selected from a plurality of discrete cosine transform coefficients, and the position and coefficient value of the selected coefficient and the coefficient values of a predetermined number of coefficients positioned before and after the selected coefficient are encoded into a predetermined bit length. After that, one coefficient is selected from the plurality of discrete cosine transform coefficients excluding the coded coefficient, and a position and a coefficient value of the selected coefficient and a predetermined coefficient located before and after the selected coefficient are selected. By performing the step of encoding the coefficient value of the number coefficient and the coefficient value to a predetermined bit length a preset number of times,
A speech compression encoding method, wherein the discrete cosine transform coefficient sequence is converted into a predetermined bit length. 8. A speech compression encoding method according to claim 7, wherein said coefficient sequence converting step includes selecting one coefficient from a plurality of discrete cosine transform coefficients obtained in said discrete cosine transform step. Selecting the coefficient having the largest absolute value of the coefficient when selecting the coefficient having the largest absolute value of the numerical value and selecting one coefficient from the plurality of discrete cosine transform coefficients excluding the encoded coefficient; A voice compression encoding method characterized by the following. 9. The audio compression encoding method according to claim 1, wherein the coefficient sequence conversion step is performed by converting the discrete cosine transform coefficient sequence into a predetermined bit length. A predetermined number of coefficients are selected from a plurality of discrete cosine transform coefficients, and a first ratio of a coefficient value of the selected coefficient to a coefficient value of a predetermined number of coefficients located before and after the selected coefficient is selected. After obtaining the pattern, a pattern of the second ratio having a property closest to the pattern of the first ratio is selected from correspondence information between a plurality of patterns of the second ratio and codes prepared in advance, and And encoding the pattern of the first ratio using a code corresponding to the pattern of the second ratio to convert the discrete cosine transform coefficient sequence into a predetermined bit length. Method. 10. The voice compression encoding method according to claim 9, wherein the coefficient sequence conversion step comprises selecting the second ratio pattern such that the Euclidean distance from the first ratio pattern is the smallest. A voice compression encoding method characterized by selecting a pattern having a small second ratio. 11. The audio compression encoding method according to claim 10, wherein the correspondence information between the plurality of second ratio patterns and the codes prepared in advance is managed by a multidimensional data structure, and The column conversion step uses the nearest point search to select the second ratio pattern.
And selecting the second ratio pattern having the smallest Euclidean distance from the ratio pattern. 12. The voice compression encoding method according to claim 9, wherein the coefficient sequence conversion step has a minimum norm with the first ratio pattern when selecting the second ratio pattern. A speech compression encoding method, wherein the pattern having the second ratio is selected. 13. A / D conversion means for inputting an analog voice waveform and converting it into a digital voice waveform, voice coding means for coding the digital voice waveform by a predetermined coding method, Storage means for storing the digital voice waveform; voice decoding means for extracting and decoding the stored digital voice waveform; and D for converting the decoded digital voice waveform into an analog voice waveform.
/ A conversion means, wherein the speech encoding means divides the digital speech waveform into processing units called frames, and obtains a spectral envelope from each of the divided frames. Using spectral envelope information extracting means for extracting and encoding the represented spectral envelope information, subframe dividing means for dividing each of the divided frames into processing units called subframes, and using the extracted spectral envelope information, Pitch information extracting means for extracting and encoding pitch information from each of the divided subframes; gain information extracting means for extracting and encoding gain information from the extracted pitch information; Noise source information as sound source information from frame, pitch information and gain information Noise source information extracting means for extracting and encoding, wherein the speech decoding means restores the encoded spectrum envelope information, pitch information, gain information, and noise source information; Excitation source signal generation means for generating an excitation source signal from the obtained pitch information, gain information, and noise source information; and synthesized speech generation means for generating a synthesized speech from the excitation source signal and the restored spectral envelope information. The noise source extracting means for extracting a secondary error signal using the extracted spectral envelope information, subframe, pitch information, and gain information; A discrete cosine transform unit for performing a discrete cosine transform on the error signal; and a discrete cosine transform coefficient sequence obtained by the discrete cosine transform unit, which is transformed into a predetermined bit length. By including a coefficient sequence converting means, the speech compression encoding apparatus characterized by encoding by extracting the noise source information. 14. The audio compression encoding apparatus according to claim 13, wherein said coefficient sequence converting means obtains said discrete cosine transform coefficient sequence when converting said discrete cosine transform coefficient sequence into a predetermined bit length. After selecting a predetermined number of coefficients from the plurality of discrete cosine transform coefficients, the position and the coefficient value of the selected coefficient and the coefficient values of the predetermined number of coefficients located before and after the selected coefficient are determined by a predetermined number. An audio compression encoding apparatus characterized in that the discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding to a bit length. 15. A speech compression encoding apparatus according to claim 14, wherein said coefficient sequence converting means selects a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained by said discrete cosine transform means. An audio compression encoding apparatus characterized by selecting a coefficient having a maximum absolute value of the coefficient value. 16. A speech compression encoding apparatus according to claim 14, wherein said coefficient sequence converting means selects a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained by said discrete cosine transform means. An audio compression encoding apparatus, wherein a predetermined number of coefficients are selected from those having the largest absolute value of coefficient values. 17. A speech compression encoding apparatus according to claim 14, wherein said coefficient sequence converting means selects a predetermined number of coefficients from a plurality of discrete cosine transform coefficients obtained by said discrete cosine transform means. And a coefficient having a coefficient value exceeding a preset threshold value is selected. 18. The audio compression encoding apparatus according to claim 13, wherein said coefficient sequence converting means obtains said discrete cosine transform coefficient sequence when converting said discrete cosine transform coefficient sequence into a predetermined bit length. After selecting a coefficient having a coefficient value exceeding a predetermined threshold value from a plurality of discrete cosine transform coefficients, the position and the coefficient value of the selected coefficient are encoded into a predetermined bit length, thereby obtaining the discrete cosine transform coefficient. An audio compression encoding device for converting a sequence into a predetermined bit length. 19. The audio compression encoding apparatus according to claim 13, wherein said coefficient sequence converting means obtains said discrete cosine transform coefficient sequence when converting said discrete cosine transform coefficient sequence into a predetermined bit length. One coefficient is selected from a plurality of discrete cosine transform coefficients, and the position and coefficient value of the selected coefficient and the coefficient values of a predetermined number of coefficients positioned before and after the selected coefficient are encoded into a predetermined bit length. After that, one coefficient is selected from the plurality of discrete cosine transform coefficients excluding the coded coefficient, and a position and a coefficient value of the selected coefficient and a predetermined coefficient located before and after the selected coefficient are selected. Converting the discrete cosine transform coefficient sequence into a predetermined bit length by executing a step of encoding a coefficient value of a number coefficient and a predetermined bit length to a predetermined bit length. Audio compression encoding device. 20. A speech compression encoding apparatus according to claim 19, wherein said coefficient sequence converting means selects one coefficient from a plurality of discrete cosine transform coefficients obtained by said discrete cosine transform means. Selecting the coefficient having the largest absolute value of the coefficient when selecting the coefficient having the largest absolute value of the numerical value and selecting one coefficient from the plurality of discrete cosine transform coefficients excluding the encoded coefficient; An audio compression encoding device characterized by the following. 21. The audio compression encoding apparatus according to claim 13, wherein the coefficient sequence converting means obtains the discrete cosine transform coefficient sequence when converting the discrete cosine transform coefficient sequence into a predetermined bit length. A predetermined number of coefficients are selected from a plurality of discrete cosine transform coefficients, and a first ratio of a coefficient value of the selected coefficient to a coefficient value of a predetermined number of coefficients located before and after the selected coefficient is selected. After finding the pattern,
A pattern of the second ratio having the property closest to the pattern of the first ratio is selected from correspondence information between a plurality of patterns of the second ratio and codes prepared in advance, and a pattern of the selected second ratio is selected. An audio compression encoding device, wherein the discrete cosine transform coefficient sequence is converted into a predetermined bit length by encoding the pattern of the first ratio using a code corresponding to the pattern. 22. The audio compression encoding apparatus according to claim 21, wherein the coefficient sequence converting means selects the second ratio pattern such that the Euclidean distance from the first ratio pattern is the shortest. An audio compression encoding apparatus, wherein a pattern having a small second ratio is selected. 23. The audio compression encoding apparatus according to claim 22, wherein the correspondence information between the plurality of second ratio patterns and codes prepared in advance is managed by a multidimensional data structure, and When the column conversion means selects the pattern of the second ratio, the first conversion is performed by using a nearest point search.
A second ratio pattern having the smallest Euclidean distance from the second ratio pattern. 24. The audio compression encoding apparatus according to claim 21, wherein the coefficient sequence conversion means has a minimum norm with the first ratio pattern when selecting the second ratio pattern. An audio compression encoding apparatus, wherein the pattern having the second ratio is selected.