JPH0981194A - Method and device for voice coding and decoding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method of voice coding by which a high-quality voice signal can be reproduced with a small amount of operation even at a low coding bit rate below about 4kbps. SOLUTION: The voice coding method which supplies the drive signals generated by using the drive vectors obtained from a drive vector coding book to a weighted synthesizing filter 113, thereby generates synthesized voice signals, searches the drive vector that minimizes the distortion of synthesized voice signals and outputs coding parameters representing the information on the above drive vector and the filter factor of the synthesizing filter 113 has plural adaptive coding books 110, 120, a noise coding book 121 and a signal classifier 130. When the nature of input voice signals is classified as to be periodical, or periodical and stationary by the signal classifier 130, the drive vectors obtained from the adaptive coding books 110, 120 are used to generate drive signals. Otherwise, the drive vectors obtained from the noise coding book 121 are used to generate drive signals.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice encoding / decoding method and a voice encoding / decoding device capable of reproducing high quality voice at a low bit rate.

[0002]

2. Description of the Related Art C is an effective encoding method for encoding telephone band voice at a transmission rate of about 4 kbps.
ELP (Code Excited Linear Prediction) is known. This method is roughly divided into a process of obtaining a voice synthesis filter that models a vocal tract from input voice divided into frame units, and a process of obtaining a drive vector that is a drive signal corresponding to the input signal of this filter. Of these processes, the latter passes a plurality of drive vectors stored in the codebook one by one through a speech synthesis filter, calculates the distortion of the obtained synthesized speech signal vector, that is, the error with respect to the input speech signal. Is the process of searching for a drive vector that minimizes. This process is called a closed loop search, and
This is a very effective method for reproducing good sound quality even at a coding bit rate of about kbps.

However, the coding bit rate is 4 kbps.
When the value is lower than the above level, there is a problem in that this method cannot reproduce sound of sufficient quality. In addition, a large amount of calculation is required for the closed loop search, which is also a problem of this method. Hereinafter, these problems will be specifically described.

[0004] Regarding the CELP system, MRSchroeder
and BSAtal, “Code Excited Linear Prediction (CEL
P): High Quality Speech at Very Low Bit Rates ”,
Proc.ICASSP, pp.937-940,1985 and WSKleijin, DJ
Krasinski et al. “Improved Speech Quality and Effic
ient Vector Quantization in SELP ”, Proc.ICASSP, p
This is detailed in p.155-158,1988.

The outline of the CELP method will be described with reference to FIG. An audio signal in frame units is input to the input terminal 460. This input speech signal is a linear prediction analysis unit 45.
0, the filter coefficient of the weighted synthesis filter 430 is obtained, and at the same time, the weighted input audio signal is input to the perceptual weighting section 440. The zero-state response of weighted synthesis filter 430 is subtracted from this weighted input speech signal to generate target vector 480.

Next, the driving vectors are read out one by one from the adaptive codebook 411, further multiplied by the gain in the gain circuit 421, and then inputted to the weighted synthesizing filter 430 as a driving signal. A signal vector is generated. The distortion of the synthesized speech signal vector, that is, the difference from the target vector 480 is evaluated by the evaluation unit 470, the drive vector is searched from the adaptive codebook 411 so that this distortion becomes smaller, and the optimum one is the first drive. It is assumed to be a vector. Next, a second drive vector is similarly searched from the noise codebook 412 in consideration of the effect of the first drive vector. Finally, the first and second drive vectors are respectively multiplied by optimum gains in the gain circuits 421 and 422 to generate a drive signal. The contents of the adaptive codebook 411 are updated by this drive signal, and the update is prepared for the input of the audio signal of the next frame.

As described above, in the CELP method, the weighted synthesis filter 430 is associated with all the drive vectors stored in the adaptive codebook 411 and the noise codebook 412.
A closed loop search is performed to search for a drive vector that minimizes this distortion by performing filtering processing and distortion calculation of the synthesized speech signal. According to this CELP method, 8k
At a coding bit rate of about bps, a voice signal of relatively good quality can be reproduced, but at a low coding bit rate of about 4 kbps or less, the adaptive codebook 4
11 and the random codebook 412 are updated with a longer cycle and the bit allocation is reduced, which causes a problem that the quality of the reproduced voice signal is deteriorated.

To explain this problem in more detail, in the configuration of FIG. 7, the drive signal sources for driving the weighted synthesis filter 430 are the first-stage adaptive codebooks 411 and 2.
The noise codebook 412 in the second stage has a two-stage configuration, and when the input speech signal is periodic, the adaptive codebook 41
1 represents the periodic component of the drive signal, and the noise codebook 412 represents the residual component of the drive signal which could not be represented by the adaptive codebook 411 alone. The book 412 is configured to divide roles.

With this configuration, the coding bit rate is 8 kb.
When it was about ps, it worked well and a reproduced voice signal of high quality was obtained. However, at a low coding bit rate of about 4 kbps or less, the period (generally called a subframe) for updating the contents of the adaptive codebook 411 and the noise codebook 412 becomes long, and the number of bits allocated becomes small. The adaptive codebook 411 cannot express the periodic component of the drive signal with sufficient accuracy. As a result, a periodic component remains in the residual component of the drive signal that cannot be expressed by the adaptive codebook 411, and this cannot be expressed by the noise codebook 412, so that the quality of the reproduced voice deteriorates.

Further, in the CELP method, since a closed loop search is executed, there is a problem that enormous calculation is required. To solve this problem of computational complexity, the method of reducing the computational complexity required for the closed loop search of the drive vector is RCRose and TPBa.
Invented by rnwell III. This method is SEV
It is called (Self Excited Vocoder), and its operation and performance are described in RCRose and TPBarnwell III: “Qu
ality Comparison of Low Complexity 4800 bps Self Ex
cited and Code Excited Vocoders ”, Proc.ofICASSP'8
7, pp.1637-1640,1987. Below, this S
The EV system will be briefly described.

FIG. 8 is a block diagram of a drive signal searching section in the SEV type speech encoding apparatus. As can be seen from the comparison with FIG. 7, the major difference between the CELP method and the SEV method lies in the method of structuring the drive signal of the synthesis filter. CEL
In the P method, the drive signal is configured using the adaptive codebook and the noise codebook, whereas in the SEV method, only one or two adaptive codebooks (referred to as Long-term predictor in the literature) 500 and 510 are used. The drive signal is configured by using a simple structure, and the special code structure of the adaptive codebook is used to reduce the calculation amount. That is, each drive vector in adaptive codebooks 500 and 510 is configured by shifting from the past drive signal by a delay corresponding to the pitch period, and as a result, each drive vector has a configuration in which its elements overlap. By utilizing that
The calculation amount is reduced to about 1/20 by recursively calculating the output of the weighted synthesis filter 520 for the drive vector.

However, in the SEV method, since the drive signal is always created from the past history by using the adaptive codebook regardless of the nature of the input voice signal, the input voice signal is different from the past signal like the unvoiced section. Generates an appropriate drive signal in a portion where the characteristics of the voice change, such as when there is little correlation, the rising or falling portion of the signal, the transient portion from the unvoiced section to the voiced section, or the transient section from the voiced section to the unvoiced section. However, there is a problem that the sound quality is further deteriorated as compared with the CELP method.

[0013]

As described above, the CELP method, which is one of the conventional audio encoding methods, cannot reproduce a high quality audio signal at a low encoding bit rate of about 4 kbps or less. There is a problem that the amount of calculation is large, and the amount of calculation is reduced in the SEV method, but there is a problem that the quality of the reproduced audio signal is further deteriorated as compared with the SELP method.

An object of the present invention is to provide a voice encoding / decoding method and a voice encoding / decoding device capable of reproducing a high quality voice signal with a small amount of calculation even at a low encoding bit rate of about 4 kbps or less. To provide.

[0015]

In order to solve the above problems, the present invention generates a drive signal using a drive vector obtained from a plurality of drive vector codebooks, and uses this drive signal as an analysis result of an input voice signal. It supplies to the synthesis filter whose filter coefficient is determined based on, and searches the drive vector codebook for a drive vector that minimizes the distortion of the synthesized voice signal output from this synthesis filter, and at least the drive vector and the synthesis filter In a speech coding method for outputting a coding parameter representing information on a filter coefficient,
At least two adaptive codebooks and at least one noise codebook are prepared as drive vector codebooks, and if the input speech signal is periodic or periodic and stationary, at least two adaptive codebooks are selected. Generate a drive signal using the obtained drive vector, and generate a drive signal using at least the drive vector obtained from the noise codebook if the input speech signal is aperiodic or periodic and nonstationary in nature. It is characterized by

The speech coding apparatus according to the present invention has at least two adaptive codebooks and at least one noise codebook prepared as drive vector codebooks, and whether the nature of the input speech signal is periodic or not. And a driving vector obtained from at least two adaptive codebooks when the characteristic of the input speech signal is classified as periodic or periodic and stationary by the classifying means. Drive signal is generated by using, and when the property of the input speech signal is classified as aperiodic or periodic and nonstationary, a drive signal is generated using at least a drive vector obtained from the noise codebook. And a signal generating means.

As described above, in the present invention, the drive signal of the synthesis filter is formed by the linear combination of the drive vectors stored in the multi-stage drive vector codebook. Then, the property of the input speech signal is classified according to whether it is periodic or periodic and stationary, and when classified as periodic or periodic and stationary, at least two-stage adaptive codebook is used. Drive signal.

As described above, in the conventional CELP system, when the input voice signal is periodic, the cyclic code component of the drive signal is expressed by the first-stage adaptive codebook and can be expressed only by this adaptive codebook. Since the residual component of the drive signal that has not existed is represented by the noise codebook in the second stage, the cycle for updating the code of the adaptive codebook or the noise codebook is long at a low coding bit rate of about 4 kbps or less. And the number of allocated bits decreases. Therefore, the adaptive codebook cannot express the periodic signal with sufficient accuracy, and the periodic component remains in the residual component of the adaptive codebook.
Due to the fact that the noise codebook cannot express the remaining periodic component, the quality of the reproduced voice signal is deteriorated.

On the other hand, in the present invention, when the input voice signal is periodic or periodic and stationary, after the periodic adaptive component of the drive signal is expressed by the first adaptive codebook in the first stage. Even if a periodic component remains in the residual of the first stage due to the reduction of the encoding bit rate, this periodic component can be represented by the second adaptive codebook of the second stage. Therefore, it is possible to obtain a reproduced audio signal of high quality even at a low coding bit rate.

On the other hand, when the input voice signal is aperiodic or periodic and non-stationary, the present invention uses at least a noise codebook and at most one adaptive codebook to generate the drive signal, Is not periodic, or is periodic but not stationary, and has a small correlation with the past drive signal, the adaptive codebook that uses the correlation with the past drive signal is not heavily used and noise is reduced. Since the codebook can be used, a good reproduced voice signal can be obtained.

As described above, according to the present invention, the drive signal is composed of a plurality of stages of adaptive codebooks or noise codebooks depending on whether the input speech signal is periodic or periodic and stationary. By switching the configuration, it is possible to generate a drive signal suitable for the property of the input audio signal, and thus it is possible to obtain a reproduced audio signal of good quality even at a low coding bit rate.

Further, in the speech decoding method according to the present invention, in order to decode the original speech signal from the coding parameters obtained by the above speech coding method, at least the driving vector and the filter coefficient of the synthesis filter are respectively set. A drive signal is generated by using a drive vector obtained from a plurality of stages of drive vector codebooks with the encoding parameter to be input as an input, and the drive signal is supplied to a synthesis filter whose filter coefficient is determined based on the encoding parameter. Accordingly, in the speech decoding method for decoding a speech signal, at least two adaptive codebooks and at least one noise codebook are prepared as the driving vector codebook, and at least two selected based on the coding parameters. Drive vectors obtained from the adaptive codebook, or at least obtained from the noise codebook And generating a driving signal by using the drive vector.

The speech decoding apparatus according to the present invention encodes at least the drive vector and the filter coefficient of the synthesis filter in order to decode the original speech signal from the coding parameters obtained by the speech coding apparatus described above. By inputting the encoding parameter and generating a drive signal using the drive vector obtained from the drive vector codebook of multiple stages, and supplying this drive signal to the synthesis filter whose filter coefficient is determined based on the encoding parameter, In a voice decoding device for decoding a voice signal, at least two adaptive codebooks and at least one noise codebook prepared as drive vector codebooks, and at least two adaptive codebooks selected based on coding parameters. Drive vector obtained from or at least the drive vector obtained from the random codebook. And having a drive signal generating means for generating a drive signal using a vector.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION

(First Embodiment) FIG. 1 is a block diagram showing the configuration of a speech coder according to a first embodiment of the present invention, in which a speech signal input terminal 100, a buffer 101, an LPC analysis section 102, and an LSP are provided. The quantizer 103, the subframe division circuit 104, the weighting filter 105, the first adaptive codebook 110 that constitutes the drive signal source of the first stage, and the gain circuit 11
1, adder 112, weighted synthesis filter 113, subtractor 114, second adaptive codebook 120 and noise codebook 121 that constitute the second stage drive signal source, switch 122, gain circuit 123, gain codebook 124, signal classifier 13
0, codebooks 110, 120, 121 and 124, search units 140, 141 and 142, a multiplexer 150, and a coding parameter output terminal 151.

Here, the signal classifier 130 classifies the characteristics of the input speech signal according to whether it is periodic or periodic and stationary, and outputs a classification determination signal. Further, the switcher 122 uses the second adaptive codebook 1 as the drive signal source of the second stage based on the classification determination signal from the signal classifier 130.
This is for selecting and extracting either the 20 or the random codebook 121.

Next, the operation of the speech coding apparatus of this embodiment will be described. First, a digitized audio signal is input from the input terminal 100, divided into sections of a constant interval (for example, 20 ms) called a frame, and the buffer 10
Stored in 1. The input speech signal read from the buffer 101 in units of frames is subjected to linear prediction analysis by the LPC analysis unit 102 in units of frames, and a linear prediction coefficient a _i which is a parameter representing a spectrum envelope of the input speech signal.
(I = 1, ..., P) is calculated and LSP quantizer 103 is calculated.
Is input to

In the LSP quantizer 103, after converting the linear prediction coefficient into an LSP (line spectrum pair) parameter,
Quantize with a predetermined number of bits. The LSP parameter quantized by the LSP quantizer 103 is input to the multiplexer 150, decoded, inversely transformed into a linear prediction coefficient, and quantized linear prediction coefficient aq.
_i (i = 1, ..., P) is obtained. Well-known methods can be used for the method of linear prediction analysis, the method of obtaining LSP parameters, and the method of quantizing LSP parameters. The quantized linear prediction coefficient is the weighting filter 1
05 and the weighted synthesis filter 113.

On the other hand, the input audio signal read from the buffer 101 on a frame-by-frame basis is the sub-frame division circuit 1
Each frame is divided into a plurality of sections called a subframe by 04, and is input to the weighting synthesis filter 105 in units of subframes to be a target vector for driving signal search. Then, this weighting synthesis filter 11
The search processing of the drive signal for driving 3 is performed in the subframe unit as follows.

In the speech coding apparatus according to the present embodiment, the first adaptive codebook 110 is the driving signal source in the first stage, and one of the second adaptive codebook 120 and the noise codebook 121 is in the second stage. It has a two-stage structure as a drive signal source, and the drive vector is sequentially searched from the codebook of each stage. This search procedure will be described with reference to the flowchart shown in FIG.

First, a drive vector is searched for the first adaptive codebook 110 in the first stage. That is, when an audio signal is input from the input terminal 100 (step S1),
Prior to searching the adaptive codebook 110, the weighting filter 1
In 05, the perceptual weighting is applied to the input speech signal, and the target vector X is generated by subtracting the influence of the weighting synthesis filter 105 from the previous subframe (step S2), and the first adaptive codebook 110 is searched ( Step S3). The first adaptive codebook 110 is used to represent the periodic component (pitch) of the speech signal, and the drive vector e (n) stored in this adaptive codebook 110 is expressed by the following equation. It is created by cutting out the past drive signal by the subframe length.

E (n) = e (nL), n = 1, ..., N (1) where L is a lag and N is a subframe length. The search of the first adaptive codebook 110 is performed by the search unit 140, in which the distortion of the synthesized speech signal vector obtained by passing the drive vector e through the weighted synthesis filter 113, that is, the error between the synthesized speech signal vector and the target vector X It is performed by searching for a lag that minimizes R according to a known method. The lag can be in whole or fractional samples.

The weighting filter 105 and the weighting synthesis filter 113 can be constructed by a known method. For example, if W (z) and Hw (z) are their transfer functions, then W (z) = A ( z / r) / A (z / r ') (2) Hw (z) = W (z) / Aq (z) (3) A (z) = 1 + Σa _i z ^-i (4) Aq (z) = It can be expressed as 1 + Σaq _i z ⁻ⁱ (5).

Next, the search unit 14 searches the second stage codebook.
Do by 1. In this case, the signal classifier 130 first classifies the input speech signal to be encoded according to whether it is periodic or whether it is periodic and stationary (step S4).
Specifically, classification is performed by a known voiced / unvoiced determination method, a method that combines pitch period continuity determination with this method, or an autocorrelation method.

FIG. 2 is a diagram showing an example of waveforms of voiced sections and unvoiced sections of a voice signal. As shown in the figure, in the transitional part where the unvoiced section transitions to the voiced section and the transitional part where the voiced section transitions to the unvoiced section are periodic, the pitch period and the amplitude change are not uniform. It is in a state, and is in a periodic and steady state except in the transient part. The signal classifier 130 may perform classification based on whether the input voice signal is periodic or not by judging whether the input voice signal is a voiced section or an unvoiced section. However, the signal classifier 130 further determines whether or not the voiced section is stationary. Therefore, classification may be performed depending on whether it is periodic and stationary.

As a result of the classification in step S4, when the input speech signal is periodic or periodic and stationary, the second adaptive codebook 120 is searched as the second stage codebook (step S5). Otherwise, the random codebook 121 is searched (step S6). The search for the second adaptive codebook 120 is performed in the same manner as the search for the first codebook 110. Also,
The random codebook 121 is searched in the same manner as the conventional CELP method. At this time, known techniques such as overlapping codebook, backward filtering, and preselection can be used to reduce the amount of calculation.

Finally, the gain codebook 124 is searched (step S7). The gain codebook 124 has, as a representative vector, a vector whose element is a gain multiplied by a drive vector searched from the codebooks of the first and second stages. Then, the drive vector searched from the first adaptive codebook 110 in the first stage and the first adaptive codebook 120 or the random codebook 12 in the second stage extracted via the switch 122.
In the drive vector searched from 1, the gain circuit 111,
The gains searched from the gain codebook 124 are multiplied by 123, respectively, and the driving vector obtained by adding them by the adder 112 is added to the weighted combined vector 1
Search unit 1 so that the distortion (error with respect to the target vector) of the synthesized speech signal vector obtained through 13 is minimized.
At 42, the gain codebook 124 is closed loop in a known manner.
Is searched.

As described above, the search units 140, 141,
The index (first index) of the first adaptive codebook 110 searched in 142, the second adaptive codebook 120
Alternatively, the index of the random codebook 121 (second index) and the index of the gain codebook 124 (third index) are the quantized parameters from the LSP quantizer 103 and the classification determination from the signal classifier 130. It is output as a coding parameter together with the signal through the multiplexer 150. Multiplexer 15
The coding parameter output from 0 is sent to a transmission line or a storage medium.

As described above, in the speech coding apparatus of this embodiment, when the input speech signal is classified as periodic or periodic and stationary by the signal classifier 130, the first
Even if the periodic component of the drive signal of the weighted synthesis filter 113 is expressed by the adaptive codebook 110 of, and the periodic component remains in the residual obtained from the subtractor 114 due to the reduction of the encoding bit rate, Since this periodic component can be expressed by the second adaptive codebook 120, it is possible to obtain a reproduced voice signal of high quality in the voice decoding device even at a low encoding bit rate.

When the signal classifier 130 classifies the input speech signal as aperiodic or periodic and nonstationary, the noise codebook 121 or the adaptive codebook 110 is used.
By generating the driving signal using the noise codebook 121 and the noise codebook 121, it is possible to obtain a good reproduced speech signal in the speech decoding device, without using the adaptive codebook that utilizes the correlation with the past driving signal. .

Next, the speech decoding apparatus according to this embodiment will be described. FIG. 4 is a block diagram showing the configuration of the speech decoding apparatus according to this embodiment. This speech decoding apparatus includes an encoded data input terminal 160, a demultiplexer 161, an LSP dequantizer 162, and a first adaptive codebook 1.
70, gain circuit 171, adder 172, synthesis filter 173, post filter 174, second adaptive codebook 18
0, a noise codebook 181, a switch 182, a gain circuit 183, a gain codebook 184, and a reproduced voice signal output terminal 190.

Coding parameters output from the speech coding apparatus shown in FIG. 1 are input to the input terminal 160 via a transmission line or a storage medium. This coding parameter is input to the demultiplexer 161, and the LSP parameter quantized by the LSP quantizer 103 of FIG. 1, the index (first index) of the first adaptive codebook 110, and the second adaptive codebook. 120 or the noise codebook 121
Index (second index), gain codebook 1
The 24 indexes (third index) and the classification determination signal from the signal classifier 130 are separated and decoded.

Of the output of the demultiplexer 161, the quantized LSP parameter is the LSP dequantizer 162.
LSP parameters are restored by inputting to the first adaptive codebook 170 and the second index is the first adaptive codebook 170, and the second index is the second adaptive codebook 180 and the random codebook 1
81, the third index is input to the gain codebook 184, and the classification determination signal is input to the switch 182.

First adaptive codebook 170, second adaptive codebook 180, random codebook 181, and gain codebook 184.
Are configured similarly to the first adaptive codebook 110, the second adaptive codebook 120, the noise codebook 121, and the gain codebook 124 in the speech coding apparatus shown in FIG.

The operation of this speech decoding apparatus will be described below.
The first drive vector obtained by searching the first adaptive codebook 170 based on the first index output from the demultiplexer 161 is based on the third index from the gain codebook 184 in the gain circuit 171. It is input to the adder 172 after being multiplied by the gain searched for. On the other hand, the second drive vector obtained by searching the second adaptive codebook 180 or the random codebook 181 based on the second index is selected by the switch 182 based on the classification determination signal, and similarly. The gain circuit 172 multiplies the gain searched by the gain codebook 184 based on the third index, and inputs the result to the adder 172.

The drive vector synthesized by the adder 172 is given as a drive signal to the synthesis filter 173 whose transfer characteristic is controlled by the LSP parameter from the LSP dequantizer 162, and this synthesis filter 173 synthesizes the synthesized voice signal vector. Is generated. This synthesized speech signal vector is input to the post filter 174 whose transfer characteristic is also controlled by the LSP parameter, and here, well-known processing for improving the subjective quality of the reproduced speech, such as pitch enhancement, formant enhancement, and high frequency enhancement. After being subjected to processing such as gain adjustment and the like, it is output from the output terminal 190 as a reproduced audio signal. This reproduced audio signal has high quality as described above.

(Second Embodiment) FIG. 5 is a block diagram of a speech encoding apparatus according to the second embodiment. The same reference numerals are given to the portions corresponding to those in FIG. 1, and the description will focus on the differences from the first embodiment. The difference between the second embodiment and the first embodiment is 2 There is a method of selecting a codebook which is a drive signal source of the stage. That is, in the first embodiment, the characteristics are classified by analyzing the input voice signal,
The codebook used as the drive signal of the second stage was switched between the second adaptive codebook 120 and the noise codebook 121.

On the other hand, in the second embodiment, the first adaptive codebook 21 is used as the codebook which is the drive signal source of the first stage.
0 and two codebooks of fixed codebook 211 in which the stored drive vector is always fixed.
One of 0 and 211 is selected by the minimum distortion criterion of the synthesized voice signal level, and the result is classified into the characteristics of the input voice signal according to whether they are periodic or not, and whether they are periodic and stationary. The second adaptive codebook 120 and the noise codebook 121 are used as the codebooks used as the drive signal source of the second stage.
Select either one of. In other words, in the first embodiment, the codebook is selected in an open loop, whereas in the second embodiment, the second-stage codebook is selected in a closed loop.

The operation of this embodiment will be described in detail.
First, LPC analysis and LSP parameter quantization are performed in frame units exactly as in the first embodiment. Next, search section 140 searches for first adaptive codebook 210 as in the first embodiment, and searches for fixed codebook 211 as in the random codebook 121 in the first embodiment. , The drive vector searched from the adaptive codebook 210 and the fixed codebook 211 is extracted via the switch 212.

Then, the gain circuit 1 is added to the search results of the first adaptive codebook 210 and the fixed codebook 211, that is, the drive vector searched from the first adaptive codebook 210.
11, the driving vector obtained by multiplying the gain searched from the gain codebook 124 by the weighting synthesis filter 1
13, the distortion of the synthetic speech vector obtained through 13 and the drive vector obtained by multiplying the drive vector searched from the fixed codebook 211 by the gain searched from the gain codebook 124 by the gain circuit 111 are weighted synthesis filters. The distortions of the synthetic speech vectors obtained through 113 are compared, and the drive vector that gives the minimum distortion is taken as the drive vector from the drive signal source of the first stage. The index of the first adaptive codebook 210 or the fixed codebook 211 obtained by the search unit 140 by this search is the multiplexer 15
0 to the signal classifier 230.

In the signal classifier 230, whether the first adaptive codebook 210 is selected as the drive signal source of the first stage from the index input from the search unit 140 or the fixed codebook 21 is selected.
By determining whether 1 is selected, the nature of the input speech signal is classified according to whether it is periodic or whether it is periodic and stationary. Then, the signal classifier 230
When the adaptive codebook 210 is selected, that is, when the input speech signal is periodic or periodic and stationary, the second adaptive codebook 220 is selected as the second stage drive signal source, Otherwise, the switch 122 is controlled so that the random codebook 121 is selected.

The search for the codebook as the second stage drive signal source is performed in the same manner as in the first embodiment. In this case, the signal classifier 230 selects the second-stage codebook in consideration of not only whether the first adaptive codebook 210 is selected as the first-stage drive signal source but also the pitch continuity. It is also possible to adopt a configuration for performing.

Since the speech decoding apparatus corresponding to the speech encoding apparatus of FIG. 5 is the same as the speech decoding apparatus of FIG. 4, description thereof will be omitted. (Third Embodiment) In the first and second embodiments, the drive signal source is composed of a two-stage codebook, but it may be three or more stages. FIG. 6 is a block diagram showing the configuration of the speech encoding apparatus according to the third embodiment when the drive signal source has three stages. As shown in the figure, in the present embodiment, in addition to the configuration of the first embodiment shown in FIG. 1, another noise codebook 330 which is the drive signal source of the third stage and the gain codebook 124 at the output thereof. The configuration is such that a gain circuit 331 for multiplying the gain searched for from and a search unit 143 for searching the random codebook 330 are added.

The codebook search operation in this embodiment is as follows.
First, the first adaptive codebook 110 that is the drive signal source of the first stage
Of the second adaptive codebook 120 or the noise codebook 121, which is the driving signal source of the second stage, is searched by the searching unit 141, and finally the driving signal source of the third stage is searched. The random codebook 330 is searched by the search unit 143.

In the above embodiment, the input voice signal is input to the signal classifier 130 as shown in FIG. 1 as means for classifying the nature of the input voice signal according to whether it is periodic or periodic and stationary. Classification is performed by voiced / unvoiced determination or the like, or whether the first adaptive codebook 210 is selected in the first stage as shown in FIG. 5, or the second adaptive codebook 210 in the second stage as shown in FIG. Although the method of indirectly classifying according to whether or not the adaptive codebook 120 is selected is used, the method is not limited to this.

For example, the power of the input voice signal may be judged to classify the property of the input voice signal periodically or periodically and whether it is stationary, or the first adaptive codebook 1
The gain circuit 111 is added to the drive vector read from
Various modifications are possible, such as determining whether the gain multiplied by exceeds a predetermined threshold and classifying the input voice signal as periodic or periodic and stationary when the threshold is exceeded. is there.

[0056]

As described above, according to the present invention, the characteristics of the input speech signal are classified according to whether they are periodic or periodic and stationary, and are classified as periodic or periodic and stationary. In this case, the drive signal is generated by the first and second two-stage adaptive codebooks, so that the periodic components that cannot be sufficiently removed by the first adaptive codebook and remain are generated by the second adaptive codebook. When the configuration is represented, when classified as non-periodic or periodic and non-stationary, a drive signal is created using a noise codebook to represent a signal having a small correlation. As a result, it is possible to always configure a drive signal that matches the characteristics of the input audio signal, and it is possible to reproduce audio of good quality even at a low bit rate of about 4 kbps.

Further, in the present invention, since the drive signal is composed of the adaptive codebook, there is an effect that the special amount of the adaptive codebook can be used to reduce the amount of calculation which has been a problem in the CELP system. That is, similar to the SEV method, each drive vector in the adaptive codebook is configured by shifting from the past drive signal by a delay corresponding to the pitch period, and as a result, each drive vector has a configuration in which its elements overlap. By taking advantage of the fact that the output of the synthesis filter for the drive vector is recursively calculated, it is possible to significantly reduce the amount of calculation.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a configuration of a speech encoding device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing an example of waveforms of voiced sections and unvoiced sections of a voice signal.

FIG. 3 is a flowchart showing a flow of main processing in the same embodiment.

FIG. 4 is a block diagram showing a configuration of a speech decoding apparatus according to the embodiment.

FIG. 5 is a block diagram showing a configuration of a speech encoding apparatus according to a second embodiment of the present invention.

FIG. 6 is a block diagram showing the configuration of a speech encoding apparatus according to a third embodiment of the present invention.

FIG. 7 is a block diagram showing a typical configuration of a conventional CELP audio encoding apparatus.

FIG. 8 is a block diagram showing a configuration of a main part of a conventional SEV audio encoding device.

[Explanation of symbols]

 100 ... Audio signal input terminal 101 ... Buffer 104 ... Subframe division circuit 102 ... LPC analysis unit 103 ... LSP quantizer 105 ... Weighting filter 110 ... First adaptive codebook 111 ... Gain circuit 112 ... Adder 113 ... Weighted Synthesis filter 114 ... Subtractor 120 ... Second codebook 121 ... Noise codebook 122 ... Switcher 123 ... Gain circuit 124 ... Gain codebook 130 ... Signal classifier 140-143 ... Search unit 150 ... Multiplexer 151 ... Encoding parameter Output terminal 210 ... First adaptive codebook 211 ... Fixed codebook 150 ... Multiplexer 160 ... Coding parameter input terminal 161 ... Demultiplexer 162 ... LSP dequantizer 170 ... First adaptive codebook 171 ... Gain circuit 173 ... Synthesis filter 174 ... Post filter 80 ... Second adaptive codebook 181 ... Random codebook 182 ... Switching device 183 ... Gain circuit 210 ... First adaptive codebook 211 ... Fixed codebook 212 ... Switching device 230 ... Signal classifier 330 ... Random codebook 331 ... Gain circuit

Claims (6)

[Claims] 1. A drive signal is generated using a drive vector obtained from a plurality of drive vector codebooks, and the drive signal is supplied to a synthesis filter whose filter coefficient is determined based on an analysis result of an input audio signal. The drive vector codebook is searched for a drive vector in which the distortion of the synthesized voice signal output from the synthesis filter is smaller, and at least an encoding parameter representing the information of the drive vector and the filter coefficient of the synthesis filter is output. In the speech coding method, at least two adaptive codebooks and at least one noise codebook are prepared as the driving vector codebook, and at least the above when the property of the input speech signal is periodic or periodic and stationary. Generating the drive signal using drive vectors obtained from two adaptive codebooks; Speech encoding method and generating said drive signal using the driving vectors obtained from at least the noise code book when the nature of the signal is non-periodic or periodic and unsteady. 2. A drive signal is generated using a drive vector obtained from a plurality of drive vector codebooks, and the drive signal is supplied to a synthesis filter whose filter coefficient is determined based on the analysis result of the input audio signal. The drive vector codebook is searched for a drive vector in which the distortion of the synthesized voice signal output from the synthesis filter is smaller, and at least an encoding parameter representing the information of the drive vector and the filter coefficient of the synthesis filter is output. In the audio encoding device, at least 2 prepared as the drive vector codebook
One adaptive codebook and at least one noise codebook, a classifying unit that classifies the properties of the input speech signal according to whether they are periodic or periodic and stationary, and the input speech signal by this classifying unit Is classified as periodic or periodic and stationary, the drive signal obtained from at least the two adaptive codebooks is used to generate the drive signal, and the input voice signal has a non-periodic nature. Or a drive signal generation means for generating the drive signal using at least one of the adaptive codebook and the drive vector obtained from the noise codebook when classified as periodic and non-stationary. Speech coding device. 3. A drive signal is generated using a drive vector obtained from a plurality of drive vector codebooks, and this drive signal is supplied to a synthesis filter whose filter coefficient is determined based on the analysis result of the input audio signal. The drive vector codebook is searched for a drive vector in which the distortion of the synthesized voice signal output from the synthesis filter is smaller, and at least an encoding parameter representing the information of the drive vector and the filter coefficient of the synthesis filter is output. In the speech coding method, at least two adaptive codebooks, at least one fixed codebook and at least one noise codebook are prepared as the driving vector codebook, and the first adaptive codebook and the fixed codebook are prepared in a first step. Of the codebooks, the one with the smaller distortion of the synthesized speech signal is selected,
In the second step, it is determined whether the first adaptive codebook or the fixed codebook is selected in the first step to determine whether the property of the input speech signal is periodic, or whether it is periodic and stationary. If the nature of the input speech signal is periodic or periodic and stationary, the driving signal is generated using at least the driving vector obtained from the two adaptive codebooks, and the input signal A voice encoding method, wherein the drive signal is generated using at least a drive vector obtained from the noise codebook when the property of the voice signal is aperiodic or periodic and nonstationary. 4. A drive signal is generated using a drive vector obtained from a plurality of drive vector codebooks, and the drive signal is supplied to a synthesis filter whose filter coefficient is determined based on an analysis result of an input voice signal. The drive vector codebook is searched for a drive vector in which the distortion of the synthesized voice signal output from the synthesis filter is smaller, and at least an encoding parameter representing the information of the drive vector and the filter coefficient of the synthesis filter is output. In the audio encoding device, at least 2 prepared as the drive vector codebook
One adaptive codebook, at least one fixed codebook and at least one noise codebook, and one of the first adaptive codebook and the fixed codebook, whichever of the first adaptive codebook and the fixed codebook has the smaller distortion of the synthesized speech signal in the first stage, Selected,
In the second stage, it is determined whether the first adaptive codebook is selected or the fixed codebook is selected, and the classification means classifies the input speech signal according to whether the property is periodic or periodic and stationary. And when the classifying means classifies the input speech signal as periodic or periodic and stationary, the driving signal is generated using the driving vector obtained from at least two adaptive codebooks. Drive signal generation for generating the drive signal using a drive vector obtained from at least one adaptive codebook and noise codebook when the characteristics of the input speech signal are classified as aperiodic or periodic and nonstationary A speech coding apparatus having means. 5. A drive signal is generated using a drive vector obtained from a plurality of drive vector codebooks, at least encoding parameters representing a drive vector and a filter coefficient of a synthesis filter are input, and the drive signal is generated by the code. In a speech decoding method for decoding a speech signal by supplying it to a synthesis filter whose filter coefficient is determined on the basis of a coding parameter, at least two adaptive codebooks and at least one noise codebook are used as the driving vector codebook. Speech decoding, which is prepared and generates the drive signal using a drive vector obtained from at least two adaptive codebooks selected based on the encoding parameter or at least a drive vector obtained from a noise codebook Method. 6. A drive signal is generated using drive vectors obtained from a plurality of drive vector codebooks, at least encoding parameters representing the drive vector and the filter coefficient of the synthesis filter are input, and the drive signal is generated by the code. In a voice decoding device for decoding a voice signal by supplying it to a synthesizing filter whose filter coefficient is determined based on a conversion parameter, at least 2 prepared as the drive vector codebook
One adaptive codebook and at least one noise codebook, and a drive vector obtained from at least two adaptive codebooks selected on the basis of the encoding parameter or at least a drive vector obtained from the noise codebook. And a drive signal generating means for generating the signal.