JP3088121B2 - Statistical excitation code vector optimization method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multistage code excitation linear predictive encoder and decoder having an adaptive excitation codebook (pitch excitation codebook) and a statistical excitation codebook (Gaussian noise excitation codebook). The present invention relates to a method for optimizing a statistical excitation code vector to be used.

[0002]

2. Description of the Related Art Excited code vectors used in code-excited linear prediction encoders and decoders are not designed as they are, but are optimized in advance so that encoding distortion is minimized. From the actual encoder and decoder.

Conventionally, as a method of optimizing an excitation code vector used in a backward type code excitation linear predictive encoder, there is a method described in the following document.

Reference: "Juin-Hwey Chen," High-Quality
16KB / S SPEECH CODING WITH AONE-WAY DELAY LESS THA
N 2MS, ”Proc.IEEE Int.Conf.Acoust., Speech, SingleP
rocessing.pp453-456 (1990). First, a backward-type code-excitation linear prediction encoder according to the optimization method will be described with reference to the block diagram of FIG.

In FIG. 2, the information transmitted to the receiving side is only the index of one of the optimized excitation code vectors stored in the excitation codebook 11. Such an optimal index is determined as follows.

The excitation signal stored in the excitation codebook 11 includes an excitation code vector and a sign (sign coefficient).
When searching for an optimal index, each stored excitation code vector is output as a candidate in time sequence. At this time, a sign coefficient is given and multiplied by an amplitude coefficient.

[0007] The excitation code vector as a candidate output from the excitation codebook 11 in this manner is applied to a gain circuit 12, multiplied by a predetermined number by the gain circuit 12, and applied to an adder 13. The gain circuit 12 is of a variable gain coefficient type, and the gain coefficient is changed by the gain control circuit 14. The gain control circuit 14 predicts a gain coefficient by performing a linear prediction analysis (LPC analysis) from a past vector sequence from the gain circuit 12 and provides the gain coefficient to the gain circuit 12.

The adder 13 is also supplied with an output signal from a synthesis filter (linear prediction filter) 15, adds the candidate excitation code vector from the gain circuit 12 and the output vector from the synthesis filter 15, and A synthesized speech vector for local reproduction using the candidate excitation code vector is obtained and given to the subtractor 16.

The linear prediction coefficient used by the synthesis filter 15 is provided from a linear prediction analysis circuit 17. Synthesis filter 1
Reference numeral 5 denotes a linear prediction analysis circuit 1 for a sequence of a locally reproduced synthesized speech vector for the past optimal excitation code vector.
The prediction synthesis processing is performed by applying the linear prediction coefficient given from 7, and the output vector is given to the adder 13. The linear prediction analysis circuit 17 obtains a linear prediction coefficient from the sequence of the synthesized speech vector with respect to the past optimal excitation code vector, and supplies it to the synthesis filter 15.

The input speech vector is also given to the subtractor 16, and the subtractor 16 subtracts the synthesized speech vector of the local reproduction in the case of using the candidate excitation code vector from the input speech vector, and obtains the subtracted speech vector. After weighting the difference vector via a perceptual weighting filter 18 according to the perceptual (audible) characteristic, the difference vector is applied to an index search circuit 19. In this way, at the time of searching for the optimal excitation code vector, the index search circuit 19 is provided with difference vectors for all the excitation code vectors.

The index search circuit 19 calculates the sum of squares of the components of each difference vector, detects the excitation code vector corresponding to the difference vector having the minimum square sum as the optimum excitation code vector, and finds the index. Give to excitation codebook 11.

As a result, the excitation codebook 11 transmits the optimum index to the receiving side as described above.
Further, the excitation codebook 11 outputs the detected optimal excitation code vector to the gain circuit 12 again, and updates the gain coefficient, updates the linear prediction coefficient, and performs local processing on the excitation code signal at the time of the next frame processing. Make the synthesized speech vector for reproduction available.

Although the encoding configuration and the encoding process have been described above, the above configuration is also used when optimizing the excitation code vector used in the encoding. That is, an input speech vector is input to execute an encoding process, values of various vectors and coefficients at that time are extracted and stored, and an excitation code vector is optimized from the stored values of various vectors and coefficients.

The conventional method of optimizing an excitation code vector described in the above document will be described below.

The amplitude coefficient of the excitation code vector used in the search processing at the time index n and the code coefficient thereof are g (n) and η (n), respectively. Let Nj be a set of time indexes at which the excitation code vector yj is selected as the optimal excitation code vector. This excitation code vector y
The total distortion Dj of the j-th cluster corresponding to j is
It is given by equation (1).

[0016]

(Equation 1)

Here, X (n) is a target vector (for example, an input speech vector or a vector obtained by perceptually weighting the input speech vector), and H (n) is a synthesis filter 15.
Response (square matrix), σ (n) is the gain circuit 1
2 is a gain coefficient.

That is, the expression (1) expresses the square of the norm of the difference between the target vector X (n) and the composite vector H (n) η (n) σ (n) g (n) yj as the excitation code vector This shows that the sum of all time indexes in which yj is optimized is represented as the total distortion Dj.

Whether the prepared excitation code vector yj is good or not is indicated by the magnitude of the total distortion Dj. When optimizing the excitation code vector yj, it is necessary to consider the excitation code vector yj as a variable and find a condition that minimizes the total distortion Dj. Therefore, as shown in equation (2), the total distortion Dj is partially differentiated with this excitation code vector yj, and its value is set to 0 which is the minimum condition.

[0020]

(Equation 2)

From this equation (2), it can be seen that the optimized excitation code vector (centroid) Yj for the j-th cluster that minimizes the total distortion Dj can be obtained by solving the equation (3).

[0022]

(Equation 3)

In equation (3), various values g (n), η (n), X (n), and H (g) when encoded by the actual encoder described above are used.
(n) and σ (n) are substituted to obtain an optimized excitation code vector Yj.

When one such optimization process is completed, a similar optimization process is performed by replacing the contents of the excitation codebook 11 with the obtained excitation code vector. When such optimization processing is repeated, the excitation code vector converges, converges to a point where the total distortion does not decrease even if the optimization processing is repeated, and the excitation code vector at that time is finalized (optimization processing). Is completed) is determined as the excitation code vector.

[0025]

By the way, recently, 2
Multi-stage code-excited linear predictive encoders and decoders having different types of excitation codebooks have been increasingly used. That is, the number of code excitation linear prediction encoders and decoders having a pitch excitation codebook (adaptive excitation codebook) and a Gaussian noise excitation codebook (statistical excitation codebook) has increased.

When the above-described conventional method is used for optimizing the statistical excitation code vector stored in the statistical excitation codebook used for such a multi-stage code excitation linear predictive encoder and decoder, the optimal adaptive excitation code vector Is determined, the information of the optimal statistical excitation code vector determined is subjected to optimization processing.

For this reason, the statistical excitation code vector is optimized once by applying the optimization method, and when the optimization method is applied again by replacing the obtained statistical excitation code vector with the previously obtained adaptive excitation code vector, Are different. As described above, the content of the adaptive excitation code vector is different every time the optimization method is applied. Therefore, even if the optimization method is applied, the total distortion of the finally obtained statistical excitation code vector cannot be reduced to a very small value. That is, the convergence characteristic is deteriorated.

The present invention has been made in view of the above points, and a statistical excitation code capable of converging a statistical excitation code vector until the total distortion becomes sufficiently small irrespective of the existence of an adaptive excitation code vector. It is intended to provide a vector optimization method.

[0029]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention provides a multi-stage code excitation linear prediction encoder and a statistical excitation code decoder having an adaptive excitation codebook and a statistical excitation codebook. A method for optimizing a code vector, which is a statistical excitation code vector for optimizing a statistical excitation code vector stored in a statistical code book based on information of each unit obtained while executing an encoding process on an input speech vector. In the optimization method, the adaptive excitation code vector in which the high-frequency component is attenuated and the statistical excitation code vector are combined during the encoding processing of the input speech vector performed to collect information of each part for performing the optimization. I did it.

Further, the second invention provides a method of optimizing a statistical excitation code vector used in a multi-stage code excitation linear prediction encoder and decoder having an adaptive excitation codebook and a statistical excitation codebook. In a statistical excitation code vector optimization method for optimizing a statistical excitation code vector stored in a statistical code book based on information of each unit obtained while performing an encoding process on an input speech vector, During the encoding process of the input speech vector performed to collect the information of each part for performing the processing, the adaptive excitation code vector in which the high frequency component is attenuated and the statistical excitation code vector subjected to the center clipping process are synthesized. I made it.

[0031]

When optimizing a statistical excitation code vector, it cannot be avoided that the information of the adaptive code vector affects the obtained optimized excitation code vector.
If this effect is large, optimization cannot be performed well.

Therefore, in the first invention, the role of the frequency in the combined excitation code vector is changed between the adaptive excitation code vector and the statistical excitation code vector. That is, encoding is performed using a combined excitation code vector obtained by combining an adaptive excitation code vector in which a high-frequency component is attenuated and a statistical excitation code vector, and various values used for optimization are obtained.

The second present invention, like the first present invention,
In order to change the role of the frequency in the combined excitation code vector between the adaptive excitation code vector and the statistical excitation code vector, the high frequency band of the adaptive excitation code vector is removed. It also emphasizes the periodicity of the adaptive excitation code vector,
By improving the contribution of the adaptive excitation code vector to the synthesized sound, the coding distortion for voiced sound is reduced, and the statistical excitation code vector is improved in order to improve the convergence characteristics of the total distortion in the optimization processing of the statistical excitation code vector. Center clipping processing.

[0034]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment Hereinafter, a first embodiment of the present invention will be described in detail with reference to the drawings. Here, FIG. 1 shows a configuration relating to a method of optimizing a statistical excitation code vector according to this embodiment.

The optimization method of this embodiment is also basically
The input speech vector is input to the code-excitation linear predictive encoder and operated, and various values when the optimal statistical excitation code vector is determined are accumulated, and when the accumulated amount reaches a certain level, the same statistical excitation is applied. This is a method of calculating and updating the optimized statistical excitation code vector that minimizes the total distortion from various values when the code vector is obtained. Therefore, as shown in FIG. 1, the configuration for optimizing the statistical excitation code vector is obtained by adding a plurality of calculation circuits to the configuration of the code excitation linear prediction encoder.

First, the configuration and operation of the code excitation linear prediction encoder according to this embodiment will be described.

In FIG. 1, the adaptive excitation codebook 20
The statistical excitation codebook 21 stores an adaptive excitation code vector (pitch excitation code vector) and a statistical excitation code vector (Gaussian noise excitation code vector). From these stored excitation code vectors, an optimum vector is searched for the input speech vector at that time (time index), and the index is output.

The search for the optimal adaptive excitation code vector and the optimal statistical excitation code vector is performed in the following order. With the output of the statistical excitation code vector stopped,
When the optimal adaptive excitation code vector is searched for and the optimal adaptive excitation code vector is searched, the optimal statistical excitation code vector is searched for with the optimal adaptive excitation code vector output from the adaptive excitation code book 20. Perform In addition, in a state where both the adaptive excitation code vector and the statistical excitation code vector are searched for the optimal one, the optimal vector is output from the adaptive excitation code book 20 and the statistical excitation code book 21 and the state of each unit is changed to the next time. Is changed in preparation for.

When the input speech vector at a certain time is processed as described above, the outputs of the adaptive excitation codebook 20 and the statistical excitation codebook 21 change depending on the processing stage. The configuration and operation of the latter stage of the book 21 will be described irrespective of whether an optimal excitation code vector is output.

The adaptive excitation code vector Va output from the adaptive excitation code book 20 is applied to the low-pass filter 22.
, The high-frequency component is removed, the vector VL after the removal is applied to the multiplier 23, multiplied by the amplitude coefficient ga, and applied to the adder 24. On the other hand, the statistical excitation code vector Vs output from the statistical excitation codebook 21 is supplied to the multiplier 25, multiplied by the amplitude coefficient gs, and supplied to the adder 24. Thus, a combined vector V of the adaptive excitation code vector and the statistical excitation code vector whose amplitude has been adjusted is obtained from the adder 24. The low-pass filter 22 is provided in consideration of the optimization of the statistical excitation code vector. The specific reason will be described later.

The combined excitation code vector V output from the adder 24 is applied to a gain circuit 26. When the optimum excitation code vector is output from both the adaptive excitation code book 20 and the statistical excitation code book 21, the combined excitation code vector V from the adder 24 is output.
opt is also given to the adaptive excitation codebook 20 and used for updating the adaptive excitation codebook 20. The gain circuit 26 multiplies the combined excitation code vector V by the gain coefficient σ given from the gain control circuit 27, and supplies the multiplied vector Vg to the adder 28. The vector Vgopt from the gain circuit 26 when the optimal excitation code vector is output from both the adaptive excitation codebook 20 and the statistical excitation codebook 21 is given to the gain control circuit 27. The gain control circuit 27 determines a new gain coefficient σ by applying a linear prediction analysis method to the past optimal vector Vgopt series provided from the gain circuit 26.

The adder 28 includes the synthesis filter 2
The output vector Sp from 9 is also given. This adder 2
8, a synthesized speech vector Sw formed based on the synthesized excitation code vector V at that time is obtained, and this vector Sw is given to the subtractor 30. When the optimal excitation code vector is output from both the adaptive excitation codebook 20 and the statistical excitation codebook 21, the optimal synthesized speech vector Swopt is calculated by the synthesis filter 2
9 and the synthesized speech linear prediction analysis circuit 31.
The linear prediction analysis circuit 31 performs a linear prediction analysis on a series of past optimal synthesized speech vectors Swopt, and obtains a linear prediction coefficient α (i) (i is 1 to m: m is an analysis order) Is given to the synthesis filter 29. Synthetic filter 2
9 synthesizes a past optimal synthesized speech vector Swopt sequence using the current linear prediction coefficient α (i),
The obtained vector Sp is given to the adder 28 as described above.

The input speech vector S is also input to the subtractor 30, and the subtracter 30 subtracts the synthesized speech vector Sw from the input speech vector S, and supplies the difference vector to the perceptual weighting filter 32. The perceptual weighting filter 32 is also provided with a linear prediction coefficient αw (i) from the input speech linear prediction analysis circuit 33. The input speech linear prediction analysis circuit 33 applies a linear prediction analysis to the input speech vector S to obtain a linear prediction coefficient αw (i).

The perceptual weighting filter 32 includes a subtractor 30
Is weighted in consideration of the perceptual characteristics (auditory characteristics), and the weighted difference vector is provided to the index search circuit 34. The perceptual weighting filter 32 specifically performs the conversion shown in Expression (4) (where z means a difference vector which is an input).

[0045]

(Equation 4)

The index search circuit 34 calculates the sum of squares of the weighted difference vector, and determines that the excitation code vector that minimizes the sum of squares is the optimum one.
That is, when searching for the optimal adaptive excitation code vector, the one with the minimum sum of squares is detected from all the difference vectors given for all the adaptive excitation code vectors. Further, when searching for an optimal statistical excitation code vector, the sum of squares is the smallest among all difference vectors given for all statistical excitation code vectors in a state where the optimal vector is output as an adaptive excitation code vector. Detect things.

The configuration and operation of the code excitation linear prediction encoder have been described above, but these configurations are also used when optimizing the statistical excitation code vector. At the time of optimization, the impulse response calculation circuit 40 and the optimum code vector calculation circuit 41 further operate.

When optimizing the statistical excitation code vector, the input speech vector S is input to the above-described code excitation linear predictive encoder to operate it, and the time when the optimal statistical excitation code vector Vsopt is determined ( Various values H (z), ga, gs, σ,
S, Vsopt and VLopt are converted to an optimal code vector calculation circuit 4
1 and accumulates, and when the accumulated amount reaches a certain level, the optimized statistical excitation code that minimizes the total distortion from various values when the optimal code vector calculation circuit 41 takes the same statistical excitation code vector The vector is calculated and updated.

As is clear from the above description of the configuration and operation of the code excitation linear prediction encoder, the values ga, gs, σ, S, Vsopt used by the optimum code vector calculation circuit 41 are described.
And VLopt are obtained from a component as an encoder. Only the impulse response H (z) cannot be obtained from the encoder configuration, so that an impulse response calculation circuit 40 is provided.

The linear prediction coefficient α (i) is given to the impulse response calculation circuit 40 from the linear prediction analysis circuit 31 for synthesized speech. The impulse response calculation circuit 40 calculates the impulse response H (z) of the transfer function Hw (z) shown in the equation (5) determined by the linear prediction coefficient α (i), and outputs it to the optimum code vector calculation circuit 41. In the following description,
Impulse response H (z) at time (time index) n
Is represented by H (n).

[0051]

(Equation 5)

The optimum code vector calculation circuit 41 calculates (6)
The target vector X (n) at each time (time index) n is calculated as shown in the equation. Now, since the optimization target is the statistical excitation code vector, instead of using the input speech vector S (n) as the target vector, the optimal adaptive excitation code vector VL ( The target vector X (n) from which the influence of (n) is removed is used.

[0053]

(Equation 6)

The optimum code vector calculation circuit 41
Groups the set of various values at a certain time with the same set of optimal statistical excitation code vectors. The optimum code vector calculation circuit 41 calculates the k-th statistical excitation code stored in the statistical excitation codebook 21 (k is 1 to t: t is the total number of statistical excitation code vectors stored in the statistical excitation codebook 21). Set N of sets related to vector Vsk
k information, the k-th statistical excitation code vector V
The statistical excitation code vector V ^* sk in which sk is optimized is obtained by solving the equation shown in Expression (7).

[0055]

(Equation 7)

This equation (7) is based on the same concept as the conventional optimization method. That is, the statistical excitation code vector V ^* sk optimized from the condition (minimum square error condition) that minimizes the total distortion of the set of time indexes Nk that optimizes the statistical excitation code vector Vsk is considered. Is obtained.

Equation (7) is an equation relating to the statistical excitation code vector Vsk. Since the information relating to the adaptive excitation code vector VL (n) is used for calculating the target vector X (n), the obtained optimal It is unavoidable that the information of the adaptive excitation code vector VL (n) affects the statistical excitation code vector V ^* sk. If this effect is large, optimization cannot be performed well.

Therefore, in this embodiment, the role of the frequency in the combined excitation code vector is changed between the adaptive excitation code vector and the statistical excitation code vector. In this way, the target vector X (n) obtained by subtracting the synthesized speech vector component related to the adaptive excitation code vector from the input speech vector S (n) is related to the same frequency as the statistical excitation code vector. As good. The low-pass filter 22 described above is provided to change the role of the frequency in the combined excitation code vector between the adaptive excitation code vector and the statistical excitation code vector. Since the statistical excitation code vector is a noise vector, the statistical excitation code vector cannot be associated with the low frequency side.

The optimized statistical excitation code vector V ^* sk obtained as described above is provided to the statistical excitation codebook 21. The statistical excitation codebook 21 updates the contents stored in the optimized statistical excitation code vector V ^* sk.

The process of optimizing the statistical excitation code vector is repeated a plurality of times. The optimum code vector calculation circuit 41 calculates the optimized statistical excitation code vector V
Determine whether to repeat the optimization process based on the total distortion Dk obtained by equation (8) for ^* sk. That is, the repetition of the optimization process is terminated when the improvement of the total distortion in the past and the current optimization process does not show any improvement in the total distortion even if the optimization process is further performed (asymptotic characteristics).

[0061]

(Equation 8)

According to the first embodiment described above, the low-pass filter 22 is provided at the next stage of the adaptive excitation codebook 20, and the roles of the frequencies of the adaptive excitation code vector and the statistical excitation code vector in the composite excitation code vector are different. Thus, the statistical excitation code vector can be optimized until the total distortion becomes sufficiently small regardless of the existence of the adaptive excitation code vector.

FIG. 3 shows a case where the low-pass filter 22 is inserted in the next stage of the adaptive excitation codebook 20 (shown by a broken line connecting circles) and a case where the low-pass filter 22 is not inserted (shown by a broken line connecting marks x). 3) shows the change (convergence characteristic) of the total distortion with respect to the number of repetitions of the optimization processing. As is clear from FIG. 3, in the case of the first embodiment (in the case where the low-pass filter 22 is interposed), the optimization can be repeated so that the total distortion becomes smaller than in the conventional case.

Second Embodiment Next, a second embodiment of the present invention will be described with reference to FIG. FIG. 4 shows a main configuration of the second embodiment, and the other configuration is the same as that of the first embodiment.

In the second embodiment, a low-pass filter 22 is provided at the next stage of the adaptive excitation codebook 20 in order to optimize the statistical excitation code vector, and a center clipping process is performed at the next stage of the statistical excitation codebook 21. The circuit 45 is provided.

The operation of the low-pass filter 22 and the significance thereof are the same as in the first embodiment.

The center clipping processing circuit 45 replaces the components of the statistical excitation code vector output from the statistical excitation codebook 21 that are smaller than a predetermined threshold with 0, and the statistical clipping after the center clipping process is performed. The excitation code vector is provided to the multiplier 25.

The center clipping processing circuit 45 is provided to enhance the periodicity of the adaptive excitation code vector and improve the contribution of the adaptive excitation code vector to the synthesized sound. That is, it is for reducing the coding distortion for voiced sound and improving the convergence characteristics of the total distortion in the optimization processing of the statistical excitation code vector.

Therefore, according to the second embodiment, the statistical excitation code vector can be optimized until the total distortion becomes sufficiently small regardless of the existence of the adaptive excitation code vector.

Other Embodiments In the above embodiments, optimization of statistical excitation code vectors relating to a so-called backward type multi-stage code excitation linear prediction encoder and decoder has been described. The present invention may be applied to optimization of a statistical excitation code vector related to a forward type multi-stage code excitation linear prediction encoder and decoder.

In the first embodiment, the low-pass filter is provided on the output side of the adaptive excitation codebook 20 so that the frequency roles of the adaptive excitation code vector and the statistical excitation code vector in the composite excitation code vector are different. Although the configuration provided with 22 is shown, a low-pass filter may be provided on the input side of the adaptive excitation codebook 20.

[0072]

As described above, according to the first aspect of the present invention,
At the time of the encoding process of the input speech vector performed to collect the information of each part for performing the optimization, since the adaptive excitation code vector in which the high-frequency component is attenuated and the statistical excitation code vector are synthesized, Regardless of the existence of the adaptive excitation code vector, the statistical excitation code vector can be optimized so that the total distortion is sufficiently small.

According to the second aspect of the present invention, the adaptive excitation code vector in which the high-frequency component is attenuated during the coding process of the input speech vector performed to collect the information of each part for performing the optimization. And the statistical excitation code vector that has been subjected to the center clipping process.
Regardless of the existence of the adaptive excitation code vector, the statistical excitation code vector can be optimized so that the total distortion is sufficiently small.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration according to an optimization method according to a first embodiment.

FIG. 2 is a block diagram showing a configuration according to a conventional optimization method.

FIG. 3 is an explanatory diagram of an effect of the first embodiment.

FIG. 4 is a block diagram illustrating a configuration according to an optimization method according to a second embodiment.

[Explanation of symbols]

20: adaptive excitation codebook, 21: statistical excitation codebook, 22: low-pass filter, 23, 25: multiplier,
24, 28 adder, 26 gain circuit, 27 gain control circuit, 29 synthesis filter, 30 subtractor, 31
Linear prediction analysis circuit for synthesized speech, 32: Perceptual weighting filter, 33: Linear prediction analysis circuit for input speech, 34: Index search circuit, 40: Impulse response calculation circuit, 4
1 ... Optimal code vector calculation circuit, 45 ... Center clipping processing circuit.

Continuation of the front page (56) References JP-A-64-40899 (JP, A) JP-A-64-54497 (JP, A) JP-A-4-51199 (JP, A) JP-A-4-114516 (JP) , A) KLEIJN W B: "Fast Methods for the CE LP Speech Coding Algorithm.", IEEE Transactions on Acoustics, Speech and Signal. 1330-1342 (1990). (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 19/00-19/14 H03M 3/00-11/00 H04B 14/00-14/08 JICST file ( JOIS)

Claims (2)

(57) [Claims] 1. A method for optimizing a statistical excitation code vector used in a multi-stage code excitation linear prediction encoder and decoder having an adaptive excitation codebook and a statistical excitation codebook, comprising: In the statistical excitation code vector optimizing method for optimizing the statistical excitation code vector stored in the statistical code book based on the information of each part obtained while executing the processing, information of each part for performing the optimization is collected. A method of optimizing a statistical excitation code vector, wherein an adaptive excitation code vector with attenuated high-frequency components and a statistical excitation code vector are combined during the encoding process of the input speech vector performed for . 2. A method for optimizing a statistical excitation code vector used in a multi-stage code excitation linear predictive encoder and decoder having an adaptive excitation codebook and a statistical excitation codebook, comprising the steps of: Based on the information of each part obtained while executing the process, in the method of optimizing the statistical excitation code vector that optimizes the statistical excitation code vector stored in the statistical code book, collect the information of each part for optimization In the encoding process of the input speech vector to be performed, the adaptive excitation code vector in which the high-frequency component is attenuated and the statistical excitation code vector subjected to the center clipping process are combined. Excitation code vector optimization method.