JP2007212637A - Voice coding device, voice decoding device, voice coding method and voice decoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To minimize sound quality deterioration and to reduce an amount of generated codes. <P>SOLUTION: A voice coding device 100 comprises: a linear prediction analysis section 1 for calculating a linear prediction coefficient from an input voice signal by a predetermined minute unit; and a representative linear prediction coefficient calculation section 2 in which similarity of each linear prediction coefficient is judged based on distance between linear prediction coefficients, and a representative linear prediction coefficient which is a representative of a plurality of linear prediction coefficients having continuous similar values is calculated, and the plurality of similar linear prediction coefficients are replaced with the representative linear prediction coefficient. An excitation signal is selected so that an error between a synthesized voice calculated by inputting the excitation signal as a driving signal to a linear prediction filter which is generated from the obtained linear prediction coefficient, and the input voice signal, may become a minimum. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a speech encoding device, speech decoding device, speech encoding method, and speech decoding method.

  Many speech compression methods used in mobile phones, etc., calculate linear prediction coefficients from input speech signals, search for excitation signals close to the speech signals, and set parameters for both the excitation signals and the linear prediction coefficients. This is based on a method of outputting as a compressed code (encoded signal) (for example, see Non-Patent Document 1).

  FIG. 7 shows the configuration of a conventional speech encoding apparatus 300. The linear prediction analysis unit 30 calculates a linear prediction coefficient by performing linear prediction analysis of the input speech signal. Here, a certain length of the audio signal on the time axis is grouped and a linear prediction coefficient is calculated in the unit of the certain length. In order to reduce the amount of generated code of the speech encoding apparatus 300, the linear prediction coefficients are encoded with every other fixed length as a whole, so that the interpolating unit 31 uses the linear prediction coefficients thinned out before and after the linear prediction coefficients. It is generated by interpolation from the prediction coefficient.

  The linear prediction filter unit 32 generates a linear prediction filter from the linear prediction coefficient after interpolation. The adaptive code extracted from the adaptive codebook of the adaptive codebook search unit 34 and the noise code extracted from the noise codebook of the noise codebook search unit 35 are amplified by the amplifiers 36 and 37, respectively, and then combined by the combining unit 38. The synthesized signal is filtered by the generated linear prediction filter.

The error calculator 33 calculates an error between the filtered signal and the input audio signal. The speech encoding apparatus 300 outputs an index representing a noise code and an adaptive code when the error calculated by the error calculation unit 33 is minimized, and every other linear prediction coefficient as a code. In general, linear prediction coefficients are calculated at intervals of about 5 to 7 ms, but linear prediction coefficients in transmission codes are transmitted at intervals of 10 to 15 ms, and interpolation processing is often performed on the receiving side.
Audio compression standard ITU-T G. 729

  However, in the conventional speech coding method described above, the linear prediction coefficient generated by the interpolation process is calculated because the linear prediction coefficient is calculated in a unit of a fixed length and the interpolation process is performed regardless of the characteristics of the input speech. However, there are cases where it is not appropriate for the input voice (particularly during the transition period), leading to a reduction in sound quality.

  An object of the present invention is to suppress a decrease in sound quality to a minimum and to reduce the amount of generated codes.

  In order to solve the above-mentioned problem, the invention described in claim 1 is a speech encoding apparatus that encodes input speech using a linear prediction filter and an excitation signal, and linear prediction is performed in predetermined minute units from the input speech signal. A linear prediction coefficient calculation means for calculating a coefficient, a distance calculation means for calculating a distance between the linear prediction coefficients calculated by the linear prediction coefficient calculation means, and a distance between the linear prediction coefficients calculated by the distance calculation means. If there are a plurality of linear prediction coefficients having similar values in succession, a representative linear prediction coefficient that is representative of the plurality of linear prediction coefficients is calculated. Representative linear prediction coefficient calculating means for performing a process of replacing the linear prediction coefficient of the representative prediction coefficient with the calculated representative linear prediction coefficient, the linear prediction coefficient calculating means, and the representative linear prediction coefficient calculating means Generation means for generating a linear prediction filter using the obtained linear prediction coefficient, synthesized speech calculated by inputting an excitation signal as a drive signal to the linear prediction filter generated by the generation means, and the input Selecting means for selecting an excitation signal that minimizes an error from the audio signal.

  According to a second aspect of the present invention, in the speech coding apparatus according to the first aspect, the distance calculation unit converts the linear prediction coefficient calculated by the linear prediction coefficient calculation unit into an LSP coefficient, and The distance between the linear prediction coefficients is calculated using the distance.

  According to a third aspect of the present invention, in the speech coding apparatus according to the first or second aspect, the representative linear prediction coefficient calculation unit is configured to perform a case where the distance between the linear prediction coefficients is smaller than a preset fixed value. It is characterized in that it is judged as having a similar value.

  According to a fourth aspect of the present invention, in the speech coding apparatus according to the first or second aspect, the representative linear prediction coefficient calculation means sets a distance between the linear prediction coefficients in advance for a specific linear prediction coefficient. It is characterized in that it is judged as having a similar value when it is smaller than the ratio.

  According to a fifth aspect of the present invention, in the speech coding apparatus according to any one of the first to fourth aspects, the linear prediction coefficient obtained by the linear prediction coefficient calculating means and the representative linear prediction coefficient calculating means A determination unit that determines whether or not a change in the time axis is greater than a predetermined value, and a linear prediction coefficient that is thinned out at a predetermined interval when the determination unit determines that the change in the linear prediction coefficient is equal to or less than a predetermined value. Interpolating means for performing the interpolating process, and when the determining means determines that the change in the linear prediction coefficient is greater than a predetermined value, the linear predictive coefficient thinning process and the interpolating process by the interpolating means are not performed. It is characterized by that.

  The speech decoding apparatus according to claim 6 is obtained by replacing a plurality of linear prediction coefficients having similar values in succession with a representative linear prediction coefficient among linear prediction coefficients calculated in a predetermined minute unit from a speech signal. Generating a linear prediction filter from the generated new linear prediction coefficient, and inputting the excitation signal generated from the encoded speech signal to the linear prediction filter generated by the generating unit and outputting the synthesized speech And an output means.

  The speech encoding method according to claim 7, wherein the speech encoding method encodes a speech signal with a linear prediction filter and an excitation signal, and calculates a linear prediction coefficient in a predetermined minute unit from the speech signal, A distance between the calculated linear prediction coefficients is calculated, a similarity between each linear prediction coefficient is determined based on the calculated distance between the linear prediction coefficients, and a plurality of linear prediction coefficients having consecutively similar values If there is, a representative linear prediction coefficient that is representative of the plurality of linear prediction coefficients is calculated, a process of replacing the plurality of linear prediction coefficients with the calculated representative linear prediction coefficient, and the calculated linear prediction coefficient And a linear prediction filter generated using the linear prediction coefficient replaced with the representative linear prediction coefficient, and an excitation signal is input to the generated linear prediction filter as a drive signal. And forming the voice, it is characterized in that error between the speech signal to select an excitation signal that minimizes.

  The speech decoding method according to claim 8 is obtained by replacing a plurality of linear prediction coefficients having similar values in succession with linear representative prediction coefficients among linear prediction coefficients calculated in a predetermined minute unit from a speech signal. A linear prediction filter is generated from the generated new linear prediction coefficient, an excitation signal generated from the encoded speech signal is input to the linear prediction filter, and a synthesized speech is output.

  According to the present invention, it is possible to minimize the influence on sound quality and reduce the amount of generated codes by replacing linear prediction coefficients having similar values in succession with representative linear prediction coefficients.

  Further, when determining the similarity of the linear prediction coefficients, the distance between the linear prediction coefficients is calculated with higher accuracy by calculating the distance between the linear prediction coefficients after converting the linear prediction coefficient into the LSP coefficient. It becomes possible.

  Furthermore, when the change in the time axis of the linear prediction coefficient (temporal change) is abrupt, it is output as it is without performing the thinning process and the interpolation process of the linear prediction coefficient, thereby efficiently finding a place where deterioration is severe. As a result, the sound quality can be improved.

Embodiments of the present invention will be described below with reference to the drawings.
First, the configuration in the present embodiment will be described.

  FIG. 1 shows the configuration of a speech encoding apparatus 100 according to this embodiment. As shown in FIG. 1, the speech encoding apparatus 100 includes a linear prediction analysis unit 1, a representative linear prediction coefficient calculation unit 2, an interpolation unit 3, a linear prediction filter unit 4, an error calculation unit 5, an adaptive codebook search unit 6, A noise codebook search unit 7, amplifiers 8 and 9, and a synthesis unit 10 are included.

ここで、｛ａ_i|i＝1,…,n｝が線形予測係数である。 The linear prediction analysis unit 1 performs linear prediction analysis in a predetermined minute unit (for example, every 5 ms) from the input speech signal, and calculates a linear prediction coefficient. The nth-order linear prediction filter is expressed as shown in Equation (1).

Here, {a _i | i = 1,..., N} is a linear prediction coefficient.

式（３）は、レビンソン・ダービンアルゴリズムを用いて解くことが可能である。このようにして得られたｊ番目の線形予測係数をＡ_j＝｛ａ_ji|i＝1,…,n｝とする。 One method of linear prediction analysis is an autocorrelation method. In the autocorrelation method, for example, if m samples of input signals used for linear prediction analysis are {s _i | i = 0,..., M−1}, the autocorrelation coefficient {
r _i | i = 0,..., n} is calculated as shown in equation (2), and a linear prediction coefficient {a _i | i = 1,..., n} that satisfies equation (3) is calculated.

Equation (3) can be solved using the Levinson-Durbin algorithm. The j-th linear prediction coefficient obtained in this way is assumed to be A _j = {a _ji | i = 1,..., N}.

The representative linear prediction coefficient calculation unit 2 includes a plurality of linear prediction coefficients having consecutively similar values in the linear prediction coefficient sequence {A _j | j = 1,..., M} obtained by the linear prediction analysis unit 1. If there is, a representative linear prediction coefficient is calculated from the plurality of linear prediction coefficients, and each of the plurality of linear prediction coefficients is replaced with the representative linear prediction coefficient. As a method for determining whether or not the linear prediction coefficients are similar, there is a method of measuring the Euclidean distance between the linear prediction coefficients. The calculation of the Euclidean distance and the calculation of the representative linear prediction coefficient are performed by converting them into LSP (Line Spectrum Pair) coefficients that are stable even after interpolation. Let Lbi coefficients obtained by converting the linear prediction coefficients {a _i | i = 1,..., N} be {b _i | i = 1,. This conversion is obtained by solving an algebraic equation (see Non-Patent Document 1).

式（４）のｄが、予め設定された固定値よりも小さい場合に、線形予測係数Ａ_jとＡ_kは類似していると判断される。ＬＳＰ係数列｛Ｂ_j|j＝1,…,m｝の中で、Ｂ_p＝｛ｂ_pi|i＝1,…,n｝〜Ｂ_q＝｛ｂ_qi|i＝1,…,n｝（ｐ＜ｑ）が類似していると判断された場合、これらを代表するＬＳＰ係数Ｂ_T＝｛ｂ_Ti|i＝1,…,n｝は式（５）のように算出される。

これにより、連続して類似した値を有する複数の線形予測係数Ａ_p〜Ａ_qが代表線形予測係数Ａ_Tに置き換えられる。 The Euclidean distance d between the LSP coefficient B _j = {b _ji | i = 1,..., N} and B _k = {b _ki | i = 1,..., N} is calculated as in equation (4).

When d in Expression (4) is smaller than a preset fixed value, it is determined that the linear prediction coefficients A _j and A _k are similar. In the LSP coefficient sequence {B _j | j = 1,..., M}, B _p = {b _pi | i = 1,..., N} to B _q = {b _qi | i = 1 _,. When it is determined that (p <q) are similar, LSP coefficients B _T = {b _Ti | i = 1,..., N} representing them are calculated as shown in Equation (5).

Thereby, a plurality of linear prediction coefficients A _{p to} A _q having similar values in succession are replaced with the representative linear prediction coefficient A _T.

In the above description, when the Euclidean distance between the two LSP coefficients is smaller than a preset fixed value, it is determined that the corresponding two linear prediction coefficients are similar. However, the Euclidean distance is constant. Similarity may be determined based on whether or not the ratio is exceeded. For example, as shown in Equation (6), for LSP coefficient B _j = {b _ji | i = 1,..., N}, B _j and B _k = {b _ki | i = 1 _,. Is smaller than a preset ratio r, it is determined that the linear prediction coefficients A _j and A _k are similar.

In this embodiment, every other linear prediction coefficient is encoded (that is, A _j−1 , A _{j + 1} , A _{j + 3} ,...) Is encoded in order to reduce the amount of generated code. 3 generates a linear prediction coefficient to be thinned out by interpolating from the preceding and subsequent linear prediction coefficients, and outputs the linear prediction coefficient after the interpolation processing to the linear prediction filter unit 4. As an interpolation method, for example, the linear prediction coefficients before and after the linear prediction coefficient to be thinned out are converted into LSP coefficients, the coefficients before and after the conversion are added, divided by 2, and then returned to the linear prediction coefficient. Do.

In this case, the interpolation unit 3, the linear prediction coefficients A _p of thinning-out object, calculates the Euclidean distance between the linear prediction coefficients A _p-1 of the previous time, and the linear prediction coefficients of the thinning-out object A _p, time Specifically, the Euclidean distance with the subsequent linear prediction coefficient A _{p + 1} is calculated, and it is determined whether or not the difference between the calculated distances exceeds a predetermined value, and is equal to or less than the predetermined value. In such a case, a thinning process is performed. If the predetermined value is exceeded, thinning is not performed. The calculation method of the Euclidean distance is the same as that in Equation (4).

  The linear prediction filter unit 4 synthesizes (generates) a linear prediction filter according to the equation (1) from the linear prediction coefficient finally obtained by replacement with the representative linear prediction coefficient or interpolation of the linear prediction coefficient. In addition, the linear prediction filter unit 4 performs filtering processing by the linear prediction filter on the adaptive code of the adaptive codebook in the adaptive codebook search processing (see FIG. 4), and outputs the result to the error calculation unit 5. The linear prediction filter unit 4 combines the noise code of the noise codebook and the excitation signal finally obtained by the adaptive codebook search process (see FIG. 4) in the noise codebook search process (see FIG. 5). The processed signal is filtered by the linear prediction filter and output to the error calculation unit 5. The output of the linear prediction filter unit 4 is a synthesized speech calculated by inputting an excitation signal as a drive signal to the linear prediction filter.

  The error calculation unit 5 calculates an error between the speech signal input to the speech encoding apparatus 100 and the signal after the filtering process in the linear prediction filter unit 4, and the adaptive codebook search unit 6 and the noise codebook search unit 7 is output.

  The adaptive codebook search unit 6 has an adaptive codebook storing the excitation signals used so far, takes out the adaptive code from the adaptive codebook, and performs filtering by the linear prediction filter unit 4 calculated by the error calculation unit 5 An adaptive code is selected such that the error between the processed adaptive code and the input speech signal is the smallest among the errors obtained so far (see FIG. 4). The adaptive codebook search unit 6 adds the finally obtained excitation signal to the adaptive codebook after the adaptive codebook search process (see FIG. 4) and the noise codebook search process (see FIG. 5). Then, the adaptive codebook is updated.

  The noise codebook search unit 7 has a noise codebook storing a white noise signal (noise code), extracts a noise code from the noise codebook, and performs filtering by the linear prediction filter unit 4 calculated by the error calculation unit 5 A noise signal is selected such that the error between the processed signal and the input audio signal is the smallest among the errors obtained so far (see FIG. 5). Here, the signal to be filtered by the linear prediction filter unit 4 is a signal obtained by adding the excitation signal determined in the adaptive codebook search process to the noise code extracted from the noise codebook.

  The amplifiers 8 and 9 amplify (adjust) the amplitude values of the adaptive code extracted from the adaptive codebook and the noise code extracted from the noise codebook, respectively, with a predetermined amplification factor.

  The synthesizer 10 synthesizes the amplified noise code extracted from the noise codebook and the amplified adaptive code determined as the excitation signal in the adaptive codebook search process.

  The speech coding apparatus 100 configured as described above includes the adaptive codebook index and the noise codebook index that are finally obtained as excitation signals in the adaptive codebook search process and the noise codebook search process. A signal representing the linear prediction coefficient thinned out and the amplification factor in the amplifiers 8 and 9 is output as an encoded signal.

  FIG. 2 shows a configuration of speech decoding apparatus 200 according to the embodiment of the present invention. Speech decoding apparatus 200 is an apparatus for decoding the signal encoded by speech encoding apparatus 100. As shown in FIG. 2, adaptive codebook search unit 21, noise codebook search unit 22, amplifier 23, 24, a synthesis unit 25, an interpolation unit 26, and a linear prediction filter unit 27.

  The adaptive codebook search unit 21 searches and extracts an adaptive code corresponding to the index of the input adaptive codebook from the adaptive codebook, and outputs it to the amplifier 23.

  The noise codebook search unit 22 extracts a noise code corresponding to the input index of the noise codebook from the noise codebook and outputs it to the amplifier 24.

  The amplifiers 23 and 24 amplify the input adaptive code and noise code, respectively, and output them to the synthesis unit 25. The synthesizer 25 synthesizes the adaptive code and the noise code input from the amplifiers 23 and 24, respectively.

  When there is a linear prediction coefficient that has not been input as an encoded signal, the interpolation unit 26 performs interpolation processing on the linear prediction coefficient, and outputs the linear prediction coefficient after the interpolation processing to the linear prediction filter unit 27. For the interpolation processing in the interpolation unit 26, a method similar to the interpolation method in the interpolation unit 3 of the speech encoding apparatus 100 can be applied.

  The linear prediction filter unit 27 synthesizes (generates) a linear prediction filter from the input linear prediction coefficient according to Expression (1), and performs filtering processing on the input excitation signal using the generated linear prediction filter. To generate and output a synthesized speech.

Next, the operation in this embodiment will be described.
First, with reference to the flowchart of FIG. 3, a speech encoding process executed in the speech encoding apparatus 100 will be described.

  First, a linear prediction analysis is performed on a speech signal input to the speech coding apparatus 100, and a linear prediction coefficient is calculated (step S1). Next, in the case where there are a plurality of linear prediction coefficients having consecutively similar values among the linear prediction coefficients calculated in step S1, a representative linear prediction coefficient is calculated from the plurality of linear prediction coefficients according to Equation (5). Each of the plurality of linear prediction coefficients is replaced with the representative linear prediction coefficient to obtain a new linear prediction coefficient (step S2).

  Next, a linear prediction coefficient to be thinned out is generated by interpolation from the preceding and subsequent linear prediction coefficients (step S3). Next, a linear prediction filter is synthesized according to the equation (1) from the linear prediction coefficients finally obtained by the processes of steps S1 to S3 (step S4).

  Next, adaptive codebook search processing and noise codebook search processing are performed in which an excitation signal that minimizes an error from the speech signal input to speech coding apparatus 100 is searched from the adaptive codebook and noise codebook ( Steps S5 and S6). The adaptive codebook search process in step S5 and the noise codebook search process in step S6 will be described later in detail with reference to FIGS. 4 and 5, respectively.

  When the adaptive codebook search process and the noise codebook search process are completed, the adaptive codebook is updated by adding the excitation signal obtained by these processes to the adaptive codebook (step S7), and the adaptation representing the excitation signal The codebook index, the noise codebook index, and every other linear prediction coefficient thinned out are output as encoded signals, and the speech encoding process ends.

  Next, the adaptive codebook search process (step S5 in FIG. 3) will be described with reference to the flowchart in FIG.

  First, the first adaptive code is extracted from the adaptive codebook and set as the processing target adaptive code (step S11). Next, it is determined whether or not the processing for all adaptive codes in the adaptive codebook has been completed (step S12). If it is determined in step S12 that the process has not been completed (step S12; NO), the current processing target adaptive code is subjected to a filtering process using the linear prediction filter synthesized in step S4 ( Step S13).

  Next, an error between the filtered signal and the input audio signal is calculated (step S14), and whether or not the calculated error is the smallest among the errors obtained after the start of the search process. Is determined (step S15).

  If it is determined in step S15 that the error is not the minimum (step S15; NO), the next adaptive code in the adaptive codebook is set as a processing target (step S17), and steps S12 to S12 are performed on the adaptive code. The process of S16 is repeated.

  If it is determined in step S15 that the error is the minimum (step S15; YES), the current adaptive code to be processed is set as an excitation signal candidate (step S16). Next, the next adaptive code in the adaptive codebook is set as a processing target (step S17), and the processes of steps S12 to S16 are repeated for the adaptive code.

  When the processing in steps S13 to S16 is completed for all the adaptive codes in the adaptive codebook (step S12; YES), the adaptive codebook search process is terminated, and finally the index of the adaptive code remaining as the excitation signal candidate is code. Is selected as the data of the digitized signal.

  Next, the noise codebook search process (step S6 in FIG. 3) will be described with reference to the flowchart in FIG.

  First, the first noise code is extracted from the noise code book and set as a noise code to be processed (step S21). Next, it is determined whether or not the processing for all the noise codes in the noise codebook has been completed (step S22). If it is determined in step S22 that the process has not been completed (step S22; NO), the adaptive code finally set as the excitation signal in the adaptive codebook search process of FIG. The codes are synthesized (step S23), and the combined signal is subjected to filtering processing by the linear prediction filter synthesized in step S4 (step S24).

  Next, an error between the filtered signal and the speech signal input to speech encoding apparatus 100 is calculated (step S25), and the calculated error is among the errors obtained after the start of the search process. It is determined whether or not it is minimum (step S26).

  If it is determined in step S26 that the error is not minimum (step S26; NO), the next noise code in the noise codebook is set as a processing target (step S28), and steps S22 to S22 are performed on the noise code. The process of S27 is repeated.

  If it is determined in step S26 that the error is minimum (step S26; YES), the current noise code to be processed is set as an excitation signal candidate (step S27). Next, the next noise code of the noise codebook is set as a processing target (step S28), and the processing of steps S22 to S27 is repeated for the noise code.

  When the processing of steps S23 to S27 is completed for all the noise codes in the noise codebook (step S22; YES), the present noise codebook search process is terminated, and finally the noise code index remaining as the excitation signal candidate is encoded. Is selected as the data of the digitized signal.

  Next, speech decoding processing executed in the speech decoding apparatus 200 will be described with reference to the flowchart of FIG.

  First, an adaptive code corresponding to the index of the adaptive codebook included in the input encoded signal is extracted from the adaptive codebook (step T1), and the noise codebook included in the encoded signal is extracted from the noise codebook. Is extracted (step T2), and an excitation signal is generated from the extracted adaptive code and noise code.

  Next, the adaptive codebook is updated by adding the generated excitation signal to the adaptive codebook (step T3). Next, when there is a linear prediction coefficient that has not been input as an encoded signal, interpolation processing of the linear prediction coefficient is performed (step T4). Next, a linear prediction filter is synthesized according to the equation (1) from the linear prediction coefficients obtained up to step T4 (step T5).

  Next, the reproduced speech is synthesized by performing filtering processing on the generated excitation signal using the linear prediction filter synthesized in step T5 (step T6), and the speech decoding processing is completed.

  As described above, according to the speech encoding apparatus 100 and the speech decoding apparatus 200 of the present embodiment, the influence on the sound quality is minimized by replacing the linear prediction coefficient having continuously similar values with the representative linear prediction coefficient. It is possible to reduce the generated code amount.

  Further, when determining the similarity of linear prediction coefficients and determining the presence or absence of thinning processing and interpolation processing, linear prediction coefficients are converted into LSP coefficients, and then the Euclidean distance between the linear prediction coefficients is calculated. It becomes possible to calculate the Euclidean distance between the prediction coefficients with higher accuracy.

  Furthermore, when the change in the time axis of the linear prediction coefficient (temporal change) is abrupt, it is output as it is without performing the thinning process and the interpolation process of the linear prediction coefficient, thereby efficiently finding a place where deterioration is severe. As a result, the sound quality can be improved.

The block diagram which shows the structure of the audio | voice coding apparatus which concerns on embodiment of this invention. The block diagram which shows the structure of the audio | voice decoding apparatus which concerns on embodiment of this invention. The flowchart which shows the audio | voice encoding process performed in the audio | voice encoding apparatus of this embodiment. The flowchart which shows an adaptive codebook search process. The flowchart which shows a noise codebook search process. The flowchart which shows the audio | voice decoding process performed in the audio | voice decoding apparatus of this embodiment. The block diagram which shows the structure of the conventional audio | voice encoding apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Linear prediction analysis part 2 Representative linear prediction coefficient calculation part 3 Interpolation part 4 Linear prediction filter part 5 Error calculation part 6 Adaptive codebook search part 7 Noise codebook search part 8, 9, 23, 24 Amplifier 10 Synthesis | combination part 21 Adaptive code Book search unit 22 Noise codebook search unit 25 Synthesis unit 26 Interpolation unit 27 Linear prediction filter unit 100 Speech encoding device 200 Speech decoding device

Claims (8)

A speech encoding device that encodes input speech using a linear prediction filter and an excitation signal,
Linear prediction coefficient calculating means for calculating a linear prediction coefficient in a predetermined minute unit from the input speech signal;
Distance calculating means for calculating a distance between the linear prediction coefficients calculated by the linear prediction coefficient calculating means;
The similarity between the linear prediction coefficients is determined based on the distance between the linear prediction coefficients calculated by the distance calculation means, and when there are a plurality of linear prediction coefficients having similar values in succession, the plurality of linear prediction coefficients Representative linear prediction coefficient calculating means for calculating a representative linear prediction coefficient that represents a coefficient, and performing processing for replacing the plurality of linear prediction coefficients with the calculated representative linear prediction coefficient;
Generating means for generating a linear prediction filter using the linear prediction coefficients obtained by the linear prediction coefficient calculating means and the representative linear prediction coefficient calculating means;
Selecting means for selecting an excitation signal that minimizes an error between the synthesized speech calculated by inputting the excitation signal as a drive signal to the linear prediction filter generated by the generating means and the input speech signal; ,
A speech encoding apparatus comprising:   The distance calculation unit converts the linear prediction coefficient calculated by the linear prediction coefficient calculation unit into an LSP coefficient, and calculates a distance between the linear prediction coefficients using a distance between the LSP coefficients. Item 2. The speech encoding device according to Item 1.   The said representative linear prediction coefficient calculation means determines as having a similar value, when the distance between linear prediction coefficients is smaller than the preset fixed value, The characteristic of Claim 1 or 2 characterized by the above-mentioned. Speech encoding device.   The representative linear prediction coefficient calculation means determines that the specific linear prediction coefficient has a similar value when the distance between the linear prediction coefficients is smaller than a preset ratio. Item 3. The speech encoding device according to Item 1 or 2. A determination unit that determines whether or not a change in a time axis of the linear prediction coefficient obtained by the linear prediction coefficient calculation unit and the representative linear prediction coefficient calculation unit is larger than a predetermined value;
Interpolating means for interpolating the linear prediction coefficient thinned out at a predetermined interval when the determination means determines that the change in the linear prediction coefficient is equal to or less than a predetermined value;
5. The linear prediction coefficient thinning-out process and the interpolation process by the interpolation unit are not performed when the determination unit determines that the change in the linear prediction coefficient is larger than a predetermined value. The speech encoding device according to claim 1. Among the linear prediction coefficients calculated from a speech signal in a predetermined minute unit, a linear prediction filter is obtained from a new linear prediction coefficient obtained by replacing a plurality of linear prediction coefficients having similar values in succession with a representative linear prediction coefficient. Generating means for generating;
Output means for inputting the excitation signal generated from the encoded speech signal to the linear prediction filter generated by the generating means and outputting the synthesized speech;
A speech decoding apparatus comprising: A speech encoding method for encoding a speech signal with a linear prediction filter and an excitation signal,
Calculating a linear prediction coefficient in a predetermined minute unit from the speech signal;
Calculating the distance between the calculated linear prediction coefficients;
Based on the distance between the calculated linear prediction coefficients, the similarity of each linear prediction coefficient is determined, and when there are a plurality of linear prediction coefficients having similar values in succession, a representative of the plurality of linear prediction coefficients and A representative linear prediction coefficient is calculated, and the plurality of linear prediction coefficients are replaced with the calculated representative linear prediction coefficient.
Generating a linear prediction filter using the calculated linear prediction coefficient and the linear prediction coefficient replaced with the representative linear prediction coefficient;
A speech encoding method, wherein an excitation signal that minimizes an error between a synthesized speech calculated by inputting an excitation signal as a drive signal to the generated linear prediction filter and the speech signal is selected. . Among the linear prediction coefficients calculated from a speech signal in a predetermined minute unit, a linear prediction filter is obtained from a new linear prediction coefficient obtained by replacing a plurality of linear prediction coefficients having similar values in succession with a representative linear prediction coefficient. Generate
A speech decoding method, wherein an excitation signal generated from an encoded speech signal is input to the generated linear prediction filter and a synthesized speech is output.

Images