JP2001222298A - Voice encode method and voice decode method and its device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve valuational precision of a noisy degree of a voice and to reproduce a high quality voice. SOLUTION: This device is provided with a noisy degree valuation means 25 for valuating a noisy degree by using the code or an encode result of a gain for an adaptive code vector outputted from an adaptive code book 9 and for a drive code vector outputted from a first drive code book 11 and a second drive code book 12 as one of elements and a noisy degree valuation means 26 for valuating the noisy degree by using the code or a decode result of the gain for an adaptive code vector outputted from an adaptive code book 19 and for a drive code vector outputted from the first drive code book 21 and the second drive code book 22 as one of the elements.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio encoding method and an audio decoding method for compressing and encoding an audio signal into a digital signal and an apparatus therefor, and more particularly to an audio for reproducing high quality audio at a low bit rate. The present invention relates to an encoding method, a speech decoding method, and an apparatus therefor.

[0002]

2. Description of the Related Art Conventionally, the degree of noise of speech in an encoding / decoding section is evaluated using a code or an encoding result, and code-driven linear prediction using a different driving codebook is performed according to the evaluation result. Code Excited Line
ar Prediction (hereinafter referred to as CELP)
As a voice encoding / decoding method, there is a method as shown in FIG.

FIG. 11 is a block diagram showing an example of the overall configuration of a speech encoding / decoding apparatus for implementing a conventional speech encoding method and speech decoding method. In FIG. 2 is a decoding unit that decodes speech based on the parameters, and 3 is a code of the linear prediction parameter encoded by the encoding unit 1 and minimizes distortion between the input speech and the encoded speech. Multiplexing means 4 for multiplexing the adaptive code, the driving code, and the gain code to output the result as an encoding result S2, and minimizes the code of the linear prediction parameter, the distortion between the input speech and the encoded speech from the encoding result S2. This is a separating means for separating the adaptive code, the drive code, and the gain code.

[0004] The encoding unit 1 comprises a linear prediction parameter analysis means 5, a linear prediction parameter encoding means 6, a synthesis filter 7, a noise degree evaluation means 8, an adaptive codebook 9, a driving codebook switching means 10, a first drive. Codebook 11, second drive codebook 1
2. It comprises a gain encoding means 13, a weighting and adding means 14, a distance calculating means 15, and the like.

[0005] The linear prediction parameter analysis means 5 analyzes the input speech S1 and extracts a linear prediction parameter which is speech spectrum information.

[0006] The linear prediction parameter encoding means 6 encodes the linear prediction parameters and sets the encoded linear prediction parameters as coefficients of the synthesis filter 7.

[0007] The synthesis filter 7 generates an encoded voice from the input drive excitation signal.

[0008] The noise degree evaluation means 8 evaluates the degree of noise in the coding section by obtaining the slope of the spectrum and the short-term prediction gain from the input coded linear prediction parameters and the pitch fluctuation from the adaptive code. The evaluation result is output.

The adaptive codebook 9 stores past driving excitation signals, and outputs a time-series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code as an adaptive code vector. .

The driving codebook switching means 10 switches the driving codebook used for encoding according to the evaluation result of the degree of noise.

The first driving codebook 11 stores a plurality of non-noise time-series vectors, and outputs a time-series vector corresponding to a driving code as a driving code vector.

The second driving codebook 12 stores a plurality of noise-like time-series vectors and outputs a time-series vector corresponding to the driving code as a driving code vector.

The gain encoding means 13 is adapted to store the adaptive codebook 9
And the first drive codebook 11 or the second drive codebook 12
, And a gain for weighting each time-series vector from.

The weighting and adding means 14 is adapted to
And the first drive codebook 11 or the second drive codebook 12
Is weighted by the gain for each time-series vector from.

The distance calculation means 15 calculates the distance between the coded voice and the input voice S1, and calculates an adaptive code that minimizes the distance.
This is to search for a driving code and a gain.

The decoding unit 2 includes a linear prediction parameter decoding unit 16, a synthesis filter 17, and a noise evaluation unit 1.
8, an adaptive codebook 19, a driving codebook switching means 20, a first driving codebook 21, a second driving codebook 22, a gain decoding means 23, a weighting and adding means 24, and the like.

The linear prediction parameter decoding means 16 decodes the linear prediction parameters from the codes of the linear prediction parameters and sets the decoded parameters as coefficients of the synthesis filter 17.

The synthesis filter 17 is for generating the output voice S3. The noise degree evaluation means 18 evaluates the degree of noise from the input decoded linear prediction parameters and the adaptive codes, and outputs the evaluation result.

The adaptive codebook 19 outputs, as an adaptive code vector, a time-series vector obtained by periodically repeating a past excitation signal corresponding to the adaptive code.

The drive codebook switching means 20 switches between the first drive codebook 21 and the second drive codebook 22 according to the result of the evaluation of the degree of noise by the noise degree evaluation means 18.

The first drive codebook 21 stores a plurality of non-noise time-series vectors and outputs a time-series vector corresponding to the drive code as a drive code vector.

The second driving codebook 22 stores a plurality of noise-like time-series vectors and outputs a time-series vector corresponding to the driving code as a driving code vector.

The gain decoding means 23 decodes the gain from the sign of the gain. Weight addition means 2
4 weights and adds the respective time-series vectors from the adaptive codebook 19 and the first driving codebook 21 or the second driving codebook 22 according to the respective gains decoded from the codes of the gains. Output.

Next, the operation will be described. In CELP audio coding, a period of about 5 to 50 msec is defined as one frame, and the audio of the frame is encoded separately into spectrum information and sound source information. First, in the encoding unit 1, the linear prediction parameter analysis unit 5 analyzes the input speech S1,
A linear prediction parameter, which is speech spectrum information, is extracted. The linear prediction parameter encoding means 6 encodes the linear prediction parameter, sets the encoded linear prediction parameter as a coefficient of the synthesis filter 7,
Output to the noise level evaluation means 8.

Next, encoding of the sound source information will be described. The adaptive codebook 9 stores past driving excitation signals, and outputs a time-series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code as an adaptive code vector. For example, as shown in FIG. 12, the noise degree evaluation means 8 obtains a spectrum inclination and a short-term prediction gain from the coded linear prediction parameters inputted from the linear prediction parameter coding means 6, and obtains a pitch variation from the adaptive code. Then, the degree of noise in the coding section is evaluated, and the evaluation result is output to the driving codebook switching means 10. The drive codebook switching means 10 switches the first drive codebook 11 according to the evaluation result of the degree of noise, for example, if the degree of noise is small,
If the degree of noise is large, the second drive codebook 12 is used, and the drive codebook used for encoding is switched.

The first driving codebook 11 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning so as to reduce distortion between the learning speech and its encoded speech. The vector is stored. The second driving codebook 12 stores a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise. Output as a vector. Adaptive codebook 9
And the first drive codebook 11 or the second drive codebook 12
Are weighted and added by the weighting and adding means 14 according to the respective gains given from the gain coding means 13, and the result of the addition is supplied to the synthesis filter 7 as a driving sound source signal and the coded speech Generate

The distance calculating means 15 finds the distance between the coded speech and the input speech S1, and searches for an adaptive code, a drive code and a gain that minimize the distance. After the above encoding is completed,
The code of the linear prediction parameter, the adaptive code for minimizing the distortion between the input voice and the coded voice, the driving code, and the code of the gain are multiplexed by the multiplexing means 3 and output as a coding result S2.

On the other hand, in the decoding section 2, the code of the linear prediction parameter, the adaptive code, the driving code, and the code of the gain, which have been multiplexed and sent, are separated by the separating means 4. Then, the linear prediction parameter decoding unit 16 of the decoding unit 2 decodes the linear prediction parameter from the sign of the linear prediction parameter, sets the coefficient as a coefficient of the synthesis filter 17, and outputs the coefficient to the noise evaluation unit 18. Next, adaptive codebook 19 outputs, as an adaptive code vector, a time-series vector obtained by periodically repeating a past excitation signal corresponding to the adaptive code. The noise degree evaluation unit 18 calculates the degree of noise from the decoded linear prediction parameters input from the linear prediction parameter decoding unit 16 and the adaptive code in the same manner as the noise degree evaluation unit 8 of the encoding unit 1. And outputs the evaluation result to the driving codebook switching means 20.

The driving codebook switching means 20, like the driving codebook switching means 10 of the encoding unit 1, responds to the evaluation result of the degree of noise by the noise degree evaluating means 18. And the second drive codebook 22.

The first driving codebook 21 has a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning such that distortion between the learning speech and its encoded speech is reduced. Are stored in the second drive codebook 22, and a plurality of time-series vectors, such as a plurality of time-series vectors generated from random noise, are stored in the second drive codebook 22. The sequence vector is output as a driving code vector.

The adaptive codebook 19 and the first driving codebook 21
Alternatively, each time-series vector from the second driving codebook 22 is weighted and added by the weighting and adding means 24 according to each gain decoded from the gain code by the gain decoding means 23, and the addition result is The output sound S3 is supplied to the synthesis filter 17 as a driving sound source signal and generates an output sound S3.

[0032]

Since the conventional speech coding method and speech decoding method and the related apparatus are constructed as described above, the degree of noise is determined based only on the spectral characteristics and fluctuations of the signal period. Therefore, if a noise signal that closely resembles an audio signal with respect to a certain short time interval, a change in spectral characteristics or a signal cycle, such as an automobile engine noise, is actually input, There is a problem in that an erroneous evaluation is made that the degree of noise is small despite the section having a large degree.

Further, in the speech coding method and speech decoding method and apparatus thereof, since the driving codebooks having different properties are switched and used, the evaluation of the degree of noise used for the switching decision is erroneous, and When an inappropriate driving codebook is used in a coding and decoding section, there is a problem that the quality of coding and decoding speech is greatly deteriorated even if it is a local short section.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and it is an object of the present invention to provide a speech coding method and a speech decoding method capable of improving the evaluation accuracy of the degree of noise of speech and reproducing high quality speech. The aim is to obtain a method and its device.

[0035]

According to the speech encoding method of the present invention, one of the parameters used for evaluating the degree of noise is used.
One is to use the sign or coding result of the gain for the adaptive code vector and the driving code vector.

A speech encoding method according to the present invention analyzes a driving excitation signal or an encoded speech obtained by linearly predicting and synthesizing a driving excitation signal, and sets a long-period prediction gain obtained as a result of the analysis as one of parameters. To evaluate the degree of noise.

In the speech coding method according to the present invention, a past evaluation result of the degree of noise is used as one of the parameters for evaluating the degree of noise.

In the speech coding method according to the present invention, a plurality of driving codebooks having different degrees of noise of stored time-series vectors are switched and used in accordance with the evaluation result of the degree of noise of speech. It was made.

In the speech coding method according to the present invention, the degree of noise of the time-series vector stored in the driving codebook is changed according to the evaluation result of the degree of noise of speech. is there.

In the speech coding method according to the present invention, the codebook of the gain is switched according to the evaluation result of the degree of noise of the speech.

In the speech decoding method according to the present invention, the coding or decoding result of the gain for the adaptive code vector and the driving code vector is used as one of the parameters for evaluating the degree of noise. .

A speech decoding method according to the present invention analyzes a driving excitation signal or a decoded speech obtained by linearly predicting and synthesizing a driving excitation signal, and determines a long-period prediction gain obtained as a result of determining a degree of noise. This is used as one of the parameters for evaluation.

In the speech decoding method according to the present invention, a past evaluation result of the degree of noise is used as one of the parameters for evaluating the degree of noise.

In the speech decoding method according to the present invention, a plurality of driving codebooks having different degrees of noise of stored time-series vectors are switched and used in accordance with the evaluation result of the degree of noise of speech. It was made.

In the speech decoding method according to the present invention, the degree of noise of the time-series vector stored in the driving codebook is changed according to the evaluation result of the degree of noise of speech. is there.

In the speech decoding method according to the present invention, the gain codebook is switched according to the evaluation result of the degree of noise in speech.

According to the speech encoding / speech decoding apparatus of the present invention, the past driving excitation signal is stored, and the time series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code is adaptively encoded. An adaptive codebook that outputs as a vector, a plurality of non-noise time-series vectors are stored, and a first driving codebook that outputs a time-series vector corresponding to the driving code as a driving code vector; A second driving codebook that stores a time-series vector and outputs a time-series vector corresponding to the driving code as a driving code vector; and a code or encoding and decoding of a gain for the adaptive code vector and the driving code vector. Noise level evaluation means for evaluating the level of noise level by using the result as one of the elements for obtaining the evaluation value of the level of noise level. It is.

A speech encoding / speech decoding apparatus according to the present invention is characterized in that a past driving excitation signal is stored, and a time series vector obtained by periodically repeating the past driving excitation signal corresponding to an adaptive code is adaptively encoded. An adaptive codebook that outputs as a vector, a plurality of non-noise time-series vectors are stored, and a first driving codebook that outputs a time-series vector corresponding to the driving code as a driving code vector; A time series vector is stored, and a second drive codebook that outputs a time series vector corresponding to the drive code as a drive code vector, and the past drive excitation signal or the drive excitation signal is obtained by linear prediction synthesis. A long-period prediction gain calculating means for analyzing the encoded voice and obtaining the long-period prediction gain;
Noise level evaluation means for evaluating the degree of noise by using the long-period prediction gain obtained by the long-period prediction gain calculation means as one of the elements for obtaining an evaluation value of the degree of noise. It was done.

According to the speech encoding / speech decoding apparatus of the present invention, the past driving excitation signal is stored, and the time series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code is adaptively encoded. An adaptive codebook that outputs as a vector, a plurality of non-noise time-series vectors are stored, and a first driving codebook that outputs a time-series vector corresponding to the driving code as a driving code vector; A second drive codebook in which a time-series vector is stored and a time-series vector corresponding to the drive code is output as a drive code vector, and an evaluation result of the degree of noise is calculated based on the evaluation result of the degree of noise in the past. Noise evaluation means for evaluating the degree of noise using one of the elements for noise reduction.

The speech coding / speech decoding apparatus according to the present invention comprises a driving codebook switching means for switching a driving codebook according to the degree of noise evaluated by the noise degree evaluation means.

The speech coding / speech decoding apparatus according to the present invention changes the degree of noise of the time-series vector stored in the driving codebook according to the degree of noise evaluated by the noise evaluation unit. This is provided with a noise change changing means.

The speech encoding / speech decoding apparatus according to the present invention stores a past excitation signal and stores a time series vector obtained by periodically repeating the past excitation signal in accordance with an adaptive code. An adaptive codebook that outputs as a vector, a plurality of noise-like time-series vectors are stored, a driving codebook that outputs a time-series vector corresponding to the driving code as a driving code vector, the adaptive code vector and the driving code Noise level evaluation means for evaluating the degree of noise by using the sign or encoding / decoding result of the gain for the vector as one of the elements for obtaining the evaluation value of the degree of noise; The degree of noise is changed by setting the amplitude value of the low-amplitude sample of the time series vector stored in the driving codebook to zero according to the evaluated degree of noise. Thereby it is obtained so as to include a noisy variation inducing means.

[0053]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below. Embodiment 1 FIG. FIG. 1 is a block diagram showing the overall configuration of a speech encoding / speech decoding apparatus for realizing the speech encoding method and speech decoding method according to the first embodiment. In the figure, reference numeral 1 denotes an encoding unit that encodes audio parameters;
2 is a decoding unit that decodes speech based on the parameters, 3 is a code of the encoded linear prediction parameter, an adaptive code that minimizes distortion between the input speech and the encoded speech, a driving code, and a gain. A multiplexing means 4 for multiplexing the code and outputting the result as a coding result S2, a code for the linear prediction parameter, an adaptive code for minimizing distortion between the input voice and the coded voice, a driving code, and a gain from the coding result S2. Is a separating means for separating the sign of.

5 is a linear prediction parameter analysis means,
Reference numeral 6 denotes a linear prediction parameter coding unit, and reference numeral 7 denotes a synthesis filter. The linear prediction parameter analysis means 5 outputs the input speech S
1 is analyzed to extract a linear prediction parameter which is spectrum information of speech. The linear prediction parameter encoding means 6 encodes the linear prediction parameters and sets the encoded linear prediction parameters as coefficients of the synthesis filter 7. The synthesis filter 7 generates a coded voice.

Reference numerals 9 and 19 are adaptive codebooks storing past excitation signals, 11 and 21 are first driving codebooks storing a plurality of non-noise time-series vectors, and 12 and 22 are noise-like. The second driving codebooks 25 and 26 in which a plurality of time-series vectors are stored are noise level evaluation means for evaluating the degree of noise using an adaptive code vector and a gain for the driving code vector, and 10 and 20 are noise levels. This is a driving codebook switching means for switching the driving codebook according to the degree of sex.

Reference numeral 13 denotes gain coding means, 14 denotes weighting and adding means, and 15 denotes distance calculating means. 16 is a linear prediction parameter decoding means, 17 is a synthesis filter, 2
3 is a gain decoding means, and 24 is a weighting addition means.

The gain encoding means 13 is adapted to
And the first drive codebook 11 or the second drive codebook 12
, And a gain for weighting each time-series vector from. In this case, as shown in FIG. 2, if the noise level of the section to be encoded is large, the gain at that time becomes unstable, and if the noise level is low, that is, if the noise is voice, the gain is stable. Becomes

The weighting and adding means 14 is adapted to
And the first drive codebook 11 or the second drive codebook 12
Is weighted by the gain for each time-series vector from.

The distance calculation means 15 calculates the distance between the coded voice generated by the synthesis filter 7 and the input voice S1, and searches for an adaptive code, a driving code and a gain that minimize the distance.

The linear prediction parameter decoding means 16 decodes the linear prediction parameters from the codes of the linear prediction parameters, and sets them as coefficients of the synthesis filter 17.

The gain decoding means 23 decodes the gain from the sign of the gain. Weight addition means 2
4 weights and adds the respective time-series vectors from the adaptive codebook 19 and the first driving codebook 21 or the second driving codebook 22 according to the respective gains decoded from the codes of the gains. Output. The synthesis filter 17 generates the output voice S3.

Next, the operation will be described. First, in the encoding unit 1, the linear prediction parameter analysis unit 5 analyzes the input speech S1, and extracts a linear prediction parameter that is speech spectrum information. The linear prediction parameter encoding unit 6 encodes the linear prediction parameter, sets the encoded linear prediction parameter as a coefficient of the synthesis filter 7, and outputs the coefficient to the noise evaluation unit 25.

Next, encoding of the sound source information will be described. The adaptive codebook 9 stores the past drive excitation signal, and outputs a time-series vector obtained by periodically repeating the past drive excitation signal corresponding to the adaptive code as an adaptive code vector.

As shown in FIG. 2, for example, as shown in FIG. 2, the noise degree evaluation means 25 calculates the slope of the spectrum and the short-term prediction gain from the coded linear prediction parameters input from the linear prediction parameter coding means 6 and the pitch fluctuation from the adaptive code. The change of the coding result of the gain of the adaptive code vector from the sign or coding result of the gain for the adaptive code vector and the driving code vector, and the ratio and the change of the coding result of the gain of the adaptive code vector and the gain of the driving code vector. As a result, an evaluation value of the degree of noise is obtained for each of the obtained parameters, and a value obtained by weighting and averaging the evaluation values of all parameters is used as a final evaluation result of the degree of noise. The degree of noise in the coding section is evaluated, and the evaluation result is output to the driving codebook switching means 10.

In this case, as described above, the gain given by the gain encoding means 13 is stable if the section to be encoded is speech, and unstable if the noise is large. By setting the sign of the gain or the coding result of the vector and the driving code vector as one of the parameters for obtaining the evaluation of the degree of noise, the evaluation accuracy of the degree of noise is improved.

The driving codebook switching means 10 responds to the evaluation result of the degree of noise by, for example, the first driving codebook 11 if the degree of noise is low, and the second driving code if the degree of noise is high. The drive codebook used for encoding is switched assuming that the book 12 is used.

The first driving codebook 11 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning such that distortion between the learning speech and its encoded speech is reduced. The vector is stored.

The second driving codebook 12 stores a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise. Output the vector as the driving code vector.

The time series vectors from the adaptive codebook 9 and the first driving codebook 11 or the second driving codebook 12 are as follows:
The weighting and adding means 14 adds weights according to the respective gains given from the gain encoding means 13 and adds the results.
To generate encoded speech. The distance calculation means 15 finds the distance between the coded voice and the input voice S1, and searches for an adaptive code, a drive code, and a gain that minimize the distance.

After the above-mentioned encoding is completed, the code of the linear prediction parameter, the adaptive code for minimizing the distortion between the input speech S1 and the encoded speech, the driving code and the gain code are output as the encoding result S2. .

Next, the decoding section 2 will be described. In the decoding unit 2, the linear prediction parameter decoding unit 16 decodes the linear prediction parameter from the code of the linear prediction parameter, sets it as a coefficient of the synthesis filter 17, and outputs the coefficient to the noise evaluation unit 26.

Next, adaptive codebook 19 outputs, as an adaptive code vector, a time-series vector obtained by periodically repeating a past excitation signal corresponding to the adaptive code.

The noise degree evaluation means 26 comprises a decoded linear prediction parameter and adaptive code inputted from the linear prediction parameter decoding means 16, and a previously decoded adaptive code vector inputted from the gain decoding means 23. From the result of the coding or decoding of the gain with respect to the drive code vector or the decoding result, the degree of noise is evaluated by the same method as the noise degree evaluation means 25 of the encoding unit 1, and the evaluation result is output to the drive codebook switching means 20. Also in this case, the evaluation accuracy of the degree of noise in the section to be decoded is improved as in the case of the noise degree evaluation means 25.

The driving codebook switching means 20 performs the first driving codebook 21 and the second driving codebook processing in the same manner as the driving codebook switching means 10 of the encoding unit 1 in accordance with the evaluation result of the degree of noise. And 22.

The first driving codebook 21 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning such that distortion between a learning speech and its encoded speech is reduced. However, the second driving codebook 22 stores a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise. Output as Adaptive codebook 19 and first driving codebook 2
Each time-series vector from the first or second driving codebook 22 is weighted and added by the weighting and adding means 24 in accordance with each gain decoded from the gain code by the gain decoding means 23, and the addition result is obtained. Is supplied to the synthesis filter 17 as a drive sound source signal to generate an output sound S3.

As described above, according to the first embodiment, the coding and decoding can be easily performed by using the gain information for the adaptive code vector and the driving code vector in evaluating the degree of noise of the input speech. Since the information on the strength and the stability of the periodicity of the input signal in the segmentation section and its stability can be obtained and used for the evaluation of the degree of noise, the accuracy of the evaluation of the degree of noise can be improved. Further, since the driving codebook adapted to the mode of the input speech is switched and used based on this evaluation, there is an effect that a speech encoding method, a speech decoding method and a device capable of reproducing high quality speech are obtained.

Embodiment 2 FIG. 3 is a block diagram showing the overall configuration of a speech encoding / speech decoding apparatus for realizing the speech encoding method and speech decoding method according to the second embodiment. 3, the same or corresponding parts as those in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted. In the figure, 27
And 29 are long-period prediction gain calculating means for analyzing the driving excitation signal to obtain its long-period prediction gain, and 28 and 30 are noise degree evaluating means for evaluating the degree of noise using the long-period prediction gain for the driving excitation signal It is.

Next, the operation will be described. First, in the encoding unit 1, the linear prediction parameter analysis unit 5 analyzes the input speech S1 and extracts a linear prediction parameter which is spectrum information of the speech. Linear prediction parameter encoding means 6
Encodes the linear prediction parameters extracted by the linear prediction parameter analysis unit 5, sets the encoded linear prediction parameters as coefficients of the synthesis filter 7, and outputs the coefficients to the noise evaluation unit 28.

Next, encoding of the sound source information will be described. The long-period prediction gain calculating unit 27 analyzes a past drive excitation signal (for example, a signal in the immediately preceding coding section) and obtains a long-period prediction gain indicating the strength of the periodicity of the signal. When the section to be encoded is speech, the periodicity of the signal becomes strong, so that it becomes stable as shown in FIG.
When the degree of noise in the section is high, the periodicity of the signal is weakened and the signal becomes unstable.) The adaptive codebook 9 stores the previous driving excitation signal, and outputs a time-series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code as an adaptive code vector.

As shown in FIG. 4, for example, the noise degree evaluation means 28 calculates the spectrum slope and short-term prediction gain from the coded linear prediction parameters input from the linear prediction parameter coding means 6 and the pitch fluctuation from the adaptive code. Is calculated from the long-period prediction gain of the past drive excitation signal input from the long-period prediction gain calculation means 27, and the evaluation value of the degree of noise is obtained for each of the obtained parameters. The value obtained by weighted averaging of the evaluation values of
The degree of noise in the coding section is evaluated, and the evaluation result is output to the driving codebook switching means 10.

In this case, based on the stability of the long-period prediction gain according to the degree of noise in the section to be encoded, the long-period prediction gain is used as one of the parameters for evaluating the degree of noise.
By using it as one, the evaluation system improves.

The driving codebook switching means 10 responds to the result of the evaluation of the degree of noise by, for example, the first driving codebook 11 if the degree of noise is low, and the second driving code if the degree of noise is high. Assuming that the codebook 12 is used, the driving codebook used for encoding is switched.

The first driving codebook 11 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning so as to reduce distortion between the learning speech and its encoded speech. The vector is stored.

The second driving codebook 12 stores a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise. Output the vector as the driving code vector.

The time series vectors from the adaptive codebook 9 and the first driving codebook 11 or the second driving codebook 12 are as follows:
The weighting and adding means 14 adds weights according to the respective gains given from the gain coding means 13, and the result of the addition is used as a driving sound source signal by the synthesis filter 7.
To generate encoded speech.

The distance calculation means 15 finds the distance between the coded speech and the input speech S1, and searches for an adaptive code, a drive code, and a gain that minimize the distance.

After the above coding is completed, the code of the linear prediction parameter, the adaptive code for minimizing the distortion between the input voice and the coded voice, the driving code and the gain code are converted to the coding result S.
Output as 2.

Next, the decoding section 2 will be described. In the decoding unit 2, the linear prediction parameter decoding unit 16 decodes the linear prediction parameter from the sign of the linear prediction parameter, sets the parameter as a coefficient of the synthesis filter 17, and
Output to the noise degree evaluation means 30.

The long-period prediction gain calculating means 29 analyzes the past drive sound source signal, obtains a long-period prediction gain indicating the strength of the periodicity of the signal, and outputs the gain to the noise degree evaluation means 30. next,
The adaptive codebook 19 outputs a time-series vector obtained by periodically repeating a past excitation signal corresponding to the adaptive code as an adaptive code vector.

The noise degree estimating means 30 includes a decoded linear prediction parameter and an adaptive code inputted from the linear prediction parameter decoding means 16 and a past driving excitation signal inputted from the long-period prediction gain calculating means 29. From the long-period prediction gain, the degree of noise is evaluated in the same manner as the noise degree evaluation means 28 of the encoding unit 1, and the evaluation result is output to the driving codebook switching means 20. In this case, the noise degree evaluation means 2
As in the case of No. 8, the evaluation accuracy of the degree of noise in the decoding section is improved.

The driving codebook switching means 20 responds to the result of the evaluation of the degree of noise by using the first driving codebook 21 and the second driving codebook in the same manner as the driving codebook switching means 10 of the encoder 1. And 22.

The first driving codebook 21 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning so that the distortion between the learning speech and the encoded speech is reduced. Are stored in the second driving codebook 22, and a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise are stored, and the time-series vectors corresponding to the driving codes are respectively driven. Output as code vector.

Each time series vector from the adaptive codebook 19 and the first driving codebook 21 or the second driving codebook 22 is
The weights are added by weighting and adding means 24 in accordance with the respective gains decoded from the gain codes by the gain decoding means 23, and the addition result is supplied to the synthesis filter 17 as a driving sound source signal to generate an output sound S3. You.

As described above, according to the second embodiment, by using the long-period prediction gain obtained by directly analyzing the driving sound source signal to evaluate the degree of noise of the input sound, the driving sound source signal is More accurate information on the strength of the periodicity of the input signal in the encoding and decoding sections and its stability than in the case of using only information on the gain for the adaptive code vector and the driving code vector that are a part of the generated parameter. Can be obtained, and by using this for the evaluation of the degree of noise, the accuracy of evaluation of the degree of noise can be improved. Further, since the driving codebook adapted to the mode of the input voice is used based on this evaluation, there is an effect that a voice encoding method, a voice decoding method, and a device capable of reproducing high quality voice can be obtained.

Embodiment 3 In the second embodiment, the long-period prediction gain calculating means 27 and 29 analyze the driving sound source signal to obtain the long-period prediction gain.
The same effect can be obtained by analyzing the encoded speech and the decoded speech output from the synthesis filters 7 and 17 obtained by linear predictive synthesis of the driving excitation signal as shown in FIG.

Embodiment 4 FIG. 6 is a block diagram showing an overall configuration of a speech encoding / speech decoding apparatus for realizing the speech encoding method and speech decoding method according to the fourth embodiment. 6, the same or corresponding parts as those in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted. In the figure, 31
And 32 are noise level evaluation means for evaluating the level of noise level using the evaluation result of the degree of noise level in the past.

Next, the operation will be described. First, in the encoding unit 1, the linear prediction parameter analysis unit 5 analyzes the input speech S1, and extracts a linear prediction parameter which is spectrum information of the speech. The linear prediction parameter encoding unit 6 encodes the linear prediction parameter, sets the encoded linear prediction parameter as a coefficient of the synthesis filter 7, and outputs the coefficient to the noise evaluation unit 31.

Next, encoding of the sound source information will be described. The adaptive codebook 9 stores the past drive excitation signal, and outputs a time-series vector obtained by periodically repeating the past drive excitation signal corresponding to the adaptive code as an adaptive code vector.

As shown in FIG. 7, for example, as shown in FIG. 7, the noise degree evaluation means 31 calculates the spectrum inclination and the short-term prediction gain from the coded linear prediction parameters input from the linear prediction parameter coding means 6, and the pitch from the adaptive code. The degree of the noise is calculated from the result of the evaluation of the degree of the past noise inputted from the degree of noise evaluation means 31, and the evaluation value of the degree of the degree of noise (the degree of the degree of the past noise) is obtained for each of the calculated parameters. With respect to the parameter of, for example, if the degree of noise for the immediately preceding predetermined number of frames is large, it is evaluated that the degree of noise of the frame in the current coding section is also large. Assume the value obtained by weighting and averaging the evaluation values of all the parameters as the final evaluation result of the degree of noise, and evaluate the degree of noise in the coding section. And it outputs the value results in driving codebook switching means 10.

According to the evaluation result of the degree of noise, the drive sound source switching means 10 determines the first drive codebook 11 if the degree of noise is small, and the second drive code if the degree of noise is large. The drive codebook used for encoding is switched assuming that the book 12 is used.

The first driving codebook 11 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning such that distortion between a learning speech and its encoded speech is reduced. The vector is stored. Further, the second driving codebook 12 stores a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise, and drives the time-series vectors corresponding to the respective driving codes. Output as code vector.

The time series vectors from the adaptive codebook 9 and the first driving codebook 11 or the second driving codebook 12 are as follows:
The weighting and adding means 14 adds weights according to the respective gains given from the gain coding means 13, and the addition result is used as a driving sound source signal by the synthesis filter 7.
To generate encoded speech.

The distance calculation means 15 finds the distance between the coded speech and the input speech S1, and searches for an adaptive code, a drive code and a gain that minimize the distance.

After the above coding is completed, the code of the linear prediction parameter, the adaptive code for minimizing the distortion between the input voice and the coded voice, the driving code and the gain code are converted to the coding result S.
Output as 2.

Next, the decoding section 2 will be described. In the decoding unit 2, the linear prediction parameter decoding unit 16 decodes the linear prediction parameter from the code of the linear prediction parameter, sets it as a coefficient of the synthesis filter 17, and outputs the coefficient to the noise evaluation unit 32.

Next, adaptive codebook 19 outputs, as an adaptive code vector, a time-series vector obtained by periodically repeating a past excitation signal corresponding to the adaptive code.

The noise degree evaluation means 32 includes a decoded linear prediction parameter and adaptive code inputted from the linear prediction parameter decoding means 16 and a past noise degree inputted from the noise degree evaluation means 32 itself. From the evaluation result of
The degree of noise is evaluated in the same manner as the noise degree evaluation means 31 of the encoding unit 1 and the evaluation result is used as the driving codebook switching means 2
Output to 0.

The driving codebook switching means 20 operates in the same manner as the driving codebook switching means 10 of the encoding unit 1 in accordance with the result of the evaluation of the degree of noise. And 22.

The first driving codebook 21 includes a plurality of non-noise time-series vectors, for example, a plurality of time-series vectors constructed by learning so as to reduce distortion between the learning speech and its encoded speech. Are stored in the second driving codebook 22. The second driving codebook 22 stores a plurality of noise-like time-series vectors, for example, a plurality of time-series vectors generated from random noise, and drives the time-series vectors corresponding to the respective driving codes. Output as code vector.

The adaptive codebook 19 and the first driving codebook 21
Alternatively, each time-series vector from the second driving codebook 22 is weighted and added by the weighting and adding means 24 according to each gain decoded from the gain code by the gain decoding means 23, and the addition result is The output sound S3 is supplied to the synthesis filter 17 as a driving sound source signal and generates an output sound S3.

As described above, according to the fourth embodiment, the evaluation result of the degree of noise in the past is used for the evaluation of the degree of noise in the input speech. Local noise level evaluation errors can be eliminated, that is, for each successive frame to be coded and decoded, the noise level evaluation result changes from large noisy to small noisy. Since it is possible to avoid a state in which the evaluation result changes suddenly many times in a short time, such as a change to a large size again, there is an effect that the evaluation accuracy of the degree of noise is improved. In addition, since the driving codebook adapted to the form of the input voice is used based on the evaluation, there is an effect that a voice encoding method, a voice decoding method, and a device capable of reproducing high-quality voice can be obtained.

Embodiment 5 FIG. In the first to fourth embodiments, two driving codebooks are switched and used. Instead, in the fifth embodiment, three or more driving codebooks are provided, and the degree of noise is reduced. Switch and use as appropriate.

Therefore, according to the fifth embodiment, a suitable driving codebook is used for not only two types of speech, that is, noise and non-noise, but also for intermediate speech that is slightly noisy. Therefore, there is an effect that an audio encoding method, an audio decoding method, and an apparatus thereof that can reproduce high quality audio can be obtained.

Embodiment 6 FIG. FIG. 8 is a block diagram showing an overall configuration of a speech encoding / decoding apparatus according to the sixth embodiment for realizing the speech encoding method and the speech decoding method according to the sixth embodiment. 8, parts that are the same as or correspond to those in FIG. 1 are given the same reference numerals, and descriptions thereof will be omitted. In the figure, reference numerals 33 and 35 denote driving codebooks storing noise-like time-series vectors, and reference numerals 34 and 36 denote sample thinning-out means (noise-change giving means) for setting the amplitude value of low-amplitude samples of the time-series vectors to zero. is there.

Next, the operation will be described. First, in the encoding unit 1, the linear prediction parameter analysis unit 5 analyzes the input speech S1, and extracts a linear prediction parameter that is speech spectrum information. The linear prediction parameter encoding unit 6 encodes the linear prediction parameters extracted by the linear prediction parameter analysis unit 5, sets the encoded linear prediction parameters as coefficients of the synthesis filter 7,
Output to the noise degree evaluation means 25.

Next, encoding of the sound source information will be described. The adaptive codebook 9 stores past driving excitation signals, and outputs a time-series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code as an adaptive code vector.

The noise degree evaluation means 25, for example, gain slope coding and short-term prediction gain from the coded linear prediction parameter input from the linear prediction parameter coding means 6, pitch fluctuation from the adaptive code, and gain coding Means 1
3. Changes in the coding result of the gain of the adaptive code vector from the coding or coding result of the gain for the adaptive code vector and the driving code vector determined in the past input from 3 and the coding of the gain of the adaptive code vector and the driving code vector The ratio of the results and the variation thereof are obtained, the evaluation value of the degree of noise is obtained for each of the obtained parameters, and the value obtained by weighting and averaging the evaluation values of all the parameters is determined as the final degree of noise. The degree of noise in the coding section is evaluated as an evaluation result or the like, and the evaluation result is output to the sample thinning means 34.

The driving codebook 33 stores, for example, a plurality of time-series vectors generated from random noise.
A time-series vector corresponding to the drive code is output as a drive code vector.

If the degree of noise is small according to the evaluation result of the degree of noise, the sample thinning means 34 fills the time series vector input from the driving codebook 33 with, for example, a predetermined amplitude value. A time-series vector in which the amplitude value of the sample is zero is output, and if the degree of noise is large, the time-series vector input from the driving codebook 33 is output as it is.

The time series vectors from the adaptive codebook 9 and the sample thinning means 34 are added to the weighting and adding means 1 in accordance with the respective gains given from the gain coding means 13.
4 and are added, and the result of the addition is supplied to the synthesis filter 7 as a drive excitation signal to generate an encoded voice.

The distance calculation means 15 finds the distance between the coded speech and the input speech S1, and searches for an adaptive code, a driving code, and a gain that minimize the distance.

After the above coding is completed, the code of the linear prediction parameter, the adaptive code for minimizing the distortion between the input voice and the coded voice, the driving code, and the gain code are converted to the coding result S.
Output as 2.

Next, the decoding unit 2 will be described. In the decoding unit 2, the linear prediction parameter decoding unit 16 decodes the linear prediction parameter from the code of the linear prediction parameter, sets it as a coefficient of the synthesis filter 17, and outputs the coefficient to the noise evaluation unit 26.

Next, decoding of sound source information will be described. The adaptive codebook 19 outputs, as an adaptive code vector, a time-series vector obtained by periodically repeating a past excitation signal in accordance with the adaptive code. The noise level evaluation means 26
Sign or decoding of the decoded linear prediction parameter and adaptive code input from the linear prediction parameter decoding means 16 and the gain for the previously decoded adaptive code vector and driving code vector input from the gain decoding means 23 The noise degree evaluation means 25 of the encoding unit 1
Then, the degree of noise is evaluated in the same manner as described above, and the evaluation result is output to the sample thinning means 36.

The driving codebook 35 outputs a time series vector corresponding to the driving code as a driving code vector. The sample decimating unit 36 outputs a time-series vector by the same processing as that of the sample decimating unit 34 of the encoding unit 1 according to the evaluation result of the degree of noise. The time series vectors from the adaptive codebook 19 and the sample thinning means 36 are weighted and added by the weighting and adding means 24 according to the respective gains given from the gain decoding means 23, and the addition result is used as a driving excitation signal. The output sound S3 supplied to the synthesis filter 17 is generated.

As described above, according to the sixth embodiment, the information on the gain for the adaptive code vector and the driving code vector is also used for evaluating the degree of noise in the coding and decoding sections, thereby reducing the noise. This has the effect of improving the evaluation accuracy of the degree. In addition, a driving codebook that stores a noise-like time-series vector as a driving code vector is provided, and according to the evaluation result of the degree of noise of the speech in the encoding and decoding sections, the input speech is In order to conform to the aspect, a signal code of the noise-like drive code vector is thinned to generate a drive code vector with a small degree of noise, so that a small amount of information can be used to reproduce a high-quality sound. There is an effect that the method and the audio decoding method and the apparatus can be obtained. Further, since there is no need to provide a plurality of drive codebooks, there is an effect that the memory capacity required for storing the drive codebooks can be reduced.

Embodiment 7 FIG. In the sixth embodiment, the time series vector samples are decimated or not decimated. In the seventh embodiment, an amplitude threshold value for decimating the samples according to the degree of noise is used instead. For example, the sample thinning rate may be changed in a plurality of stages, such as changing the number of samples.

Therefore, according to the seventh embodiment, a time-series vector suitable for not only two types of speech, that is, noise / non-noise, but also intermediate speech such as a little noise Since it can be generated and used, there is an effect that an audio encoding method and an audio decoding method and an apparatus thereof that can reproduce high quality audio can be obtained.

Embodiment 8 FIG. FIG. 9 is a block diagram showing the overall configuration of a speech encoding / speech decoding apparatus for realizing the speech encoding method and speech decoding method according to the eighth embodiment. In the figure, reference numerals 37 and 40 denote first driving codebooks storing noise-like time-series vectors, and 38 and 41.
Is a second driving codebook that stores a non-noise time-series vector, and 39 and 42 are weight determining means (noisy change providing means).

Next, the operation will be described. First, in the encoding unit 1, the linear prediction parameter analysis unit 5 analyzes the input speech S1, and extracts a linear prediction parameter that is speech spectrum information. The linear prediction parameter encoding unit 6 encodes the linear prediction parameters extracted by the linear prediction parameter analysis unit 5, sets the encoded linear prediction parameters as coefficients of the synthesis filter 7, and outputs the coefficients to the noise evaluation unit 25. .

Next, encoding of the sound source information will be described. The adaptive codebook 9 stores past driving excitation signals, and outputs a time-series vector obtained by periodically repeating the past driving excitation signal corresponding to the adaptive code as an adaptive code vector.

The noise degree evaluation means 25 includes, for example, the spectrum inclination and short-term prediction gain from the coded linear prediction parameters input from the linear prediction parameter coding means 6, the pitch fluctuation from the adaptive code, and the gain coding means. 13
Of the gain of the adaptive code vector and the gain of the adaptive code vector and the result of the coding of the gain of the adaptive code vector and the drive code vector obtained from the past, And the variation thereof, an evaluation value of the degree of noise is obtained for each of the obtained parameters, and a value obtained by weighting and averaging the evaluation values of all parameters is the final evaluation result of the degree of noise. For example, the degree of noise in the coding section is evaluated, and the evaluation result is output to the weight determining means 39.

The first driving codebook 37 stores a plurality of noise-like time-series vectors generated from, for example, random noise, and outputs a time-series vector corresponding to the driving code as a driving code vector. Second drive codebook 38
Stores, for example, a plurality of time-series vectors configured by learning so that distortion between the learning speech and the encoded speech is reduced, and a time-series vector corresponding to the drive code is used as a drive code vector. Output.

The weight determining means 39 determines the degree of noise based on the result of the noise evaluation input from the noise evaluation means 25.
For example, according to FIG. 10, the weight given to the time series vector from the first driving codebook 37 and the time series vector from the second driving codebook 38 are determined. Each time series vector from the first driving codebook 37 and the second driving codebook 38 is
The weights are added according to the weights given by the weight determination means 39.

The time series vector output from adaptive codebook 9 and the time series vector generated by weighting and adding by weight determining means 39 are weighted and added according to the respective gains given from gain coding means 13. The weighting and addition are performed by the means 14, and the result of the addition is supplied to the synthesis filter 7 as a drive excitation signal to generate an encoded voice.

The distance calculating means 15 finds the distance between the coded speech and the input speech S1, and searches for an adaptive code, a drive code and a gain that minimize the distance. After this coding is completed, the code of the linear prediction parameter, the adaptive code for minimizing the distortion between the input voice and the coded voice, the driving code, and the gain code are output as the coding result S2.

Next, the decoding unit 2 will be described. In the decoding unit 2, the linear prediction parameter decoding unit 16 decodes the linear prediction parameter from the code of the linear prediction parameter, sets it as a coefficient of the synthesis filter 17, and outputs the coefficient to the noise evaluation unit 26.

Next, decoding of sound source information will be described. The adaptive codebook 19 outputs a time-series vector obtained by periodically repeating a past excitation signal corresponding to the adaptive code as an adaptive code vector. The noise degree evaluation means 26 includes a decoded linear prediction parameter and an adaptive code inputted from the linear prediction parameter decoding means 16, and a previously decoded adaptive code vector and driving code inputted from the gain decoding means 23. The noise degree evaluation means 25 of the encoding unit 1 is obtained from the result of encoding or decoding the gain of the vector.
The degree of noise in the decoding section is evaluated in the same manner as described above, and the evaluation result is output to the weight determining means 42.

The first drive codebook 40 and the second drive codebook 41 output a time-series vector corresponding to the drive code as a drive code vector. It is assumed that the weight determining means 42 assigns weights in the same manner as the weight determining means 39 of the encoding unit 1 in accordance with the evaluation result of the degree of noise input from the noise degree evaluating means 26.

Each time-series vector from the first driving codebook 40 and the second driving codebook 41 is weighted according to each weight given from the weight determining means 42 and added.

The time-series vector output from the adaptive codebook 19 and the time-series vector generated by the weighted addition by the weight determining means 42 are obtained by decoding the respective gains decoded from the gain code by the gain decoding means 23. Are added by weighting by the weighting and adding means 24, and the result of the addition is supplied to the synthesis filter 17 as a driving sound source signal to generate an output sound S3.

As described above, according to the eighth embodiment, the information on the gain for the adaptive code vector and the driving code vector is also used for evaluating the degree of noise of the speech in the coding and decoding sections. This has the effect of improving the evaluation accuracy of the degree of sex. In addition, since a noise-like time series vector and a non-noise time-series vector are weighted and used so as to conform to the form of the input speech based on this evaluation, high-quality speech can be reproduced with a small amount of information. There is an effect that a voice encoding method and a voice decoding method and a device thereof are obtained.

Embodiment 9 FIG. In Embodiments 6 to 8, the gains for the adaptive code vector and the drive code vector are used to evaluate the degree of noise in speech, but in Embodiment 9, a drive sound source is used instead. Long-period prediction gain of the signal, or long-period prediction gain of coded speech obtained by linear prediction synthesis of the driving excitation signal,
Alternatively, the evaluation result of the past noise level is used, and the same effect can be obtained.

Embodiment 10 FIG. In the first to ninth embodiments, the gain for the adaptive code vector and the driving code vector, or the long-term prediction gain of the driving excitation signal, or the linear excitation synthesis of the driving excitation signal is used to evaluate the degree of noise. Although the obtained long-term prediction gain of the coded speech or the evaluation result of the degree of noise in the past is individually used, in Embodiment 10, two or more of these are used instead. .

According to the tenth embodiment, since the evaluation accuracy of the degree of noise is further improved, a speech encoding method and a speech decoding method capable of reproducing high quality speech and an apparatus thereof are obtained. .

Embodiment 11 FIG. The codebook of the gain in the gain encoding unit 13 and the gain decoding unit 23 may be changed in accordance with the evaluation result of the degree of noise in the first to tenth embodiments.

According to the eleventh embodiment, since a codebook having an optimum gain can be used according to a driving codebook, a speech encoding method, a speech decoding method, and a speech decoding method capable of reproducing high quality speech. The effect is obtained.

Embodiment 12 FIG. In the first to eleventh embodiments, the degree of noise in speech is evaluated, and the driving codebook or gain codebook is switched according to the evaluation result. However, voiced rising or bursting consonants are used. May be determined and evaluated, and the driving codebook or gain codebook may be switched according to the evaluation result.

According to the twelfth embodiment, not only the noise state of voice but also voiced rising and explosive consonants are further subdivided, and a driving codebook or a codebook of gain suitable for each is used. Therefore, there is an effect that an audio encoding method, an audio decoding method, and an apparatus thereof that can reproduce high-quality audio can be obtained.

[0150]

As described above, according to the present invention, in speech coding, as one of the parameters for evaluating the degree of noise, the coding or coding of the gain with respect to the adaptive code vector and the driving code vector is performed. Since the configuration is made to use the result, it is possible to easily obtain information on the strength of the periodicity of the input signal in the coding section and the stability thereof, and use this information for the evaluation of the degree of noise. Is improved, and a driving codebook according to the evaluation is used, so that there is an effect that encoding that enables reproduction of high-quality sound can be realized.

According to the present invention, in speech encoding,
A drive excitation signal or a coded speech obtained by linear prediction synthesis of a drive excitation signal is analyzed, and a long-period prediction gain obtained as a result of the analysis is used as one of parameters for evaluating the degree of noise. With this configuration, the strength and periodicity of the periodicity of the input signal in the coding section and its stability can be obtained more accurately, and this is used for the evaluation of the degree of noise. Is improved, and a driving codebook according to this evaluation is used.
There is an effect that encoding that enables reproduction of high-quality audio can be realized.

According to the present invention, in speech coding,
Since the past noise level evaluation results are used as one of the parameters for evaluating the noise level, local noise level evaluation errors are made based on the continuity of the noise level. Can be eliminated, the evaluation accuracy of the degree of noise is improved, and a driving codebook suitable for the mode of the input voice is used based on the evaluation, so that it is possible to realize encoding that enables high-quality voice reproduction. effective.

According to the present invention, in speech coding,
Since a plurality of driving codebooks with different degrees of noise of the stored time-series vectors are switched and used in accordance with the evaluation result of the degree of noise of speech, an intermediate level such as slightly noisy is used. Since it is possible to use a drive codebook suitable for a typical voice, there is an effect that it is possible to realize coding that enables high-quality voice reproduction.

According to the present invention, in speech coding,
Since the degree of noise of the time-series vector stored in the driving codebook is changed according to the evaluation result of the degree of noise of speech, high-quality speech can be reproduced with a small amount of information. There is an effect that the encoding that can be performed can be realized.

According to the present invention, in speech encoding,
Since the gain codebook is switched in accordance with the evaluation result of the degree of noise of the voice, there is an effect that it is possible to realize encoding that enables high-quality voice reproduction.

According to the present invention, in speech decoding,
Since one of the parameters for evaluating the degree of noise is configured to use the sign or decoding result of the gain for the adaptive code vector and the driving code vector, the periodicity of the input signal in the decoding section is easily determined. The strength and stability information can be obtained and used for the evaluation of the degree of noise, and the accuracy of the evaluation of the degree of noise is improved. Since the driving codebook is used, high-quality sound can be reproduced.

According to the present invention, in speech decoding,
As one of the parameters for evaluating the degree of noise, a driving excitation signal or a decoded speech obtained by linear prediction synthesis of the driving excitation signal is analyzed, and a long-period prediction gain obtained as a result is used. As a result, the strength of the periodicity of the input signal in the decoding section and the information on the stability thereof can be more accurately obtained. By using this information for the evaluation of the degree of noise, the evaluation accuracy of the degree of noise can be improved. Since the driving codebook adapted to the mode of the input speech to be decoded based on the evaluation is used, high-quality speech can be reproduced.

According to the present invention, in speech decoding,
Since the past noise level evaluation result is used as one of the parameters for evaluating the noise level, the local noise level at the time of decoding is determined based on the continuity of the noise level. Since the evaluation error of the degree of noise can be eliminated, the evaluation accuracy of the degree of noise can be improved, and the driving codebook adapted to the mode of the input speech to be decoded based on this evaluation is used.
This has the effect of reproducing high quality audio.

According to the present invention, in speech decoding,
Since a plurality of driving codebooks having different degrees of noise in the stored time-series vectors are switched and used in accordance with the evaluation result of the degree of noise in speech, decoding is somewhat noise-like. For example, a driving codebook suitable for the intermediate sound can be used, so that high-quality sound can be reproduced.

According to the present invention, in speech decoding,
Since the degree of noise of the time series vector stored in the driving codebook is changed in accordance with the evaluation result of the degree of noise of the voice, high-quality voice can be reproduced with a small amount of information. effective.

According to the present invention, in speech decoding,
Since the gain codebook is switched in accordance with the evaluation result of the degree of noise of the voice, high quality voice can be reproduced.

According to the present invention, there is provided an adaptive codebook which stores a past driving excitation signal and outputs a time series vector obtained by periodically repeating the past driving excitation signal corresponding to an adaptive code as an adaptive code vector. A plurality of non-noise time-series vectors are stored, a first drive codebook that outputs a time-series vector corresponding to the drive code as a drive code vector, and a plurality of noise-like time-series vectors are stored; A second drive codebook that outputs a time-series vector corresponding to the drive code as a drive code vector, and evaluates the adaptive code vector and the result of coding or encoding and decoding gains for the drive code vector for the degree of noise. Noise level evaluation means for evaluating the degree of noise using one of the elements for obtaining the value. Alternatively, information on the strength and the stability of the periodicity of the input signal in the decoding section and its stability can be obtained and used for the evaluation of the degree of noise, and the evaluation accuracy of the degree of noise can be improved. Since the driving codebook adapted to the mode of the input voice is used based on the above, there is an effect that the encoding and the decoding enabling the reproduction of the high quality voice can be realized.

According to the present invention, there is provided an adaptive codebook which stores a past excitation signal and outputs a time series vector obtained by periodically repeating the past excitation signal in accordance with an adaptive code as an adaptive code vector. A plurality of non-noise time-series vectors are stored, a first drive codebook that outputs a time-series vector corresponding to the drive code as a drive code vector, and a plurality of noise-like time-series vectors are stored; A second drive codebook that outputs a time-series vector corresponding to the drive code as a drive code vector, and analyzes the past drive excitation signal or encoded speech obtained by linear prediction synthesis of the drive excitation signal, A long-period prediction gain calculating unit for obtaining a long-period prediction gain; and an element for obtaining an evaluation value of a degree of noise by using the long-period prediction gain obtained by the long-period prediction gain calculating unit. And noise level evaluation means for evaluating the degree of noise by using the information as described above, so that information on the strength of the periodicity of the input signal of the encoding section or the decoding section and its stability can be obtained more accurately. By using this for the evaluation of the degree of noise, the accuracy of the evaluation of the degree of noise is improved, and based on this evaluation, a driving codebook suitable for the form of the input speech is used. There is an effect that encoding and decoding that enable reproduction of high sound can be realized.

According to the present invention, the noise level evaluation means for evaluating the level of the noise level using the past evaluation result of the level of the noise level as one of the elements for obtaining the evaluation value of the noise level is provided. Since it is configured so as to be provided, it is possible to eliminate a local noise degree evaluation error based on the noise degree continuity, improve the noise degree evaluation accuracy, and based on this evaluation, the input voice Since the driving codebook suitable for the mode is used, there is an effect that encoding and decoding that enable reproduction of high-quality sound can be realized.

According to the present invention, since the driving codebook switching means for switching the driving codebook according to the degree of noise evaluated by the noise degree evaluating means is provided, the code for enabling high-quality speech reproduction is provided. And decryption can be realized.

According to the present invention, there is provided a noise change imparting means for changing the noise degree of the time series vector stored in the driving codebook according to the noise degree evaluated by the noise degree evaluation means. With such a configuration, there is an effect that encoding and decoding that enable reproduction of high-quality audio with a small amount of information can be realized.

According to the present invention, the amplitude value of the low-amplitude sample of the time-series vector stored in the driving codebook is reduced to zero according to the degree of noiseness evaluated by the noise degree evaluation means. Since the apparatus is provided with the noise property change providing means for changing the degree of the nature, there is an effect that it is possible to realize encoding and decoding which enables reproduction of high quality sound with a small amount of information.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an overall configuration of a speech encoding / speech decoding apparatus that realizes a speech encoding method and a speech decoding method according to Embodiment 1 of the present invention.

FIG. 2 is a diagram showing speech encoding and coding according to Embodiment 1 of the present invention.
FIG. 4 is an explanatory diagram showing a degree of noise for each parameter obtained by a noise degree evaluation unit of the speech decoding device.

FIG. 3 is a block diagram showing an overall configuration of a speech encoding / speech decoding apparatus that realizes a speech encoding method and a speech decoding method according to Embodiment 2 of the present invention.

FIG. 4 is a diagram illustrating speech encoding / coding according to a second embodiment of the present invention.
FIG. 4 is an explanatory diagram showing a degree of noise for each parameter obtained by a noise degree evaluation unit of the speech decoding device.

FIG. 5 is a block diagram showing an overall configuration of a speech encoding / speech decoding apparatus for realizing a speech encoding method and a speech decoding method according to Embodiment 3 of the present invention.

FIG. 6 is a block diagram showing an overall configuration of a speech encoding / speech decoding apparatus which realizes a speech encoding method and a speech decoding method according to Embodiment 4 of the present invention.

FIG. 7 shows a speech encoding / coding system according to a fourth embodiment of the present invention.
FIG. 4 is an explanatory diagram showing a degree of noise for each parameter obtained by a noise degree evaluation unit of the speech decoding device.

FIG. 8 is a block diagram showing an overall configuration of a speech encoding / speech decoding apparatus that realizes a speech encoding method and a speech decoding method according to Embodiment 6 of the present invention.

FIG. 9 is a block diagram illustrating an overall configuration of a speech encoding / speech decoding apparatus that realizes a speech encoding method and a speech decoding method according to Embodiment 8 of the present invention.

FIG. 10 is an explanatory diagram of weights given to time-series vectors by weight determining means of a speech encoding / speech decoding apparatus according to an eighth embodiment of the present invention.

FIG. 11 is a block diagram illustrating an overall configuration of a speech encoding / speech decoding apparatus that realizes a conventional speech encoding method and speech decoding method.

FIG. 12 is an explanatory diagram showing the degree of noise for each parameter obtained by the noise degree evaluation means of the conventional speech encoding / decoding apparatus.

[Explanation of symbols]

1 encoding unit, 2 decoding unit, 9 adaptive codebook, 10,
20 drive codebook switching means, 11, 37, 40 first drive codebook, 12, 38, 41 second drive codebook, 1
3 gain encoding means, 23 gain decoding means, 2
5,26,28,30,31,32 Noise degree evaluation means,
27, 29 Long-period prediction gain calculating means, 33, 35 driving codebook, 34, 36 Sample thinning means (noise change giving means), 39, 42 Weight determining means (noise change giving means).

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5D045 CA04 5J064 AA01 BA13 BB03 BB13 BC11 BC26 BD03 9A001 EE04 HH16

Claims (18)

[Claims] 1. A code that evaluates the degree of noise of speech in an encoding section using parameters such as a spectrum inclination, a short-term prediction gain, and a pitch variation, and uses a codebook according to the evaluation result. A speech encoding method to which a driving linear prediction method is applied, wherein a code of an adaptive code vector and a gain code for the driving code vector or an encoding result is used as one of the parameters for evaluating a degree of noise. Encoding method. 2. A code that uses a parameter such as a spectrum inclination, a short-term prediction gain, and a pitch variation to evaluate the degree of noise of speech in the coding section, and uses a codebook according to the evaluation result. In a speech encoding method to which a driving linear prediction method is applied, a driving excitation signal or a coded speech obtained by performing linear prediction synthesis on a driving excitation signal is analyzed, and a long-period prediction gain obtained as a result of the analysis is set to 1 A speech coding method characterized by being used for evaluating the degree of noise. 3. A code using a parameter such as a spectrum inclination, a short-term prediction gain, and a pitch variation to evaluate a degree of noise of speech in a coding section and using a codebook according to the evaluation result. A speech coding method to which a driving linear prediction method is applied, wherein a past noise level evaluation result is used as one of the parameters for noise level evaluation. 4. The method according to claim 1, wherein a plurality of driving codebooks having different degrees of noise of the stored time-series vectors are switched and used according to the evaluation result of the degree of noise of speech. The speech encoding method according to claim 3. 5. The method according to claim 1, wherein the degree of noise of the time-series vector stored in the driving codebook is changed according to the evaluation result of the degree of noise of speech. A speech encoding method according to any one of the preceding claims. 6. The speech encoding method according to claim 1, wherein a codebook of gain is switched according to an evaluation result of a degree of noise of speech. 7. A code for evaluating the degree of noise of speech in a decoding section thereof using parameters such as a spectrum inclination, a short-term prediction gain, and a pitch variation, and using a codebook according to the evaluation result. A speech decoding method to which a driving linear prediction method is applied, wherein a code or decoding result of a gain for an adaptive code vector and a driving code vector is used as one of the parameters for evaluating a degree of noise. Method. 8. A code for evaluating a degree of noise of speech in a decoding section using parameters such as a spectrum inclination, a short-term prediction gain, and a pitch fluctuation, and using a codebook according to the evaluation result. In a speech decoding method to which the driving linear prediction method is applied, a driving excitation signal or a decoded speech obtained by linear prediction synthesis of the driving excitation signal is analyzed to evaluate a degree of noise, and a long-term prediction gain obtained as a result is obtained. Is used as one of the parameters. 9. A code that evaluates the degree of noise of speech in a decoding section using parameters such as a spectrum inclination, a short-term prediction gain, and a pitch variation, and uses a codebook according to the evaluation result. A speech decoding method to which the driving linear prediction method is applied, wherein a past noise degree evaluation result is used as one of the parameters for noise degree evaluation. 10. The method according to claim 7, wherein a plurality of driving codebooks having different degrees of noise of the stored time-series vectors are switched according to the evaluation result of the degree of noise of speech. The speech decoding method according to claim 9. 11. The method according to claim 7, wherein the degree of noise of the time series vector stored in the driving codebook is changed according to the evaluation result of the degree of noise of speech. The speech decoding method according to any one of the preceding claims. 12. The speech decoding method according to claim 7, wherein a gain codebook is switched according to a result of evaluating the degree of noise of speech. 13. A speech coding / speech decoding apparatus to which a code-driven linear prediction method is applied, wherein a past driving excitation signal is stored, and said past driving excitation signal is periodically repeated corresponding to an adaptive code. Codebook for outputting a time-series vector obtained as an adaptive code vector, and a first drive codebook for storing a plurality of non-noise time-series vectors and outputting a time-series vector corresponding to the drive code as a drive code vector A second driving codebook that stores a plurality of noise-like time-series vectors and outputs a time-series vector corresponding to the driving code as a driving code vector; and a gain of gain for the adaptive code vector and the driving code vector. Noise level that evaluates the degree of noise by using the code or the result of encoding or decoding as one of the elements for calculating the evaluation value of the degree of noise Speech coding and speech decoding apparatus characterized by comprising a valence means. 14. A speech coding / speech decoding apparatus to which a code-driven linear prediction method is applied, wherein a past driving excitation signal is stored, and said past driving excitation signal is periodically repeated corresponding to an adaptive code. Codebook for outputting a time-series vector obtained as an adaptive code vector, and a first drive codebook for storing a plurality of non-noise time-series vectors and outputting a time-series vector corresponding to the drive code as a drive code vector A second driving codebook that stores a plurality of noise-like time-series vectors and outputs a time-series vector corresponding to the driving code as a driving code vector; A long-period prediction gain calculating means for analyzing a coded voice obtained by the linear prediction synthesis and obtaining a long-period prediction gain; Noise encoding means for evaluating the degree of noise by using the periodic prediction gain as one of the elements for obtaining the evaluation value of the degree of noise. Device. 15. A speech coding / speech decoding apparatus to which a code-driven linear prediction method is applied, wherein a past driving excitation signal is stored, and the past driving excitation signal is periodically repeated corresponding to an adaptive code. Codebook for outputting a time-series vector obtained as an adaptive code vector, and a first drive codebook for storing a plurality of non-noise time-series vectors and outputting a time-series vector corresponding to the drive code as a drive code vector A second driving codebook in which a plurality of noise-like time-series vectors are stored and which outputs a time-series vector corresponding to the driving code as a driving code vector; Noise evaluation means for evaluating the degree of noise by using one of the elements for obtaining an evaluation value of the degree of noise, and speech coding / speech decoding characterized by comprising: Location. 16. A driving codebook switching means for switching a driving codebook according to the degree of noise evaluated by the noise degree evaluation means.
The speech encoding / speech decoding apparatus according to any one of the preceding claims. 17. The apparatus according to claim 1, further comprising: a noise level changing means for changing a level of the noise level of the time series vector stored in the driving codebook according to the level of the noise level evaluated by the noise level evaluation unit. The speech encoding / speech decoding apparatus according to any one of claims 13 to 15, characterized in that: 18. A speech coding / speech decoding apparatus to which a code-driven linear prediction method is applied, wherein a past driving excitation signal is stored, and the past driving excitation signal is periodically repeated corresponding to an adaptive code. An adaptive codebook that outputs a time-series vector obtained as an adaptive code vector, a driving codebook that stores a plurality of noise-like time-series vectors and outputs a time-series vector corresponding to the driving code as a driving code vector, Noise degree evaluation means for evaluating the degree of noise by using the code vector and the sign or encoding of the gain for the driving code vector as one of the elements for obtaining the evaluation value of the degree of noise, The amplitude value of the low-amplitude sample of the time-series vector stored in the driving codebook is set to zero according to the degree of noise evaluated by the noise degree evaluation means. Speech coding and speech decoding apparatus characterized by comprising a noise characteristic variation applying means for varying the degree of noise resistance by.