JP3475958B2 - Speech encoding / decoding apparatus including speechless encoding, decoding method, and recording medium recording program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device which reduces sound quality deterioration when a voiceless section is encoded at a very low bit rate. <P>SOLUTION: The device is equipped with: a switching means for passing a signal code sequence to a voice part decoding means or voiceless part decoding means according to whether an input signal is in a voiced or voiceless section on the basis of a VAD decision code outputted from a bit sequence decomposing means; a parameter decoding means which outputs a filter coefficient and RMS that the voiceless part decoding means derives from the signal code sequence; a smoothing means which smoothes and outputs the RMS and filter coefficient outputted from the parameter decoding means; a pulse generating means which generates a pulse train signal consisting of pulses having positions and amplitudes generated with random numbers; a pitch generating means which generates a pitch signal; a mixing means which generates an excitation signal of a synthesizing filter by performing linear sum processing between a random number signal, the pulse train signal passed from the pulse generating means and the pitch signal; and a synthesizing means which decodes and outputs the excitation signal by inputting the excitation signal to a filter composed of a smoothed filter coefficient. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for coding / decoding digital information such as a voice signal, and more particularly to a coding / decoding technique for a non-voice part.

[0002]

2. Description of the Related Art A conventional voice encoding / decoding device of this type encodes a section without voice (referred to as "non-voice section") at a bit rate very lower than that of encoding a voice section. The average bit rate for transmission is reduced by, for example, Document 1 (IEEE Communications
Magazine, pp. 64-73, Sep, 1997), etc. are referred to.

In this conventional coding apparatus, it is determined whether the input signal is a voice section or a non-voice section for each predetermined frame (10 msec), and when it is a voice section,
The input signal is coded / decoded by a normal speech coding method (ITU-T Recommendation G.729), while in the case of no speech section, the coding device intermittently codes and decodes the characteristic parameter of the input signal. Transmit to device. The decoding apparatus calculates characteristic parameters of all frames by repeating or smoothing the characteristic parameters received intermittently instead of all frames, and decodes the signals using these.

As a method for discriminating between a voice section and a non-voice section, as described in the above-mentioned reference 1, a root mean sq calculated from an input signal for each frame.
uares; "RMS"), RMS for low frequency range,
There is a method of using the number of zero crossings and a filter coefficient representing the spectral envelope characteristic. Based on the difference between these variables and the average value in each unvoiced section, threshold value determination is performed.

As a method of encoding a voice section, for example, reference 2 (ITU-T Recommendation G.729, COM15-152 July 199) is used.
5) CELP (Code Excited Linear Pr)
ediction Coding; code excitation linear prediction coding). For the CELP method, refer to Reference 3 (Code-Excited Line
ar Prediction: High Quality Speech at Very LowBit
Rates (IEEE Proc. ICASSP−85, pp.937−940, 198
5)) is also referred to.

In the encoding processing of the conventional apparatus, the input signal is subjected to linear prediction analysis for each predetermined frame to calculate a linear prediction (filter) coefficient representing the spectral envelope characteristic of the voice signal, and the spectral envelope characteristic is calculated. The corresponding LP synthesis filter is driven to calculate the excitation signal, and each excitation signal is encoded.

The excitation signal is coded for each subframe by further dividing the frame into subframes. Here, the excitation signal is composed of a periodic component representing the pitch period of the input signal, the remaining residual component, and their gains. A periodic component representing the pitch period of the input signal is represented as an adaptive code vector stored in a codebook holding a past excitation signal called an “adaptive codebook”, and the residual component is a multi-pulse composed of a plurality of pulses. It is represented as a pulse signal.

In the decoding process, the excitation signal obtained from the decoded pitch period component and the residual signal is input to the synthesis filter composed of the decoded filter coefficient to decode the voice signal.

As a method for encoding a non-voice section, as described in the above-mentioned document 1, first, in an encoding device,
The RMS and the filter coefficient representing the spectral characteristic are encoded as the characteristic parameters of the input signal.

Next, in the decoding device, the linear sum of the random number signal, the pulse signal generated in a random number, and the pitch signal is adjusted by RMS, and this is input to the synthesizing filter configured by using the filter coefficient. Decode the unvoiced signal.

The characteristic parameter is transmitted only in the frame in which the signal property changes in the non-voice section, and nothing is transmitted in the other frames. However, information on whether to transmit the characteristic parameter is transmitted separately.

In a frame in which no characteristic parameter is transmitted, past transmitted characteristic parameters are repeatedly used. However, to prevent discontinuity on the waveform,
The RMS has been smoothed.

FIG. 8 is a block diagram showing the structure of a conventional coding apparatus. Referring to FIG. 8, this encoding device includes a voice part encoding circuit 12, a voiceless part encoding circuit 14,
Signal determination circuit 16, switching circuit 18, bit generation circuit
It has 20 and.

The input terminal 10 inputs an input signal in a unit of a fixed frame, for example, a unit of 10 msec. The signal determination circuit 16 is
The input signal from the input terminal 10 is used to determine whether the frame is a voice section or a non-voice section, and the determination result (VAD determination code) is passed to the switching circuit 18 and the bit string generation circuit 20.

The voice part coding circuit 12 codes the input signal from the input terminal 10 for each frame and passes the signal code string to the switching circuit 18.

The voiceless section coding circuit 14 codes the input signal from the input terminal 10 for each frame and passes the signal code string to the switching circuit 18. Also, the judgment information (DTX judgment code) as to whether or not to transmit the signal code string in the non-voice section is passed to the bit generation circuit 20.

Based on the VAD decision code passed from the signal decision circuit 16, the switching circuit 18 judges the VAD decision code from the signal code string passed from the voice part coding circuit 12 when the input signal is in the voice section. When the code indicates that the input signal is in the non-voice section, the signal code sequence passed from the non-voice encoding circuit 14 is passed to the bit sequence generation circuit 20.

The bit string generation circuit 20 multiplexes the VAD judgment code passed from the signal judgment circuit 16, the DTX judgment code passed from the non-voice part coding circuit 10, and the signal code string passed from the switching circuit 18. , Generate bit string and output terminal
Output from 22.

FIG. 9 is a block diagram for explaining a conventional decoding device.

Referring to FIG. 9, this decoding device is provided with a bit string decomposition circuit 26, a switching circuit 28, and a voice part decoding circuit.
30 and a voiceless section decoding circuit 34. The bit string decomposition circuit 26 analyzes the bit string input from the input terminal 24.
Decompose into VAD decision code, DTX decision code and signal code string, and
The D decision code and the signal code string are passed to the switching circuit 28, and the DTX decision code is passed to the voiceless section decoding circuit 34.

The switching circuit 28, on the basis of the VAD decision code passed from the bit string decomposition circuit 26, when the input signal is in the voice section, the signal code string passed from the bit string decomposition circuit 26 to the audio part decoding circuit 30. When the input signal is in the non-voice section by the VAD judgment code, the non-voice section decoding circuit 34
Pass to.

The audio part decoding circuit 30 decodes the signal using the signal code string passed from the switching circuit 28, and outputs it to the output terminal 32.
Output from.

The non-voice part decoding circuit 34 is a bit string decomposition circuit.
The DTX judgment code passed from 26 and the signal code string passed from the switching circuit 28 are used to decode the signal of the non-voice part and output from the output terminal 32.

FIG. 10 is a block diagram showing the configuration of the voiceless decoding circuit 34 in the conventional decoding device. Referring to FIG. 10, the voiceless decoding circuit 34 includes a parameter decoding circuit 54.
A random number circuit 56, a pulse circuit 53, a mixing circuit 61, a smoothing circuit 66, and a synthesizing circuit 68.

The parameter decoding circuit 54 passes the filter coefficient and RMS obtained from the signal code string input at the input terminal 52 to the synthesizing circuit 68 and the smoothing circuit 66, respectively.

The smoothing circuit 66 smoothes the RMS passed from the parameter decoding circuit 54 and obtains the smoothed RMS obtained by the smoothing RMS.
Pass to 61. However, if the DTX decision code input from the input terminal 50 indicates that the signal code string is not transmitted, smoothing is performed using the RMS of the previous frame.

The smoothed RMS P (n) used in the nth frame counting from the beginning in each unvoiced section is calculated by the following equation (1) using the RMS p (n) input in the nth frame. To do. However, in a frame in which nothing is transmitted, the following equation (1) is calculated using the RMS transmitted immediately before instead of p (n).

[0028] P (n) = (1−α) ・ p (n−1) ＋ α ・ p (n)… (1)

Here, α is a smoothing coefficient that determines the degree of smoothing, and in the above-mentioned document 1, a fixed value of 0.125 is used. Also, P (−1) = 0.

The random number circuit 56 generates a random number, and the mixing circuit 61
Pass to. The pulse circuit 53 generates a pulse train signal composed of pulses each having a position and an amplitude generated by random numbers, and passes the pulse train signal to the mixing circuit 61.

The pitch circuit 58 generates a pitch signal composed of the above-mentioned adaptive code vector and passes it to the mixing circuit 61. Since the pitch period defining the adaptive code vector is not transmitted, a random number signal is used instead.

In the mixing circuit 61, the random number signal r (i) passed from the random number circuit 56, the pulse train signal p (i) passed from the pulse circuit 53, and the pitch signal q (i passed from the pitch circuit 58. ) And the excitation signal x (i) of the synthesizing filter is calculated and passed to the synthesizing circuit 68.

As a method of calculating the coupling coefficient of the linear sum,
For example, the method described in Document 1 is used.

First, the pitch signal coupling coefficient Gq is randomly selected from a value within a limited range.

Next, using the calculated pitch signal coupling coefficient Gq, the pulse train signal coupling coefficient Gp is calculated so that the RMS calculated from the linear sum of the pitch signal and the pulse train signal becomes the same as the smoothed RMS. To do.

A linear sum e (i) of the pitch signal and the pulse train signal is calculated by the following equation (2) using the coupling coefficient calculated above.

E (i) = Gq.q (i) + Gp.p (i) (2)

Further, the coupling coefficient Gr of the linear sum e (i) is calculated so that the new linear sum of the linear sum e (i) and the random number signal becomes the same as the smoothed RMS. Here, a fixed value γ = 0.6 is used as the coupling coefficient of the random number signal.

Therefore, the excitation signal x (i) of the synthesis filter is calculated by the following equation (3).

[0040]     x (i) = Gr ・ [Gq ・ q (i) + Gp ・ p (i)] + γ ・ r (i)… (3)

The synthesizing circuit 68 decodes the signal by inputting the excitation signal passed from the mixing circuit 61 to the filter constituted by the filter coefficient passed from the parameter decoding circuit 54, and outputs it from the output terminal 70.

[0042]

However, the above-mentioned conventional apparatus has the following problems.

The first problem is that, in the decoding apparatus, the filter coefficient used when decoding the non-voice section may change discontinuously, resulting in deterioration of the quality of the decoded signal. is there.

The reason is that the filter coefficient transmitted intermittently is used as it is.

The second problem may be affected by the immediately preceding voiced section in the first section (for example, several hundred msec) in the unvoiced section, and as a result, the amplitude of the decoded signal becomes higher than it actually is. Or, the sound quality of the decoded signal deteriorates due to the inclusion of echo.

The reason is that in the non-voice section, the RMS smoothing process is always performed so that the reproduced signal in the non-voice section does not become discontinuous.

The third problem is that the decoded signal in the non-voice section may be significantly different from the background noise of the input signal, and as a result, the background noise included in the voiced section and the auditory discontinuity. Is to occur.

The reason is that the ratio of the pulse component to the pitch component to the random number component is set to a constant value when the excitation signal of the reproduction filter is generated in the non-voice section.

Therefore, the present invention has been made in view of the above-mentioned problems, and its main purpose is to encode a non-voice section with high performance, and to introduce a non-voice section coding so as to achieve a transmission bit rate. An object of the present invention is to provide a device that realizes high coding quality even if the average value of is reduced.

Another object of the present invention is to provide a decoding apparatus that reduces the deterioration of the decoded sound quality due to the discontinuity of the filter coefficients during the non-voice section decoding.

[0051]

[Means for Solving the Problems] First to achieve the above object
The present invention, in a voice decoding device for switching the method of decoding a signal from the characteristic parameter of the decoded signal according to the discrimination information whether the decoded signal is a voice section or a non-voice section in each frame, in the characteristic parameter There is provided means for decoding using a value obtained by smoothing the characteristic parameter representing the spectral envelope characteristic of the decoded signal in the time direction.

A second aspect of the present invention is a voice decoding apparatus for switching a method of decoding a signal from a characteristic parameter of the decoded signal according to the discrimination information as to whether the decoded signal is a voice section or a non-voice section in each frame. There is provided means for decoding using a value in which a degree of smoothing in at least one of the characteristic parameters is changed in accordance with a lapse of time after the section is switched to the non-voice section.

A third aspect of the present invention is a voice decoding apparatus for switching a method of decoding a signal from a characteristic parameter of the decoded signal according to whether the decoded signal is a voice section or a non-voice section in each frame. In the section immediately after switching from the non-voice section, at least one of the transmitted characteristic parameters is directly used, and thereafter,
There is provided means for decoding by using a value smoothed in the time direction for at least one of the characteristic parameters in signal decoding.

A fourth aspect of the present invention is the speech decoding apparatus, wherein the method of decoding a signal from the characteristic parameter of the decoded signal is switched according to whether the decoded signal in each frame is a speech section or a non-speech section. There is provided means for decoding using a value in which the degree of smoothing in the time direction is changed for at least one of the characteristic parameters according to at least one of the parameters.

A fifth aspect of the present invention is the speech decoding apparatus, wherein the method for decoding a signal from the characteristic parameter of the decoded signal is switched according to whether the decoded signal in each frame is a speech section or a non-speech section. Decoding means using at least one of the parameters and a value obtained by changing the degree of smoothing in the time direction for at least one of the characteristic parameters according to the time elapsed after switching from the voice section to the non-voice section. Prepare

According to a sixth aspect of the present invention, in the speech decoding apparatus, the method for decoding the signal from the characteristic parameter of the decoded signal is switched according to the determination as to whether the decoded signal is in the voice section or the non-voice section in each frame. At least one of the transmitted characteristic parameters is directly used in the section where the parameter satisfies a predetermined condition, and thereafter, a value smoothed in the time direction for at least one of the characteristic parameters is used for signal decoding. A speech decoding apparatus comprising means for decoding the speech.

A seventh aspect of the present invention is the speech decoding apparatus, wherein the method for decoding a signal from the characteristic parameter of the decoded signal is switched according to whether the decoded signal in each frame is a speech section or a non-speech section. Decoding means using at least one of the parameters and a value in which the degree of smoothing in the time direction is changed for at least one of the characteristic parameters according to the passage of time after switching from the voice section to the non-voice section. Equipped with.

An eighth aspect of the present invention is a speech decoding apparatus for switching a method of decoding a signal from a characteristic parameter of the decoded signal in accordance with a determination as to whether the decoded signal is a speech section or a non-speech section in each frame. Immediately after switching from the non-voice section to the non-voice section and in the section where the characteristic parameter satisfies a predetermined condition, at least one of the transmitted characteristic parameters is directly used, and thereafter, at least one of the characteristic parameters is temporally And means for decoding by using the smoothed value in the signal decoding.

A ninth aspect of the present invention switches the method of decoding a signal from the characteristic parameter corresponding to the decoded signal according to the discrimination information indicating whether the decoded signal is in the voice section or in the non-voice section in each frame, and at least partly In a speech decoding device for generating a signal in a non-voice section by inputting an excitation signal composed of a plurality of types of signals to a synthesis filter in the section, in the non-voice section based on at least one of the received characteristic parameters. A means for determining a coefficient when adding a plurality of kinds of signals is provided.

A tenth aspect of the present invention switches the method of decoding a signal from the characteristic parameter corresponding to the decoded signal according to the discrimination information whether the decoded signal is a voice section or a non-voice section in each frame, In a speech decoding apparatus for generating the signal of 1) by inputting an excitation signal composed of a plurality of kinds of signals into a synthesis filter, at least one of smoothing parameters smoothed in the time direction of the received characteristic parameter in at least a part of the section. Based on the above, a coefficient for adding the plurality of types of signals in the non-voice section is determined.

In an eleventh aspect based on the first to tenth aspects, the characteristic parameter includes at least one of an amount representing a spectrum envelope corresponding to the decoded signal and an amount representing power.

A twelfth aspect of the present invention is an encoding apparatus for discriminating whether an input signal is in a voice section or a non-voice section in each frame and encoding characteristic parameters of the input signal, and first to eleventh aspects. And a voice decoding device according to any one of 1.

[0063]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described. A speech decoding apparatus according to the first embodiment of the present invention provides a method of decoding a signal from a characteristic parameter of the decoded signal according to the discrimination information in each frame whether the decoded signal is a voice section or a non-voice section. Switching means (28 in FIG. 9), means for smoothing the characteristic parameter representing the spectral envelope characteristic of the decoded signal in the characteristic parameter in the time direction (64 in FIG. 1), and the smoothed characteristic parameter Means for performing a decoding process by using (56 in FIG. 1,
53, 58, 61 and 68).

The speech decoding apparatus of the present invention, in the second embodiment, decodes a signal from the characteristic parameter of the decoded signal according to whether the decoded signal in each frame is a speech section or a non-speech section. Means for switching the method (28 in FIG. 2), and smoothing in time direction with respect to at least one of the characteristic parameters according to at least one of the characteristic parameters and time elapsed after switching from the voice section to the non-voice section. Means for converting (36 in FIG. 2 and FIG. 3)
49 and 51) and means for performing a decoding process using the smoothed characteristic parameters (56, 53, 58, 61 and 68 in FIG. 3).
It has and.

The speech decoding apparatus of the present invention, in the third embodiment, decodes a signal from a characteristic parameter of the decoded signal in accordance with the determination of whether the decoded signal is a speech section or a non-speech section in each frame. Means for switching the method (28 in FIG. 2) and directly using at least one of the transmitted characteristic parameters in the section in which the characteristic parameters satisfy a predetermined condition immediately after switching from the voice section to the non-voice section, After that, means for generating a value smoothed in the time direction with respect to at least one of the characteristic parameters (36 in FIG. 2, 49 and 51 in FIG. 3), and means for performing a decoding process using the smoothed value ( 56, 53, 5 in FIG.
8, 61 and 68).

According to the fourth embodiment of the speech decoding apparatus of the present invention, a signal is extracted from a characteristic parameter corresponding to the decoded signal according to whether the decoded signal in each frame is a speech section or a non-speech section. A means for switching the decoding method (28 in FIG. 4) and a means for generating a signal in the non-voice section by inputting an excitation signal composed of a plurality of kinds of signals to a synthesis filter (56, 53, 58, 60 in FIG. 5). , 68)
And a means (38 in FIG. 4) for determining a coefficient for adding the plurality of types of signals in the non-voice section based on at least one of the received characteristic parameters.

According to the fifth embodiment of the speech decoding apparatus of the present invention, a signal is extracted from a characteristic parameter corresponding to the decoded signal according to whether the decoded signal in each frame is a speech section or a non-speech section. A means for switching the decoding method (28 in FIG. 6) and a means for generating a signal in the non-voice section by inputting an excitation signal composed of a plurality of kinds of signals to a synthesis filter (56, 53, 58, 60 in FIG. 7). , 68)
And means for calculating a smoothing parameter smoothed in the time direction of the received characteristic parameter (64 and 50 in FIG. 7) and the plurality of types of signals in the non-voice section based on at least one of the calculated smoothing parameters. And a means (38 in FIG. 6) for deciding the coefficient when adding.

In the sixth embodiment of the speech decoding apparatus of the present invention, the characteristic parameter includes at least one of a quantity representing a spectrum envelope corresponding to the decoded signal and a quantity representing power.

In a preferred embodiment of the encoding / decoding apparatus of the present invention, it is determined in each frame whether the input signal is a voice section or a non-voice section, and the characteristic parameters of the input signal are encoded. Means (see FIG. 8) and the speech decoding apparatus according to the first to sixth embodiments.

The operation and principle of the embodiment of the present invention will be described below.

In the present invention, when decoding a non-voice section in the speech decoding apparatus, the filter coefficient intermittently transmitted is smoothed in the same manner as RMS and then used in the synthesis filter. By this means, it is possible to prevent the filter coefficients that occur due to intermittent transmission from changing discontinuously, and as a result, it is possible to improve decoded sound quality.

When the filter coefficient or RMS smoothed in the non-voice section is used in the speech decoding apparatus, the filter coefficient or RMS transmitted in the past frame by the smoothing process.
Will be affected.

Since the signal in the head section of the non-voice section includes the characteristic of the immediately preceding voice section, the smoothing process is performed in this section to use the characteristic parameter including the characteristic of the section. Will be decrypted. as a result,
The waveform amplitude of the decoded signal may become larger than it actually is, or the decoded signal may include echoes, etc., resulting in deterioration of the decoded speech.

In order to prevent this, when the fixed time after the voice section enters the non-voice section, the fixed number of frames, and the decoded characteristic parameters satisfy the predetermined conditions, for example, the RMS representing the amplitude is predetermined. The smoothing coefficient is set so that smoothing is not performed when the value is still larger than the above value. As a result, it is possible to reduce the influence from the immediately preceding voiced section, which is caused by the smoothing in the head section.

Depending on the type of background noise superimposed on the input signal, an audible difference may occur between the background noise included in the signal decoded by the speech decoding circuit and the signal decoded by the non-voice decoding circuit. . This is because the speechless decoding circuit calculates the addition ratio of the excitation signal of the synthesis filter only under the condition that the RMS becomes the same as the smoothed value of the transmitted RMS.

In the present invention, by determining the addition ratio in consideration of the characteristics of the input signal, it is possible to reduce the deterioration of the decoded sound quality due to the auditory difference. As a method of consideration, for example, when the average RMS is small, mainly random noise is used, and when the average RMS is large, or when the spectrum calculated from the filter coefficient is not flat, it is mainly a pulse signal or pitch. Use signals.

[0077]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In order to describe the embodiment of the present invention described above in more detail, an embodiment of the present invention will be described below with reference to the drawings. The encoding apparatus in the embodiment of the present invention described below has the same basic configuration as that shown in FIG. The basic configuration of the decoding device according to the embodiment of the present invention is the same as that shown in FIG.

FIG. 1 is a block diagram showing the structure of a voiceless decoding circuit in a decoding apparatus according to the first embodiment of the present invention. Referring to FIG. 1, the voiceless portion decoding circuit according to the first embodiment of the present invention is the voiceless portion decoding circuit shown in FIG.
The difference from 34 is that a smoothing circuit 64 is further provided. In the following, differences from the conventional device will be mainly described, and description of the same parts will be appropriately omitted.

The parameter decoding circuit 54 passes the filter coefficient and RMS obtained from the signal code string input from the input terminal 52 to the smoothing circuit 64 and the smoothing circuit 66, respectively.

The smoothing circuit 64 smoothes the filter coefficient passed from the parameter decoding circuit 54 and passes it to the synthesizing circuit 68. However, when the DTX determination code input from the input terminal 50 indicates that the signal code string is not transmitted, smoothing is performed using the filter coefficient of the previous frame.

Smoothing filter coefficients F (n, i), (i = 1, ..., M) used in the nth frame counting from the beginning in each unvoiced section
Is the filter coefficient f (n, i), (i
= 1, ..., M) is calculated by the following equation (4). However, in a frame in which nothing is transmitted, the following equation (4) is calculated using the filter coefficient transmitted immediately before instead of f (n, i).

[0082] F (n, i) = (1−β) F (n−1, i) + βf (n, i)… (4)

Here, β is a smoothing coefficient that determines the degree of smoothing. Also, F (−1, i) = 0, (i = 1, ..., M).

M is the order of the filter. Synthesis circuit 68
Inputs the excitation signal passed from the mixing circuit 61 into a filter constituted by the filter coefficients passed from the smoothing circuit 64, thereby decoding the signal and outputting it from the output terminal 70.

FIG. 2 is a diagram showing the configuration of a decoding device according to the second embodiment of the present invention. The second embodiment of the present invention is different from the conventional decoding device shown in FIG. 9 in that the structure of the non-voice part decoding circuit 35 is different and that the smoothing control circuit 36 is provided. is there. In the following, differences from the conventional device will be mainly described, and description of the same parts will be appropriately omitted.

The bit string decomposition circuit 26 decomposes the bit string input from the input terminal 24 into a VAD judgment code, a DTX judgment code and a signal code string, passes the VAD judgment code to the smoothing control circuit 36 and the switching circuit 28, and outputs the signal code. The column is passed to the switching circuit 28, and the DTX determination code is passed to the voiceless section decoding circuit 35.

The switching circuit 28 passes the signal code string passed from the bit string decomposing circuit 26 to the audio part decoding circuit 30 when the VAD decision code passed from the bit string decomposing circuit 26 indicates that the input signal is in the voice section. If the VAD decision code indicates that the input signal is in the non-voice section, it is passed to the non-voice section decoding circuit.

The smoothing control circuit 36 includes a bit string decomposition circuit 26.
Smoothing coefficient α according to the change of VAD judgment code passed from
(n) and β (n) are passed to the voiceless decoding circuit 35. Here, n is the frame number counted from the beginning in each non-voice section.

For example, when the VAD decision code indicates that there is a non-voice section, by setting the smoothing coefficients α (n) and β (n) to 1 at the initial specific frame number or specific time length, It is possible to remove the influence of the immediately preceding voiced portion remaining in the head portion of the voice section. Similarly, while the transmitted filter coefficient, RMS, etc. satisfy certain conditions,
By setting the smoothing coefficients α (n) and β (n) to 1, it is possible to remove the influence of the immediately preceding voiced portion remaining in the head portion in the non-voiced section. As an example of the condition, RM
As a method for detecting that S is affected by the immediately preceding voiced section, "RMS is greater than or equal to a predetermined threshold" or "RMS and RMS of the first subframe in the silent section are predetermined. Is less than or equal to the threshold. " Further, in order to detect that the filter coefficient is similar to the average spectrum of the voice section, "the distance between the filter coefficient and the standard filter coefficient that is set in advance (for example, the squared distance) is equal to or less than a predetermined threshold value", etc. There is.

Further, when the length of the immediately preceding voice section is shorter than the fixed number of frames or the fixed time length, it is considered that the characteristics of the input signal are similar to those of the non-voice section immediately before the voice section, and the filter is used. As the initial values P (-1), F (-1, i), (i = 1, ..., M) when calculating the smoothed value of the coefficient and RMS, in the last frame of the immediately preceding non-voice section. The smoothed value of can be used.

The non-voice part decoding circuit 35 includes a smoothing control circuit 36.
Decodes the signal in the non-voice section using the smoothing coefficients α (n) and β (n) passed from the DTX decision code passed from the bit string decomposition circuit 26, and the signal code string passed from the switching circuit 28. Output from the output terminal 32.

FIG. 3 is a diagram showing the configuration of the non-voice part decoding circuit 35 in the second embodiment of the present invention. Second of the present invention
The embodiment is different from the non-voice part decoding circuit in the first embodiment in the configuration of the smoothing circuit 49 and the smoothing circuit 51.

The parameter decoding circuit 54 passes the filter coefficient and RMS obtained from the signal code string input at the input terminal 52 to the smoothing circuit 49 and the smoothing circuit 51, respectively.

The smoothing circuit 49 smoothes the filter coefficient passed from the parameter decoding circuit 54 using the smoothing coefficient β (n) input from the input terminal 65, and passes it to the synthesizing circuit 68.
However, when the DTX determination code input from the input terminal 50 indicates that the signal code string is not transmitted, the filter coefficient of the previous frame is repeatedly used.

Smoothing filter coefficients F (n, i), (i = 1, ..., M) used in the nth frame counting from the beginning in each non-voice section
Is the filter coefficient f (n, i), (i
= 1, ..., M) is used to calculate by the following equation (5) similar to the above equation (4).

F (n, i) = (1−β (n)) · F (n−1, i) + β (n) · f (n, i) (5) where β (n) Is a value that changes according to the number of frames that have elapsed from the beginning in each unvoiced section, and takes a value near 1 so that the effect from past frames is forgotten when the number of elapsed frames is small. For example, β (1) = β (2) = 1.0, β (3) =
β (4) =… = β (L) = 0.7. L is the number of frames in each non-voice section.

The smoothing circuit 51 smoothes the RMS passed from the parameter decoding circuit 54 and passes it to the mixing circuit 61. However,
If the DTX judgment code input from the input terminal 50 indicates that the signal code string is not transmitted, it was transmitted immediately before.
Smooth using RMS. The smoothed RMS P (n) used in the nth frame counting from the beginning in each unvoiced section is n
Using RMS p (n) input in the frame, the above equation (1)
It is calculated by the following equation (6) similar to.

P (n) = (1−α (n)) · P (n−1) + α (n) · p (n) (6) where α (n) is β (n) Similarly, is a value that changes according to the number of frames that have elapsed from the beginning in each non-voice section,
When the number of elapsed frames is small, a value near 1 is set so as to forget the influence of past frames. For example, α (1) =
α (2) = 1.0, α (3) = α (4) = ... = α (L) = 0.7. L is the number of frames in each non-voice section.

It is also possible to perform only one of the smoothing circuit 49 and the smoothing circuit 51. In that case, the filter coefficient or RMS passed from the parameter decoding circuit 54 is directly combined with the synthesis circuit 68 or the mixing circuit 61.
Will be passed to.

The mixing circuit 61 uses the smoothed RMS passed from the smoothing circuit 51 to generate the random number signal r (i) passed from the random number circuit 56 and the pulse train signal p (i) passed from the pulse circuit 53. The excitation signal x (i) of the synthesis filter is calculated by performing the linear sum processing with the pitch signal q (i) passed from the pitch circuit 58, and passed to the synthesis circuit 68.

The synthesizing circuit 68 decodes the signal by inputting the excitation signal passed from the mixing circuit 61 to the filter constituted by the filter coefficient passed from the smoothing circuit 49,
Output from the output terminal 70.

FIG. 4 is a diagram showing the structure of a decoding device according to the third embodiment of the present invention. The decoding device of the third embodiment of the present invention is different from the conventional decoding device in the non-voice part verification circuit 38 and the non-voice part decoding circuit 37.

The bit string decomposition circuit 26 decomposes the bit string input from the input terminal 24 into a VAD judgment code, a DTX judgment code and a signal code string, passes the VAD judgment code and the signal code string to the switching circuit 28, and outputs the DTX judgment code. It is passed to the non-voice part decoding circuit 37.

The switching circuit 28 sends the signal code string passed from the bit string decomposing circuit 26 to the audio part decoding circuit 30 when the VAD decision code passed from the bit string decomposing circuit 26 determines that the input signal is a voice section. When the VAD decision code indicates that the input signal is in the non-voice section, it is passed to the non-voice section decoding circuit 37.

The voiceless section verification circuit 38 is a voiceless section decoding circuit.
Using the filter coefficient and RMS passed from 37, the setting parameter for adjusting the combination coefficient of the linear sum used in the mixing circuit 62 in FIG. 5 is determined and passed to the non-voice section decoding circuit 37. The calculation of this adjustment parameter will be described later together with the processing in the mixing circuit 62.

The non-voice part decoding circuit 37 is a bit string decomposition circuit.
The signal in the non-voice section is decoded using the DTX determination code passed from 26 and the signal code string passed from the switching circuit 28, and output from the output terminal 32.

FIG. 5 is a diagram showing the configuration of the non-voice part decoding circuit 37 in the third embodiment of the present invention. Third of the present invention
The non-voice part decoding circuit 37 in this embodiment differs from the non-voice part decoding circuit 35 in the first embodiment in the output destinations of the mixing circuit 62 and the parameter decoding circuit 54. In the following, differences from the conventional device will be mainly described, and description of the same parts will be appropriately omitted.

The parameter decoding circuit 54 obtains the filter coefficient and RMS from the signal code string input at the input terminal 52, passes the filter coefficient to the smoothing circuit 64 and the output terminal 23, and transfers the RMS to the smoothing circuit 66 and the output terminal 25. Pass to.

The smoothing circuit 66 smoothes the RMS passed from the parameter decoding circuit 54 and passes it to the mixing circuit 62. However,
If the DTX judgment code input from the input terminal 50 indicates that the signal code string is not transmitted, it was transmitted immediately before.
Smooth using RMS. In this case, the smoothing coefficient α (n) or β (n) can be set to zero so that the smoothed RMS is not updated.

The random number circuit 56 generates a random number, and the mixing circuit 62
Pass to.

The pulse circuit 53 generates a pulse train signal composed of pulses having a position and amplitude generated by random numbers, and passes the pulse train signal to the mixing circuit 62. The pitch circuit 58 generates a pitch signal composed of the above-mentioned adaptive code vector and passes it to the mixing circuit 62.

The mixing circuit 62 uses the setting parameter input from the input terminal 60 and the smoothing RMS passed from the smoothing circuit 66 to calculate the above-described linear sum coupling coefficient.

Further, by using this coupling coefficient, the random number circuit 56
The linear sum signal of the random number signal passed from the pulse circuit 53, the pulse train signal passed from the pulse circuit 53, and the pitch signal passed from the pitch circuit 53 is calculated and passed to the synthesis circuit 68.

The synthesizing circuit 68 decodes the signal by inputting the excitation signal passed from the mixing circuit 62 to the filter constituted by the filter coefficient passed from the smoothing circuit 64,
Output from the output terminal 70.

The voiceless section verification circuit 38 and the mixing circuit 62 will be described.

The nature of background noise in the non-voice part is determined in the non-voice part verification circuit 38, and the calculation method of the coupling coefficient of the pitch signal, the pulse train signal and the random number signal in the mixing circuit 62 is changed according to this property. The setting parameters to be changed include the order of determining the coupling coefficient and the coupling coefficient γ.
There is.

The information used by the silence section test circuit 38 to test the nature of the background noise in the unvoiced section is, for example, RMS and a filter coefficient.

As a method of operating the setting parameter from this information, for example, when the RMS is smaller than a predetermined threshold value and it is considered that there is no background noise, or when the spectral slope of the input signal calculated from the filter coefficient is If it is regarded as flat white noise, there is a method of increasing the contribution of the random number signal. This is equivalent to reducing γ while keeping the calculation order of the coupling coefficient.

The setting parameter of the non-voice signal can be included in the signal code string for transmission.

FIG. 6 is a block diagram showing the arrangement of the decoding apparatus according to the fourth embodiment of the present invention. The decoding device according to the fourth embodiment of the present invention is different from the decoding device according to the second embodiment in a voiceless section verification circuit 38 and a voiceless section decoding circuit 39.

The bit string decomposition circuit 26 decomposes the bit string input from the input terminal 24 into a VAD judgment code, a DTX judgment code and a signal code string, passes the VAD judgment code to the smoothing control circuit 36 and the switching circuit 28, and outputs the signal code. The column is passed to the switching circuit 28, and the DTX determination code is passed to the voiceless section decoding circuit 39.

The switching circuit 28 passes the signal code string passed from the bit string decomposition circuit 26 to the audio part decoding circuit 30 when the VAD decision code passed from the bit string decomposition circuit 26 indicates that the input signal is in the voice section. , If the input signal is in the voiceless section by the VAD determination code, it is passed to the voiceless section decoding circuit 39.
The signal code string is passed to the unvoiced part verification circuit 38 and the unvoiced part decoding circuit 39.

The smoothing control circuit 36 includes a bit string decomposition circuit 26.
The smoothing coefficients α (n) and β (n) corresponding to the change in the VAD determination code passed from are passed to the non-voice part decoding circuit 39.

The voiceless section verification circuit 38 is a voiceless section decoding circuit.
Using the smoothed RMS passed from 39, the setting parameter for adjusting the combination coefficient of the linear sum used in the mixing circuit 62 in FIG. 7 is determined and passed to the non-voice part decoding circuit 39.

For the setting parameter determination processing in the voiceless section test circuit 39, the same processing as that of the voiceless section test circuit 38 described above can be applied by replacing RMS with smoothed RMS.

The voiceless decoding circuit 39 is a bit string decomposition circuit.
26, the DTX determination code passed from 26, the signal code string passed from the switching circuit 28, the smoothing coefficients α (n) and β (n) passed from the smoothing control circuit 36, and the non-voice part test circuit 38. The signal in the non-voice section is decoded by using the setting parameter passed from and output from the output terminal 32.

Further, the smoothing RMS calculated by the smoothing circuit 50 of FIG. 7 and the smoothing filter coefficient calculated by the smoothing circuit 64 are passed to the non-voice part test circuit 38.

FIG. 7 is a diagram showing the structure of the non-voice part decoding circuit 39 in the fourth embodiment of the present invention. The voiceless portion decoding circuit 39 in the fourth embodiment of the present invention is different from the voiceless portion decoding circuit in the second embodiment in that the outputs from the averaging circuit 50 and the smoothing circuit 64. Is the output terminal
69 and the output terminal 63.

In each of the above embodiments, the pitch signal, the pulse train signal and the random number signal are all used when calculating the excitation signal of the synthesizing filter, but any of them may be omitted.

The present invention, together with the encoding device described in the section of the prior art, can be installed in a subject wireless terminal or a wireless base station to easily construct a wireless audio communication system using audio signal compression technology. You can It is also possible to build a voice terminal by storing a program for executing the decoding method already described in a recording medium such as a floppy disk and loading the program on a personal computer to which a speaker or the like is connected. You can easily.

[0131]

As described above, the present invention has the following effects.

The first effect of the present invention is to reduce the deterioration of the decoded sound quality due to the discontinuous change of the filter coefficient used when decoding the non-voice section in the decoding device.

The reason is that, in the present invention, the filter coefficient transmitted intermittently is used after being smoothed.

The second effect of the present invention is that the decoding apparatus reduces the deterioration of the decoded sound quality due to the influence of the immediately preceding voiced section at the beginning of the non-voiced section.

The reason is that, in the present invention, the smoothing coefficient is set so that the smoothing process of the characteristic parameter is not performed in the head portion of the non-voice section.

A third effect of the present invention is that the decoding apparatus reduces the auditory discontinuity caused by switching between the voice section and the non-voice section.

The reason is that, in the present invention, when the excitation signal of the reproduction filter is generated in the non-voice section, the ratio of the pulse component to the pitch component to the random number component is changed according to the property of the input signal.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a voiceless portion decoding circuit according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a decoding device according to a second exemplary embodiment of the present invention.

FIG. 3 is a diagram showing a configuration of a voiceless portion decoding circuit according to a second embodiment of the present invention.

FIG. 4 is a diagram showing a configuration of a decoding device in a third exemplary embodiment of the present invention.

FIG. 5 is a diagram showing a configuration of a voiceless portion decoding circuit according to a third embodiment of the present invention.

FIG. 6 is a diagram showing a configuration of a decoding device according to a fourth exemplary embodiment of the present invention.

FIG. 7 is a diagram showing a configuration of a voiceless section decoding circuit according to a fourth example of the present invention.

[Fig. 8] Fig. 8 is a diagram illustrating a configuration of an encoding device according to a conventional example and an example of the present invention.

FIG. 9 is a diagram showing a configuration of a conventional decoding device.

FIG. 10 is a diagram showing a configuration of a non-voice part decoding circuit in a conventional decoding device.

[Explanation of symbols]

10, 24, 50, 52, 65, 67, 60, 63 Input terminal 12 Voice part coding circuit 14 Non-voice part coding circuit 18, 28 Switching circuit 20 Bit string generation circuit 22, 23, 25, 32, 70 Output terminal 26 bit string division circuit 30 voice part decoding circuit 34, 35, 37, 39 voiceless part decoding circuit 36 smoothing control circuit 38 voiceless part verification circuit 49, 51, 64, 66 smoothing circuit 53 pulse circuit 54 parameter decoding circuit 56 Random number circuit 58 Pitch circuit 61, 62 Mixing circuit 68 Synthesizing circuit

─────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP 10-83200 (JP, A) JP 10-39898 (JP, A) JP 8-305398 (JP, A) JP 62- 253200 (JP, A) JP 60-262200 (JP, A) JP 2000-267700 (JP, A) JP 2001-249698 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 11/02 G10L 13/00 G10L 19/00-19/14 H04B 14/04 H03M 7/30 JISC file (JOIS)

Claims (43)

(57) [Claims] 1. A voice decoding device for decoding a voice signal from a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section. In at least a part of the section, a smoothed value obtained by weighted addition of a smoothed value calculated in the past from a characteristic parameter representing the spectral envelope characteristic of the decoded signal among the characteristic parameters and a characteristic parameter received in the current frame is used. A speech decoding apparatus characterized by using a means for decoding. 2. A voice decoding apparatus for decoding a voice signal from a received characteristic parameter according to whether the voice signal is a voice section or a voiceless section, wherein the decoding of the voice signal in the voiceless section is performed by the voiceless section. In at least a part of the section, weighting a and b of the smoothed value Xpre calculated in the past from the characteristic parameter representing the spectral envelope characteristic of the decoded signal among the characteristic parameters and the characteristic parameter Yn received in the current frame are used. A speech decoding apparatus characterized by using a means for decoding using a smoothed value Xn = aXpre + bYn obtained by the addition. 3. A speech decoding apparatus for decoding a signal from a received characteristic parameter according to whether the decoded signal is a speech section or a non-speech section, and a coefficient for smoothing at least one of the characteristic parameters. Is changed according to the time elapsed after switching from the voice section to the non-voice section, and at least one of the characteristic parameters is smoothed by using the changed coefficient value to obtain the voice signal of the non-voice section. A voice decoding device comprising a non-voice section decoder for decoding. 4. The non-voice section decoder uses at least one of the transmitted characteristic parameters as it is immediately after switching from the voice section to the non-voice section, and thereafter, at least one of the characteristic parameters. The speech decoding apparatus according to claim 3, wherein decoding is performed by using one of the smoothed characteristic parameters. 5. A speech decoding apparatus for decoding a signal from a received characteristic parameter according to whether the decoded signal is a speech section or a non-speech section, and a coefficient for smoothing at least one of the characteristic parameters. According to the characteristic parameter, and using the changed coefficient value, at least one of the characteristic parameters is smoothed to decode the speech signal in the non-speech section. A voice decoding device characterized by the above. 6. The non-voice section decoder uses at least one of the transmitted characteristic parameters as they are while the characteristic parameter satisfies a predetermined condition, and thereafter, at least one of the characteristic parameters is used. The speech decoding apparatus according to claim 5, wherein decoding is performed using one of the smoothed characteristic parameters. 7. A speech decoding apparatus for decoding a signal from a received characteristic parameter according to whether the decoded signal is a speech section or a non-speech section, and a coefficient for smoothing at least one of the characteristic parameters. According to the information indicating whether the characteristic parameter has been transmitted, using the changed coefficient value,
A speech decoding apparatus comprising: a speechless section decoder that smoothes at least one of the characteristic parameters to decode the speech signal in the speechless section. 8. The non-voice section decoder sets a coefficient for smoothing at least one of the characteristic parameters according to a time lapse after switching from a voice section to a non-voice section and the characteristic parameter. The speech decoding apparatus according to claim 3, wherein the speech decoding device is a speechless section decoder that changes and smoothes at least one of the characteristic parameters using the changed coefficient value, and decodes the signal in the speechless section. 9. The non-voice section decoder uses at least one of the transmitted characteristic parameters as it is, and in the following non-voice section, the time lapse after switching from the voice section to the non-voice section and the characteristics. 7. The speech decoding apparatus according to claim 4, wherein the speech decoding apparatus is a non-voice section decoder that decodes using a value obtained by smoothing at least one of the characteristic parameters according to at least one of the parameters. 10. The non-voice section decoder transmits the transmitted characteristic parameter immediately after the decoder switches from the voice section to the non-voice section and while the characteristic parameter satisfies a predetermined condition. The speech decoding apparatus according to claim 3, wherein at least one of the characteristic parameters is used as it is, and thereafter, a speech signal in a non-speech section is decoded using a value obtained by smoothing at least one of the characteristic parameters. 11. The non-voice section decoder includes a coefficient for smoothing at least one of the characteristic parameters,
Characterized in that the characteristic parameter is changed according to information indicating whether or not it is transmitted, and using the changed coefficient value, at least one of the characteristic parameters is decoded using a smoothed characteristic parameter. Claims 1 to 6, 8
10. The audio decoding device according to any one of 10 to 10. 12. The speech decoding apparatus according to claim 11, wherein the non-voice section decoder receives information indicating whether or not the characteristic parameter is transmitted on the transmitting side. 13. When the length of the voice section immediately before the unvoiced section is smaller than a predetermined value, the characteristic parameter transmitted last in the voiceless section immediately before this voice section is smoothed. The speech decoding apparatus according to claim 1, wherein the speech decoding apparatus is used as an initial value. 14. The feature parameter includes at least one of a quantity representing a spectral envelope and a quantity representing power corresponding to the decoded signal.
3. The audio decoding device according to any one of 3 above. 15. An encoding device for determining, in each frame, whether an input signal is a voice section or a non-voice section, and encoding and outputting a characteristic parameter of the input signal.
A voice encoding / decoding device comprising the voice decoding device according to claim 1. 16. A base station apparatus comprising the voice encoding / decoding apparatus or the decoding apparatus according to any one of claims 1 to 15. 17. A communication terminal device comprising the voice encoding / decoding device or the decoding device according to any one of claims 1 to 15. 18. A voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section, in at least a part of the non-voice section. A smoothing step of calculating a smoothed value obtained by weighted addition of a smoothed value calculated in the past from a characteristic parameter representing the spectral envelope characteristic of the decoded signal among the characteristic parameters, and a characteristic parameter received in the current frame; And a voiceless section voice signal decoding step of decoding the voiceless section signal using the smoothed value calculated in the smoothing step. 19. A voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section, in at least a part of the non-voice section. , A weighting a of the smoothed value Xpre calculated in the past from the feature parameter representing the spectral envelope characteristic of the decoded signal and the feature parameter Yn received in the current frame among the feature parameters.
Smoothed value Xn = aXpre + b obtained by addition using B and b
A voice decoding method comprising: a smoothing step of calculating Yn; and a voiceless section voice signal decoding step of decoding the signal of the voiceless section using the smoothed value calculated in the smoothing step. . 20. A voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section, wherein a voice section is switched to a non-voice section. Smoothing step for smoothing at least one of the characteristic parameters according to the passage of time, and a speech-free speech signal for decoding the speech-free signal using the smoothed characteristic parameter A speech decoding method comprising: a decoding step. 21. The smoothing step comprises:
21. The method according to claim 20, comprising the step (b).
The voice decoding method described in. (A) At least one of the transmitted characteristic parameters is used as it is in a certain section immediately after switching from the voice section to the non-voice section, and (b) after that, at least one of the characteristic parameters is used.
Smooth one. 22. A voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is in a voice section or a non-voice section, wherein: It is characterized by including a smoothing step of smoothing at least one of the characteristic parameters, and a non-voice section voice signal decoding step of decoding the signal of the non-voice section using the smoothed feature parameter. Speech decoding method. 23. The smoothing step comprises:
23. The method according to claim 22, comprising the step (b).
The voice decoding method described in. (A) At least one of the transmitted characteristic parameters is used as it is while the characteristic parameter satisfies a predetermined condition, and (b) At least one of the characteristic parameters is thereafter used.
Smooth one. 24. A voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section, wherein the feature parameter is transmitted or not. A smoothing step of smoothing at least one of the characteristic parameters according to information indicating whether or not; and a speech-free speech signal decoding for decoding the speech-free signal using the smoothed characteristic parameter. A voice decoding method comprising the steps of: 25. The smoothing step smoothes at least one of the characteristic parameters according to a time lapse after switching from a voice section to a non-voice section and the characteristic parameter. Item 22. The speech decoding method according to Item 20. 26. After said smoothing step uses at least one of the transmitted feature parameters as is,
24. At least one of the characteristic parameters is smoothed according to a time lapse after switching from a voice section to a non-voice section and at least one of the characteristic parameters. Speech decoding method. 27. The smoothing step comprises:
21. The method according to claim 20, comprising the step (b).
The voice decoding method described in. (A) Immediately after switching from the voice section to the non-voice section and while the characteristic parameter satisfies a predetermined condition, at least one of the transmitted characteristic parameters is directly used, and (b) after that, At least one of the characteristic parameters is smoothed in the time direction. 28. The smoothing step changes a coefficient for smoothing at least one of the characteristic parameters according to information indicating whether or not the characteristic parameters have been transmitted. Claims 18 to 23, 25
27. The speech decoding method according to any one of 27 to 27 . 29. The speech decoding method according to claim 28, wherein the characteristic parameter is the Japanese <br/> Features, further comprising receiving information indicating whether or not transmitted. 30. The feature parameter according to claim 18, wherein the feature parameter includes at least one of an amount representing a spectral envelope and an amount representing power corresponding to the decoded signal. The voice decoding method described. 31. An encoding step of determining whether an input signal is in a voice section or a non-voice section in each frame and encoding a characteristic parameter of the input signal and outputting the encoded characteristic parameter. A speech encoding / decoding method in combination with the step of carrying out the speech decoding method described in any one of the above. 32. A recording medium recording a program for executing a voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section. , Obtained by weighted addition of a smoothed value calculated in the past from the characteristic parameter representing the spectral envelope characteristic of the decoded signal and the characteristic parameter received in the current frame among the characteristic parameters in at least a part of the non-voice section. A record characterized by storing a smoothing step of calculating a smoothed value and a voiceless section voice signal decoding step of decoding the signal of the voiceless section using the smoothed value calculated in the smoothing step. Medium. 33. A recording medium recording a program for executing a voice decoding method for decoding a voice signal by changing a decoding operation of a received characteristic parameter according to whether the voice signal is a voice section or a non-voice section. In at least a part of the non-voice section, a weighting a of the smoothed value Xpre calculated in the past from the characteristic parameter representing the spectral envelope characteristic of the decoded signal and the characteristic parameter Yn received in the current frame among the characteristic parameters a.
Smoothed value Xn = aXpre + b obtained by addition using B and b
A recording medium storing a smoothing step of calculating Yn, and a voiceless section voice signal decoding step of decoding the signal of the voiceless section using the smoothed value calculated in the smoothing step. . 34. A program for executing a voice decoding method for decoding a voice signal by changing a decoding operation of a plurality of types of received characteristic parameters according to whether the voice signal is a voice section or a non-voice section. In the recorded recording medium, a smoothing step of smoothing at least one of the characteristic parameters according to the time elapsed after switching from the voice section to the non-voice section, and using the smoothed characteristic parameter And a non-voice section voice signal decoding step for decoding the signal in the non-voice section. 35. The smoothing step comprises:
35. The method according to claim 34, comprising the step (b).
The recording medium described in. (A) Immediately after switching from the voice section to the non-voice section, at least one of the transmitted characteristic parameters is used as it is. (B) After that, at least one of the characteristic parameters is smoothed. 36. A program for executing a voice decoding method for decoding a voice signal by changing the decoding operation of a plurality of types of received characteristic parameters according to whether the voice signal is a voice section or a non-voice section. In the recorded recording medium, a smoothing step of smoothing at least one of the characteristic parameters according to the characteristic parameter, and a step of decoding the signal in the non-voice section using the smoothed characteristic parameter. A recording medium storing a voice section voice signal decoding step. 37. The smoothing step comprises:
37. The method according to claim 36, comprising the step (b).
The recording medium described in. (A) At least one of the transmitted characteristic parameters is used as it is while the characteristic parameter satisfies a predetermined condition, and (b) After that, at least one of the characteristic parameters is smoothed. 38. A program for executing a voice decoding method for decoding a voice signal by changing a decoding operation of a plurality of types of received characteristic parameters according to whether the voice signal is a voice section or a non-voice section. In the recorded recording medium, a smoothing step of smoothing at least one of the characteristic parameters according to information indicating whether the characteristic parameters have been transmitted, and the smoothing step using the smoothed characteristic parameters. A non-voice section voice signal decoding step for decoding a signal in a non-voice section is stored. 39. The smoothing step smoothes at least one of the characteristic parameters according to a time lapse after switching from a voice section to a non-voice section and the characteristic parameter. Item 34. The recording medium according to Item 34. 40. After the smoothing step uses at least one of the transmitted feature parameters as is,
Claim, characterized in that said smoothing at least one of said feature parameters according to at least one of the time and the feature parameters from the switches to no-speech interval from the speech interval 35 The recording medium according to 37 . 41. The smoothing step comprises:
35. The method according to claim 34, comprising the step (b).
The recording medium described in. (A) Immediately after switching from a voice section to a non-voice section and in a section in which the characteristic parameter satisfies a predetermined condition, at least one of the transmitted characteristic parameters is directly used, and (b) after that, the characteristic is At least one of the parameters is smoothed in the time direction. 42. The smoothing step changes a coefficient for smoothing at least one of the characteristic parameters according to information indicating whether or not the characteristic parameters have been transmitted. Claims 32 to 37, 39
The recording medium according to any one of items 1 to 41 . 43. A recording medium of claim 42, wherein the characteristic parameter is the Japanese <br/> Features, further comprising receiving information indicating whether or not transmitted.