JPH08202398A - Voice coding device - Google Patents

Abstract

PURPOSE: To provide a voice coding device by which good voice quality can be obtained without degrading voice quality by temporal variation for discriminating a mode even if a frame length is shortened to 5ms-10ms or less in order to make low delay. CONSTITUTION: A hearing sense weighing signal is inputted from an input terminal 2010 with a unit of a frame, and a spectrum parameter is inputted from an input terminal 2020. A feature quantity calculating circuit A 2030 calculates, for example, pitch prediction gain PG as feature quantity and outputs it. A feature quantity calculating circuit B 3040 calculates short time prediction gain SG as feature quantity and output it. A mode discriminating circuit 2050 compares an output value PG of the input terminal 2020 and an output value SG of the input terminal 2030 with plural threshold values previously decided in accordance with mode information of a frame of one frame before stored in a delay circuit 2060, performs mode discrimination, and outputs mode information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech coding apparatus for coding a speech signal with a low delay, particularly in a short frame unit of 5 ms-10 ms or less with high quality.

[0002]

2. Description of the Related Art As a method for encoding a voice signal, for example, K. "M-LCELPS S by Ozawa et al.
Peach Coding at 4kb / s wit
h Multi-Mode and Multi-Co
"Debook" (IEICE Trans. Commu
n. vol. E77-B, No. 9, pp. 1114-
A paper (reference 1) entitled "1121, 1994) is known. In this conventional example, on the transmission side, linear prediction (LP) is performed from the audio signal for each frame (for example, 40 ms).
C) Using analysis, a spectrum parameter representing the spectral characteristic of the audio signal is extracted, and a feature amount obtained from the frame unit signal or a signal obtained by perceptually weighting the frame unit signal is calculated. A mode discrimination (for example, a vowel part and a consonant part) is performed by using it, and coding is performed by switching the algorithm or codebook according to the mode discrimination result. The encoding unit further divides the frame into subframes (for example, 8 ms), extracts parameters (delay parameters and gain parameters corresponding to the pitch period) in the adaptive codebook based on past excitation signals for each subframe, and adapts them. Optimum sound source from a sound source codebook (vector quantization codebook) composed of a noise signal of a predetermined type with respect to the residual signal obtained by pitch-estimating the voice signal of the subframe by the codebook and pitch-predicting The source signal is quantized by selecting a code vector and calculating the optimum gain. The excitation code vector is selected so as to minimize the error power between the signal synthesized from the selected noise signal and the residual signal. Then, the index and the gain indicating the type of the selected code vector, the spectrum parameter and the parameter of the adaptive codebook are combined by the multiplexer unit and transmitted. A description of the receiving side is omitted.

[0003]

In the conventional method described above, when the frame length is reduced to, for example, 5 ms or less in order to reduce the processing delay, if mode information or pitch extraction and level extraction are obtained in frame units, these Mode change due to instability due to large time fluctuation of value,
Alternatively, there is a problem that unstable and incorrect pitch extraction and unstable and incorrect level extraction occur, resulting in deterioration of sound quality. SUMMARY OF THE INVENTION It is an object of the present invention to solve the above problems, provide correct mode determination, correct pitch extraction, and correct level extraction, and suppress sound quality deterioration due to these errors.

[0004]

According to the present invention, a frame dividing section for dividing an audio signal into predetermined frame units, a mode determining section for calculating a feature amount from the audio signal and performing a mode determination, and the determination In a voice encoding device that encodes the voice signal based on the result, at least one type of feature amount obtained from each of the present frame and at least one past frame and mode discrimination information obtained from at least one past frame A speech coding apparatus having a function of discriminating the mode of the current frame is obtained by using.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and performs mode determination, and the above-mentioned determination result is used. In a voice encoding device for encoding a voice signal, at least one type of feature amount obtained from each of a current frame and at least one past frame and mode discrimination information obtained from at least one past frame In a speech coding apparatus having a function of discriminating a mode of a frame, a speech coding apparatus can be obtained in which, as the feature quantity, at least one kind of time-dependent change ratio of the feature quantity is included as the feature quantity.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and performs mode determination, and the above-mentioned determination result is used. In a voice encoding device for encoding a voice signal, at least one type of feature amount obtained from each of a current frame and at least one past frame and mode discrimination information obtained from at least one past frame A voice encoding device having a function of discriminating a mode of a frame, wherein the feature amount is a feature amount for each feature amount of two frames of a current frame or at least one past frame. As a result, it is possible to obtain a speech coding apparatus in which the ratio of the two feature quantities is included as the feature quantity.

Further, according to the present invention, a frame dividing section that divides the audio signal into predetermined frame units, a mode determining section that calculates a feature amount from the audio signal and determines a mode, and the above-mentioned determination result is used. In a voice encoding device for encoding a voice signal, at least one type of feature amount obtained from each of a current frame and at least one past frame and mode discrimination information obtained from at least one past frame A speech coding apparatus having a function of discriminating a mode of a frame, characterized in that the feature quantity includes at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch. A speech coder is obtained.

Further, according to the present invention, a frame dividing section that divides the audio signal into predetermined frame units, a mode determining section that calculates a feature amount from the audio signal and performs mode determination, and the above-mentioned determination result is used. In a voice encoding device for encoding a voice signal, at least one type of feature amount obtained from each of a current frame and at least one past frame and mode discrimination information obtained from at least one past frame A voice encoding device having a function of discriminating a mode of a frame, wherein the feature amount includes at least one time change ratio of feature amounts as a feature amount.
A speech coding apparatus is obtained which includes at least one of pitch prediction gain, short-term prediction gain, level, and pitch as the feature amount as the feature amount.

Further, according to the present invention, a frame dividing section that divides the audio signal into predetermined frame units, a mode determining section that calculates a feature amount from the audio signal and performs mode determination, and the above-mentioned determination result is used. In a voice encoding device for encoding a voice signal, at least one type of feature amount obtained from each of a current frame and at least one past frame and mode discrimination information obtained from at least one past frame A voice encoding device having a function of discriminating a mode of a frame, wherein the feature amount is a feature amount for each feature amount of two frames of a current frame or at least one past frame. , A speech coding apparatus including a ratio of the two feature quantities as a feature quantity, wherein the feature quantity is a pitch prediction gain, Period predicted gain, a level, the speech coding apparatus is obtained which is characterized in that included as a feature quantity of at least one or more of the pitch.

Further, according to the present invention, a frame dividing section that divides an audio signal into predetermined frame units, a mode determining section that calculates a feature amount from the audio signal and determines a mode, and a pitch is extracted from the audio signal. In the pitch extraction unit and in the speech coding apparatus for coding the speech signal based on the discrimination result, at least one type of feature amount and at least one past characteristic value respectively obtained from the current frame and at least one past frame. A speech encoding apparatus is obtained which has a pitch extraction unit that corrects the pitch of the current frame using the mode discrimination information obtained from one frame.

Further, according to the present invention, a frame dividing unit that divides an audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a pitch is extracted from the audio signal. In the pitch extraction unit and in the speech coding apparatus for coding the speech signal based on the discrimination result, at least one type of feature amount and at least one past characteristic value respectively obtained from the current frame and at least one past frame. A speech coding apparatus characterized by having a pitch extraction unit that corrects the pitch of the current frame by using the mode discrimination information obtained from one frame, wherein at least one or more kinds of the characteristic quantities change with time as the characteristic quantity. A speech coding apparatus including a ratio as a feature amount can be obtained.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a pitch is extracted from the audio signal. In the pitch extraction unit and in the speech coding apparatus for coding the speech signal based on the discrimination result, at least one type of feature amount and at least one past characteristic value respectively obtained from the current frame and at least one past frame. A speech coding apparatus characterized by having a pitch extraction unit for correcting the pitch of the current frame by using the mode discrimination information obtained from one frame, wherein the feature amount is at least one of the current frame and the past. The ratio of the two feature amounts is included as the feature amount for each feature amount of any two frames. Voice encoding device can be obtained.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a pitch is extracted from the audio signal. In the pitch extraction unit and in the speech coding apparatus for coding the speech signal based on the discrimination result, at least one type of feature amount and at least one past characteristic value respectively obtained from the current frame and at least one past frame. Using the mode discrimination information obtained from one frame, a speech coding apparatus characterized by having a pitch extraction unit for correcting the pitch of the current frame, as the feature amount, pitch prediction gain, short-term prediction gain, level, There is provided a speech coding apparatus characterized by including at least one kind of pitch as a feature amount.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a pitch is extracted from the audio signal. In the pitch extraction unit and in the speech coding apparatus for coding the speech signal based on the discrimination result, at least one type of feature amount and at least one past characteristic value respectively obtained from the current frame and at least one past frame. A speech coding apparatus characterized by having a pitch extraction unit that corrects the pitch of the current frame by using the mode discrimination information obtained from one frame, wherein at least one or more kinds of the characteristic quantities change with time as the characteristic quantity. A speech coding apparatus including a ratio as a feature amount, wherein the feature amount includes a pitch prediction gain, a short-term prediction gain, a level, Speech coding device is obtained, characterized in that included as the feature quantity of at least one or more pitch.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a pitch is extracted from the audio signal. In the pitch extraction unit and in the speech coding apparatus for coding the speech signal based on the discrimination result, at least one type of feature amount and at least one past characteristic value respectively obtained from the current frame and at least one past frame. A speech coding apparatus characterized by having a pitch extraction unit for correcting the pitch of the current frame by using the mode discrimination information obtained from one frame, wherein the feature amount is at least one of the current frame and the past. The ratio of the two feature amounts is included as the feature amount for each feature amount of any two frames. In voice encoding device, as the feature amount, a pitch prediction gain, short-term prediction gain, level, the speech coding apparatus is obtained which is characterized in that included as a feature quantity of at least one or more of the pitch.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a level is extracted from the audio signal. And a level extraction unit for encoding the audio signal based on the determination result, and at least one type of feature amount and at least one of the past obtained respectively from the current frame and at least one frame in the past. A speech coding apparatus is obtained which has a level extraction unit for correcting the level of the current frame using the mode discrimination information obtained from one frame.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a level is extracted from the audio signal. And a level extraction unit for encoding the audio signal based on the determination result, and at least one type of feature amount and at least one of the past obtained respectively from the current frame and at least one frame in the past. A speech coding apparatus characterized by having a level extraction unit that corrects the level of the current frame by using the mode discrimination information obtained from one frame, wherein at least one or more of the characteristic amounts as the characteristic amount changes with time. A speech coding apparatus including a ratio as a feature amount can be obtained.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a level is extracted from the audio signal. And a level extraction unit for encoding the audio signal based on the determination result, and at least one type of feature amount and at least one of the past obtained respectively from the current frame and at least one frame in the past. A speech coding apparatus characterized by having a level extraction unit for correcting the level of the current frame by using the mode discrimination information obtained from one frame, wherein the feature amount is at least one of the current frame and the past. The ratio of the two feature amounts is included as the feature amount for each feature amount of any two frames. Voice encoding device can be obtained.

Further, according to the present invention, a frame dividing section that divides the audio signal into predetermined frame units, a mode determining section that calculates a feature amount from the audio signal and determines a mode, and a level is extracted from the audio signal. And a level extraction unit for encoding the audio signal based on the determination result, and at least one type of feature amount and at least one of the past obtained respectively from the current frame and at least one frame in the past. Using the mode discrimination information obtained from one frame, a speech encoding device characterized by having a level extraction unit for correcting the level of the current frame, as the feature amount, pitch prediction gain, short-term prediction gain, level, There is provided a speech coding apparatus characterized by including at least one kind of pitch as a feature amount.

Further, according to the present invention, a frame dividing unit that divides the audio signal into predetermined frame units, a mode determining unit that calculates a feature amount from the audio signal and determines a mode, and a level is extracted from the audio signal. And a level extraction unit for encoding the audio signal based on the determination result, and at least one type of feature amount and at least one of the past obtained respectively from the current frame and at least one frame in the past. A speech coding apparatus characterized by having a level extraction unit that corrects the level of the current frame by using the mode discrimination information obtained from one frame, wherein at least one or more of the characteristic amounts as the characteristic amount changes with time. A speech coding apparatus including a ratio as a feature amount, wherein the feature amount includes a pitch prediction gain, a short-term prediction gain, a level, Speech coding device is obtained, characterized in that included as the feature quantity of at least one or more pitch.

Further, according to the present invention, a frame dividing section that divides the audio signal into predetermined frame units, a mode determining section that calculates a feature amount from the audio signal and determines a mode, and a level is extracted from the audio signal. And a level extraction unit for encoding the audio signal based on the determination result, and at least one type of feature amount and at least one of the past obtained respectively from the current frame and at least one frame in the past. A speech coding apparatus characterized by having a level extraction unit for correcting the level of the current frame by using the mode discrimination information obtained from one frame, wherein the feature amount is at least one of the current frame and the past. The ratio of the two feature amounts is included as the feature amount for each feature amount of any two frames. In voice encoding device, as the feature amount, a pitch prediction gain, short-term prediction gain, level, the speech coding apparatus is obtained which is characterized in that included as a feature quantity of at least one or more of the pitch.

[0022]

With the above structure, the correct mode information, the correct pitch extraction, and the correct level extraction are shown before and after the current frame. However, when these values are exceeded only in the current frame, the correct mode information in the past or the correct pitch is extracted. By applying the information from the frame indicating the extraction and the correct level extraction to the current frame, the mode information of the current frame or the pitch extraction and the level extraction can be corrected using the information over a long time length. Therefore, correct mode information or correct pitch and correct level extraction can be provided to the audio encoding process, and the sound quality deterioration due to these errors can be suppressed.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the speech encoding apparatus according to claim 4, "at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch is included as the feature amount as the feature amount". A related example is shown in FIG.

In the figure, a voice signal is input from an input terminal 100, a frame division circuit 110 divides the voice signal into frames (for example, 5 ms), and a sub-frame division circuit 120 divides the voice signal of the frame into shorter than the frame. It is divided into subframes (for example, 2.5 ms).

In the spectrum parameter calculation circuit 200, a speech signal is cut out by applying a window (for example, 24 ms) longer than the subframe length to a speech signal of at least one subframe, and spectrum parameters are set to predetermined orders (for example, P = 10th order) Here, well-known LPC analysis, Burg analysis, or the like can be used for the calculation of the spectrum parameter. Here, Berg analysis is used. For more information on Burg analysis, see 82-of the book (Corona Publishing Co., Ltd., 1988) entitled "Signal Analysis and System Identification" by Nakamizo.
The description is omitted because it is described on page 87 (reference 2). Furthermore, in the spectrum parameter calculation unit, Burg
Linear prediction coefficient α _i (i = 1, ..., 1) calculated by the method
0) is converted into an LSP parameter suitable for quantization and interpolation. Here, the conversion from the linear prediction coefficient to the LSP is performed by Sugamura et al., "Speech information compression by line spectrum pair (LSP) speech analysis and synthesis method" (The Institute of Electronics and Communication Engineers, J64-A, pp. 599). -606, 1981)
(Reference 3) can be referred to. That is, the linear prediction coefficient obtained by the Burg method in the second subframe is
It is converted into SP parameters, the LSP of the first sub-frame is obtained by linear interpolation, the LSP of the first sub-frame is inversely transformed back to the linear prediction coefficient, and the linear prediction coefficient α _il (i = 1, ..., 10, l = 1, ...,
5) is output to the perceptual weighting circuit 230. In addition, the LSP of the first and second subframes is output to the spectrum parameter quantization circuit 210.

The spectrum parameter quantization circuit 210 efficiently digitizes the LSP parameters of a predetermined subframe. In the following, it is assumed that vector quantization is used as the quantization method and the LSP parameter of the second subframe is quantized. LSP
A well-known method can be used as a method of vector quantization of parameters. A specific method is, for example, Japanese Patent Laid-Open No. 4-1
71500 (Japanese Patent Application No. 2-297600) (Reference 4) and Japanese Patent Laid-Open No. 4-363000 (Japanese Patent Application No. 3-26).
1925) (Reference 5), Japanese Patent Laid-Open No. 5-6199 (Japanese Patent Application No. 3-155049) (Reference 6), T.I. No
mura et al. , By "LSPCoding
Using VQ-SVQ With Interp
automation in 4.075 kbps ML
The paper entitled "CELP Speech Coder" (Proc. Mobile Multimedia C
communications, pp. B. 2.5, 19
93) (Reference 7) and the like, so description thereof will be omitted here. Further, the spectrum parameter quantization circuit 210 restores the LSP parameters of the first and second subframes based on the LSP parameters quantized in the second subframe. Here, the quantized LSP parameter of the second subframe of the current frame and the quantized LSP of the second subframe of the frame one past are linearly interpolated to restore the LSP of the first and second subframes. Where LS before quantization
After selecting one type of code vector that minimizes the error power between P and the quantized LSP, first to
The LSP of the fourth subframe can be restored. In order to further improve the performance, after selecting a plurality of code vectors that minimize the error power, the cumulative distortion is evaluated for each candidate, and the candidate that minimizes the cumulative distortion and the interpolation L are calculated.
A set of SPs can be selected.

The LSP of the first and second subframes and the quantized LSP of the second subframe restored as described above are linearly predicted for each subframe by a linear prediction coefficient _α'il (i = 1, ..., 10, l =
1, ..., 5), and the impulse response calculation circuit 310
Output to. In addition, the multiplexer 40 transmits the impulse representing the code vector of the quantized LSP of the second subframe.
Output to 0.

In the above, instead of linear interpolation, LS
P interpolation patterns are prepared for a predetermined number of bits (for example, 2 bits), and LSPs of 1 and 2 subframes are restored for each of these patterns and a code vector for minimizing the cumulative distortion. A set of interpolation patterns may be selected. In this way, the transmission information increases by the number of bits of the interpolation pattern, but it is possible to more accurately represent the temporal change in the LSP frame. Here, the interpolation pattern is the LSP for training.
It may be created by learning in advance using data,
A predetermined pattern may be stored. As the predetermined pattern, for example,
T. Taniguchi et al. By “Imp
lovedCELP speech coding a
t 4 kb / s and below ”(Proc. ICSLP, pp. 41-44, 199).
2) The pattern described in (Reference 8) or the like can be used. Further, in order to further improve the performance, after selecting an interpolation pattern, an error signal between the true value of the LSP and the interpolation value of the LSP is obtained in a predetermined subframe, and the error signal is further converted into an error code. It may be represented in a book.

The perceptual weighting circuit 230 receives from the spectral parameter calculation circuit 200 a linear prediction coefficient α _il (i = 1, ..., 10, l = 1, 1) before quantization for each subframe.
, 5) is input, and the perceptual weighting is performed on the audio signal of the sub-frame based on the reference 1 and the perceptual weighting signal is output.

The proposed mode discrimination circuit 2000 receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, receives the spectrum parameter from the spectrum parameter calculation circuit 200, and outputs the mode discrimination information. The configuration of the proposed mode discrimination circuit is shown in FIG.

In FIG. 2, the perceptual weighting signal is input from the input terminal 2010 on a frame-by-frame basis, and the input terminal 20
Input the spectral parameters from 20. The characteristic amount calculation circuit A2030 calculates and outputs, for example, a pitch prediction gain PG as a characteristic amount. Feature amount calculation circuit B3040
Then, for example, a short-term prediction gain SG is calculated and output as the feature amount.

In the mode discrimination circuit 2050, the delay unit 20
According to the mode information of the previous frame stored in 60, the output value PG of 2030 and the output value SG of 2040 are compared with a plurality of predetermined thresholds to determine the mode. And output the mode information.
The mode discrimination circuit 2050 outputs the mode discrimination result to the adaptive codebook circuit 500 and the excitation quantization circuit 350.

Returning to FIG. 1, the response signal calculation circuit 240
Is the linear prediction coefficient α _il input from the spectrum parameter calculation circuit 200 for each subframe, and the linear prediction coefficient α ′ _il restored by quantization and interpolation from the spectrum parameter quantization circuit 210 for each subframe. The input signal d is input using the value of the filter memory that is input and stored.
The response signal with (n) = 0 is calculated for one subframe,
Output to the subtractor 235. Here, the response signal x _z (n)
Is represented by the following formula (1).

[0034]

[Equation 1]

Here, γ is a weighting coefficient for controlling the perceptual weighting amount, and has the same value as the following equation (3).

The subtractor 235 subtracts the response signal for one subframe from the perceptual weighting signal by the following equation (2),
It outputs x ′ _w (n) to the adaptive codebook circuit 300.

X ′ _w (n) = x _w (n) −x _z (n) (2) In the impulse response calculation circuit 310, z conversion is represented by the following formula (3).
Impulse response h of the weighting filter represented by
_w (n) is calculated by a predetermined number L of points and output to the adaptive codebook circuit 300 and the excitation quantization circuit 350.

[0038]

[Equation 2]

Adaptive codebook circuit 500 determines pitch parameters. For details, refer to Reference 2 above. Also, pitch prediction is performed by the adaptive codebook according to the following equation (4), and the adaptive codebook prediction difference signal z (n) is output.

X (n) = x ′ _w (n) −b (n) (4) Here, b (n) is an adaptive codebook pitch prediction signal and can be expressed by the following equation (5).

B (n) = βv (n−T) * h _w (n) (5) where β and T represent the gain and delay of the adaptive codebook, respectively. v (n) is an adaptive code vector. The symbol * indicates a convolution operation.

The nonuniform pulse number type sparse excitation codebook 351 is a sparse codebook in which the number of non-zero components of each vector is different.

In the excitation quantization circuit 350, the best excitation code vector c _j (n) is minimized so that the expression (6) is minimized for all or some of the excitation code vectors stored in the excitation codebook 351. ) Is selected. At this time, one of the best code vectors may be selected, or 2
It is also possible to select more than one type of code vector and make a final selection for one type during gain quantization. Here, it is assumed that two or more types of code vectors are selected.

D _j = Σ _n (z (n) −γ _j c _j (n) h _w (n)) ² (6) Note that Expression (6) is applied only to some sound source code vectors. In this case, it is also possible to preselect a plurality of sound source code vectors in advance and apply equation (6) to the preselected sound source code vectors.

The gain quantization circuit 365 reads the gain code vector from the gain code book 355, and for the selected excitation code vector, the excitation code vector and the gain code vector are set so as to minimize the equation (7). Select the combination of.

D _{j, k} = Σ _n (x _w (n) −β ′ _k v (n−T) h _w (n) −γ ′ _k c _j (n) h _w (n)) ² (7) Here, β ′ _k and γ ′ _k are gain codebook 355.
It is the k-th code vector in the two-dimensional gain codebook stored in. The indexes representing the selected sound source code vector and gain code vector are output to the multiplexer 400.

The weighting signal calculation circuit 360 inputs the output parameter of the spectrum parameter calculation circuit and each index, reads the code vector corresponding to the index from the index, and first, the drive source signal v (based on the following equation (8). n) is calculated.

V (n) = β ′ _k v (n−T) + γ ′ _k c _j (n) (8) Next, the output parameter of the spectrum parameter calculation circuit 200 and the output parameter of the spectrum parameter quantization circuit 210 are set. Using the following equation (9), the weighting signal s
_w (n) is calculated for each subframe and output to the response signal calculation circuit 240.

[0049]

(Equation 3)

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch are included as the feature quantity in the speech coding apparatus according to claim 4. The description of the embodiment relating to "is completed.

The speech coding apparatus according to claim 2, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as a feature amount as the feature amount. An example is shown in FIG.

According to the present invention, at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as the feature quantity in the claim 4 as the feature quantity. 1 is an embodiment relating to the "encoding device", and the proposed mode discrimination circuit 2000 of FIG.
Since the configuration of the proposed mode discrimination circuit of this embodiment is different from that of the present embodiment, the configuration of the proposed mode discrimination circuit of this embodiment will be described with reference to FIG.

Like the proposed mode discrimination circuit 2000, the proposed mode discrimination circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis and the spectrum parameter from the spectrum parameter calculation circuit 200, and outputs the mode discrimination information. The configuration of the proposed mode discrimination circuit is shown in FIG.

In FIG. 3, the perceptual weighting signal is input from the input terminal 3010 in units of frames, and the input terminal 30
Input the spectral parameters from 20.

The characteristic amount calculation circuit A3030 calculates and outputs, for example, a pitch prediction gain PG as a characteristic amount. In the characteristic amount calculation circuit B3040, as the characteristic amount, for example, RM
Calculate and output the S ratio RR. Feature amount calculation circuit C3050
Then, for example, the short-term prediction gain SG and the short-term prediction gain ratio SGR are calculated and output as the feature amount.

In the mode discrimination circuit 3060, the delay unit 30
According to the mode information of the previous frame stored in 70, the output value PG of 3030, the output value RR of 3040, and the output values SG and SGR of 3050 are stored in a predetermined number. The mode information is compared and the mode information is output. Mode discrimination circuit 3060
Outputs the mode discrimination result to the adaptive codebook circuit 500 and the excitation quantization circuit 350, similarly to the proposed mode discrimination circuit 2000.

The structure of the characteristic amount calculation circuit B3040 is shown in FIG. In FIG. 4, the perceptual weighting signal is input from the input terminal 4010 in frame units, and the RMS calculation circuit 40
20 calculates the RMS value R, and uses this value and the past RMS value stored in the delay unit 4030 to calculate the RMS calculation circuit 4
040 calculates the RMS ratio RR and outputs it to the output terminal 405.
Output by 0. Here, the RMS ratio RR is the rate of change of the RMS when the time axis is taken in frame units.

The configuration of the characteristic amount calculation circuit C3050 is shown in FIG. In FIG. 5, the perceptual weighting signal is input from the input terminal 5010 in frame units, the spectrum parameter is input from the input terminal 5020 in frame units, the short-term prediction gain calculation circuit 5030 calculates the short-term prediction gain SG, and this value is calculated. Is output from the output terminal 5070. Also, the short-term prediction gain ratio calculation circuit 5050 calculates the short-term prediction gain ratio using the short-term prediction gain SG calculated in 5030 and the short-term prediction gain of the past frame stored in the delay unit 5040, and this is output terminal. Output by 5060.

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch are included as the feature amount in the claim 4 as the feature amount. The description of the embodiment relating to "is completed.

The present invention relates to the speech coder according to claim 3, wherein at least one of a pitch prediction gain, a short-term prediction gain, a level, and a pitch is included as the feature quantity as the feature quantity. An example is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch are included as the feature amount in the claim 4 "feature amount. The configuration of the feature amount calculation circuit C3050 of FIG. 3 which is an embodiment relating to the "encoding device" and the feature amount calculation circuit C of this embodiment are different. Therefore, the configuration of the feature amount calculation circuit C of this embodiment is shown in FIG. Explain.

In FIG. 9, the perceptual weighting signal is input from the input terminal 8010 in frame units, and the input terminal 80
The spectrum parameter is input from 20 in frame units, the short-term prediction gain calculation circuit 8030 calculates the short-term prediction gain SG, and this value is output from the output terminal 8070. Further, the short-term prediction gain ratio is calculated by the short-term prediction gain ratio calculation circuit using the short-term prediction gain SG calculated in 8030 and the short-term prediction gain of the previous frame stored in the delay unit 8050. Is output from the output terminal 8060.

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch are included as the feature amount as the feature amount in the speech encoding device according to claim 4. The description of the embodiment relating to "is completed.

An embodiment relating to claim 3 is shown in FIG.

In the present invention, the speech according to claim 4 is characterized in that at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as the feature amount as the feature amount. Since the configuration of the proposed mode discriminating circuit which is the embodiment relating to the "encoding device" is different from that of the proposed mode discriminating circuit of the present embodiment, the configuration of the proposed mode discriminating circuit of the present embodiment will be described with reference to FIG. .

The proposed mode discrimination circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis and the spectrum parameter from the spectrum parameter calculation circuit 200, and outputs the mode discrimination information. The configuration of the proposed mode discrimination circuit is shown in FIG.

In FIG. 10, the perceptual weighting signal is input from the input terminal 9010 in frame units, and the input terminal 9
Input the spectrum parameter from 020.

The characteristic amount calculation circuit A 9030 calculates and outputs, for example, a pitch prediction gain PG as a characteristic amount. In the feature amount calculation circuit B9040, as the feature amount, for example, RM
The S value R and the RMS ratio RR are calculated and output. The characteristic amount calculation circuit C9050 calculates and outputs, for example, the short-term prediction gain SG and the short-term prediction gain ratio SGR as the characteristic amount.

In the mode discrimination circuit 9060, the delay device 90
A plurality of predetermined output values PG of 9030, output values R and RR of 9040, and output values SG and SGR of 9050 are stored in accordance with the mode information of the previous frame stored in 70. The mode information is compared with the threshold value and the mode information is output. Mode discrimination circuit 90
Reference numeral 60 denotes the adaptive codebook circuit 50 which indicates the mode discrimination result.
0, output to the excitation quantization circuit 350.

FIG. 11 shows the configuration of the characteristic amount calculation circuit B9040.
Shown in In FIG. 11, the perceptual weighting signal is input from the input terminal 1010 in frame units, the RMS calculation circuit 11020 calculates the RMS value R, and outputs the RMS value R from the output terminal 11060. Also, the RMS ratio calculation circuit 11040 is calculated using the output value R of the RMS calculation circuit 11020 and the RMS value of the frame two frames before in the past stored in the delay unit 21030.
Then, the RMS ratio RR is calculated, and this is output from the output terminal 11050.

The feature quantity calculating circuit C9050 according to claim 4, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as the feature quantity as the feature quantity. Feature amount calculation circuit C3050 of FIG.
Is the same as

This is the end of the description of the embodiment according to claim 3.

An embodiment relating to claim 2 is shown in FIG.

In the present invention, the characteristic amount calculating circuit B of FIG. 10 which is an embodiment relating to claim 3 and the characteristic amount calculating circuit B of the present embodiment have different configurations. explain.

The proposed mode discriminating circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis and the spectrum parameter from the spectrum parameter calculating circuit 200, and outputs the mode discriminating information. FIG. 27 shows the configuration of the proposed mode discrimination circuit.

In FIG. 27, a perceptual weighting signal is input from the input terminal 10010 in frame units, the RMS calculation circuit 10020 calculates the RMS value R, and the output terminal 100
Output from 60. Further, the RMS ratio calculation circuit 10040 calculates the RMS ratio RR using the output value R of the RMS calculation circuit 10020 and the RMS value of the past frame stored in the delay device 10030, and outputs the RMS ratio RR from the output terminal 10050. Here, the RMS ratio RR is the rate of change of the RMS when the time axis is taken in frame units.

This is the end of the description of the embodiment according to claim 2.

An embodiment relating to claim 1 is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch are included as the feature quantity in the claim 4 as the feature quantity. The configuration of the proposed mode discrimination circuit of FIG. 3 which is an embodiment related to the "encoding device" and the proposed mode discrimination circuit of this embodiment are different, and therefore the configuration of the proposed mode discrimination circuit will be described with reference to FIG.

The proposed mode discriminating circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis and the spectrum parameter from the spectrum parameter calculating circuit 200, and outputs the mode discriminating information.

The characteristic amount calculation circuit A12030 calculates and outputs, for example, a pitch prediction gain PG as a characteristic amount.
The characteristic amount calculation circuit B12040 calculates and outputs, for example, an RMS value R as a characteristic amount. Feature amount calculation circuit C120
In 50, for example, a short-term prediction gain SG is calculated and output as a feature amount.

In the mode discrimination circuit 12060, the delay unit 1
According to the mode information of the previous frame stored in 2070, the output value PG of 12030 and 12040
Output value R of 12050 and output value SG of 12050 are compared with a plurality of predetermined thresholds, mode determination is performed, and mode information is output. Mode discrimination circuit 1206
0 indicates that the result of mode discrimination is the adaptive codebook circuit 500,
Output to the excitation quantization circuit 350.

With the above, the description of the embodiment according to claim 1 is completed.

The present invention relates to the "speech coding apparatus according to claim 6," wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as the feature quantity as the feature quantity. An example is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch are included as the feature amount in the claim 4 "feature amount. 1 is an embodiment relating to the "encoding device", and the proposed mode discrimination circuit 2000 of FIG.
And the adaptive codebook circuit 500, the proposed mode discrimination circuit 4000 and the adaptive codebook circuit 5000 of this embodiment.
Since their configurations are different from each other, the configurations of these and the proposed pitch extraction circuit 6000 will be described in this embodiment with reference to FIG.

The mode discrimination circuit 4000 receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, calculates the pitch prediction gain PG, and compares this with a predetermined plurality of thresholds to discriminate the mode. And output the mode information. Mode discrimination circuit 400
0 indicates that the result of mode discrimination is the adaptive codebook circuit 5000.
And excitation quantization circuit 350 and proposed pitch extraction circuit 6
Output to 000.

The proposed pitch extraction circuit 6000 receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, the mode discrimination information from the mode discrimination circuit 4000, and the output value of the proposed pitch extraction circuit 6000. Proposed pitch extraction circuit 6000
The pitch CP extracted in is output.

The structure of the proposed pitch extraction circuit 6000 is shown in FIG.

In FIG. 7, mode discrimination information is input from the input terminal 6010, a perceptual weighting signal is input from the input terminal 6020, and a pitch is input from the input terminal 6070.

The characteristic amount D calculating circuit 6040 calculates and outputs, for example, the current frame pitch CP, the past frame pitch PP, and the pitch ratio DR as the characteristic amount. here,
The pitch ratio DR is the rate of change in pitch when the time axis is taken in frame units.

In the characteristic amount D correction calculation circuit 6050, the current mode information from the input terminal 6010 and the delay device 603 are used.
According to the mode information of the previous frame stored in 0, the output value pitch ratio DR of 6040 is compared with a predetermined threshold value, and the pitch CP of the current frame is set to the pitch PP of the past frame. The corrected value CPP is output.

FIG. 8 shows the configuration of the characteristic amount D calculation circuit 6040. In FIG. 8, the pitch is input from the perceptual weighting signal input terminal 7080 in frame units from the input terminal 7010, the pitch CP is calculated by the pitch calculation circuit 7020, and output at the output terminal 7070. Further, the pitch ratio calculation circuit 7 is calculated using the pitch CP calculated in 7020 and the pitch PP of the past frame stored in the delay unit 7030.
The pitch ratio DR is calculated at 040, and this is output terminal 706.
Output by 0. The pitch PP of the past frame stored in the delay unit 7030 is also output from the output terminal 7050.

The adaptive codebook circuit 5000 is basically the same as the adaptive codebook circuit 500 of the first invention, but the pitch search range from the past signal is obtained by the proposed pitch extraction circuit 6000. It is characterized in that it is in the vicinity of the pitch CPP.

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch are included as the feature quantity in the speech coding apparatus according to claim 8. The description of the embodiment relating to "is completed.

The speech coding apparatus according to claim 5, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as a feature amount as the feature amount. An example is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch are included as the feature amount in the claim 8 "feature amount. The configuration of the pitch extraction circuit of FIG. 7 which is an embodiment relating to the "encoding device" is different from that of the pitch extraction circuit of this embodiment.
Will be explained.

The proposed pitch extraction circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis and the mode discrimination information from the mode discrimination circuit 4000, and outputs the extracted pitch CPP to the adaptive codebook circuit 5000.

The structure of the proposed pitch extraction circuit is shown in FIG.

In FIG. 13, mode discrimination information is input from an input terminal 13010, and a perceptual weighting signal is input from an input terminal 13020.

The characteristic amount D calculating circuit 13040 calculates and outputs, for example, the pitch CP of the current frame as a characteristic amount.

In the characteristic amount D correction calculation circuit 13050, the current mode information from the input terminal 13010 and the delay unit 1
A value C obtained by correcting the pitch CP of the current frame in accordance with the mode information of the previous frame stored in 3030.
Output PP.

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch are included as the feature quantity in the speech coding apparatus according to claim 8. The description of the embodiment relating to "is completed.

The speech coding apparatus according to claim 7, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as a feature amount as the feature amount. An example is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch is included as the feature amount in the claim 8 "feature amount. The feature amount D calculation circuit 6040 of FIG. 7 which is an embodiment relating to the encoding device and the feature amount D calculation circuit of the present embodiment have different configurations, which will be described with reference to FIG.

FIG. 14 shows the configuration of the feature amount D calculation circuit.
In FIG. 14, the perceptual weighting signal is input from the input terminal 14010 in frame units, and the pitch calculation circuit 140
The pitch CP is calculated at 20, and output at the output terminal 14070. Also, the pitch ratio DR is calculated by the pitch ratio calculation circuit 14040 using the pitch CP calculated in 14020 and the pitch PPP of the frame of the previous frame stored in the delay unit 14030, and this is calculated by the output terminal 14060. Output. Further, the pitch PP of the past frame stored in the delay device 14030 is also output from the output terminal 14050.

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch is included as the feature quantity in the speech coding apparatus according to claim 8. The description of the embodiment relating to "is completed.

An embodiment relating to claim 7 is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch are obtained as the feature amount as the feature amount in the eighth aspect. The pitch extraction circuit of FIG. 7 which is an embodiment related to the "encoding device" has a different structure from the pitch pitch extraction circuit of the proposed type of this embodiment. Therefore, in this embodiment, these structures will be described with reference to FIG. .

The proposed pitch extraction circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, the mode discrimination information from the mode discrimination circuit 4000 and the pitch from the proposed pitch extraction circuit, and the adaptive codebook circuit 5
000 and the pitch CP extracted to the proposed pitch extraction circuit.

The structure of the proposed pitch extraction circuit is shown in FIG.

In FIG. 15, mode discrimination information is input from an input terminal 15010, and a perceptual weighting signal is input from an input terminal 15020.

The characteristic amount D calculating circuit 15040 calculates and outputs, for example, the pitch CP and the pitch ratio DR of the current frame as the characteristic amount.

In the characteristic amount D correction calculation circuit 15050, the current mode information from the input terminal 15010 and the delay unit 1
A value obtained by comparing the output value pitch ratio DR of 15040 with a predetermined threshold value according to the mode information of the previous frame stored in 5030 and correcting the pitch CP of the current frame with the pitch ratio DR. Output CPP.

FIG. 1 shows the configuration of the feature amount D calculation circuit 15040.
6 is shown. In FIG. 16, the perceptual weighting signal is input from the input terminal 16010 in frame units, the pitch calculation circuit 16020 calculates the pitch CP, and the output terminal 160
Output at 70. Further, the pitch ratio calculation circuit 160 is calculated by using the pitch CP calculated in 16020 and the pitch PP of the frame of the previous frame stored in the delay unit 26030.
The pitch ratio DR is calculated at 40, and this is output terminal 1606
Output by 0.

With the above, the description of the embodiment according to claim 7 is completed.

An embodiment relating to claim 6 is shown in FIG.

In the present invention, the feature amount D calculating circuit 15040 of FIG. 15 which is an embodiment relating to claim 7 has a different structure from the proposed pitch extracting circuit of the present embodiment. Therefore, in the present embodiment, These configurations will be described with reference to FIG.

FIG. 17 shows the configuration of the characteristic amount D extraction / calculation circuit. In FIG. 17, the perceptual weighting signal and the pitch are input from the input terminal 17080 from the input terminal 17010 in frame units, and the pitch calculation circuit 17020 inputs the pitch CP.
Is calculated and output at the output terminal 17070. Also, 17
The pitch ratio DR is calculated by the pitch ratio calculation circuit 17040 using the pitch CP calculated in 020 and the pitch PP of the past frame stored in the delay device 17030, and this is output from the output terminal 17060.

With the above, the description of the embodiment according to claim 6 is completed.

An embodiment relating to claim 5 is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch are included as the feature amount in the claim 8 "feature amount. The proposed pitch extraction circuit 6000 of FIG. 6 which is an embodiment relating to the "encoding device".
Since the proposed pitch extraction circuit of this embodiment is different from that of this embodiment, only this configuration will be described with reference to FIG.

The proposed pitch extraction circuit receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, the mode discrimination information from the mode discrimination circuit 4000 and the pitch from the proposed pitch extraction circuit, and the adaptive codebook circuit 6
000 and the pitch CP extracted to the proposed pitch extraction circuit.

The structure of the proposed pitch extraction circuit is shown in FIG.

In FIG. 18, mode discrimination information is input from an input terminal 18010, a perceptual weighting signal is input from an input terminal 18020, and a pitch is input from an input terminal 18070.

The characteristic amount D calculating circuit 18040 outputs, for example, the pitch CP of the present frame and the pitch PP of the past frame as the characteristic amount.

In the characteristic amount D correction calculation circuit 18050, the current mode information from the input terminal 18010 and the delay unit 1
According to the mode information of the previous frame stored in 8030, the output value 18040 of the past frame is compared with a preset threshold value, and the pitch CP of the current frame is compared with that of the previous frame. The value CPP corrected with the pitch PP is output.

The configuration of the feature amount D calculation circuit 18040 is shown in FIG.
9 shows. 19, a perceptual weighting signal is input from an input terminal 19010 in frame units, a pitch CP is calculated by a pitch calculation circuit 19020, and an output terminal 190 is output.
Output at 70. Further, the output terminal 19060 outputs the pitch CP calculated in 19020 and the pitch PP of the past frame stored in the delay device 19030.

The description of the embodiment relating to claim 5 is finished.

An embodiment relating to claim 9 is shown in FIG.

According to the present invention, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch is included as the feature amount in the claim 4 "feature amount. 1 is an embodiment relating to the "encoding device", and the proposed mode discrimination circuit 2000 of FIG.
However, since the configuration is different from the proposed mode discrimination circuit 20000 of this embodiment, this and the proposed RM are used in this embodiment.
The configuration of the S extraction circuit 30000 will be described with reference to FIG.

The mode discrimination circuit 20000 receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, calculates the pitch prediction gain PG, and compares it with a plurality of predetermined thresholds to discriminate the mode. Output the mode information. Mode discrimination circuit 20
000 indicates the mode discrimination result based on the adaptive codebook circuit 50.
0, and outputs to the proposed RMS extraction circuit 30000 and the excitation quantization circuit 350.

The proposed RMS extraction circuit 30000 receives a voice signal from the frame division circuit 110 on a frame-by-frame basis, the mode discrimination information from the mode discrimination circuit 20000, and several RMS code vectors from the RMS codebook 40000. Output a vector.

The structure of the proposed RMS extraction circuit 30000 is shown in FIG.

In FIG. 21, mode discrimination information is input from the input terminal 31010, an audio signal in frame units is input from the input terminal 31020, and an RMS code vector signal is input from the input terminal 31080.

The RMS calculation circuit 31040 calculates the RMS value R in frame units.

In the RMS correction calculation circuit 31050, the current mode information from the input terminal 31010 and the delay unit 31 are used.
According to the mode information of the previous frame stored in 030, the output value R of 31040 is compared with a predetermined threshold value, and the value IR obtained by correcting the RMS value of the current frame is output.

RMS quantization vector selection circuit 31060
Then, the RMS correction calculation circuit 3105 is selected from the code vectors stored in advance in the RMS codebook 40000.
A vector close to the output value IR of 0 is selected and output.

With the above, the description of the embodiment according to claim 9 is completed.

The speech encoding apparatus according to claim 9, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as a feature amount as the feature amount. FIG. 22 shows an example.
Shown in

In the present invention, the proposed RMS extraction circuit 30000 of FIG. 20, which is an embodiment relating to claim 9, is
Since the configuration is different from the proposed RMS extraction circuit of this embodiment, this configuration will be described in this embodiment with reference to FIG.

In FIG. 22, mode discrimination information is input from the input terminal 32010, an audio signal in frame units is input from the input terminal 32020, and an RMS code vector signal is input from the input terminal 32080.

The RMS calculation circuit 32040 calculates the RMS value R in frame units.

In the RMS correction calculation circuit 32050, the current mode information from the input terminal 32010 and the delay unit 32
According to the mode information of the previous frame stored in 030 in the past and the RMS value of the past frame stored in the delay unit 32090, the output value R of 32040 is compared with a predetermined threshold value, A value IR obtained by correcting the RMS value of the current frame is output.

RMS quantization vector selection circuit 32060
Then, the RMS correction calculation circuit 3205 is selected from the code vectors stored in advance in the RMS codebook 40000.
A vector close to the output value IR of 0 is selected and output.

According to the above, at least one or more of a pitch prediction gain, a short-term prediction gain, a level, and a pitch is included as the feature quantity in the speech coding apparatus according to claim 12. The description of the embodiment relating to "is completed.

An embodiment relating to claim 10 is shown in FIG.

In the present invention, the structure of the proposed RMS extraction circuit of FIG. 20, which is an embodiment relating to claim 9, is different from that of the proposed RMS extraction circuit of the present embodiment. The configuration will be described with reference to FIG.

In FIG. 23, mode discrimination information is input from the input terminal 33010, an audio signal in frame units is input from the input terminal 33020, and an RMS code vector signal is input from the input terminal 33080.

The RMS calculation circuit 33040 calculates the RMS value R in frame units.

In the RMS correction calculation circuit 33050, the current mode information from the input terminal 33010 and the delay unit 33 are used.
The output value R of 33040 is compared with a predetermined threshold value according to the mode information of the previous frame stored in 030 and the RMS ratio RR stored in the RMS calculation circuit 43090, and the current frame is compared. A value IR obtained by correcting the RMS value of is output.

In the RMS ratio calculation circuit 43090, the RMS
The ratio between the output value R of the calculation circuit 33040 and the RMS value of the past frame stored in the delay unit 33090 is calculated and output.

RMS quantization vector selection circuit 33060
Then, the RMS correction calculation circuit 3305 is selected from the code vectors stored in advance in the RMS codebook 40000.
A vector close to the output value IR of 0 is selected and output.

With the above, the description of the embodiment according to claim 10 is completed.

According to a twelfth aspect of the present invention, the "speech coding apparatus according to the tenth aspect" is characterized in that at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as a feature amount as the feature amount. Example of FIG.
4 shows.

In the present invention, the structure of the proposed RMS extraction circuit of FIG. 24, which is an embodiment relating to claim 10, is different from that of the proposed RMS extraction circuit of the present embodiment.
In this embodiment, this structure will be described with reference to FIG.

In FIG. 24, mode discrimination information is input from the input terminal 34010, an audio signal in frame units is input from the input terminal 34020, and an RMS code vector signal is input from the input terminal 34080.

The RMS calculation circuit 34040 calculates the RMS value R in frame units.

In the RMS correction calculation circuit 34050, the current mode information from the input terminal 34010 and the delay unit 34 are used.
Mode information of the previous frame stored in 030 and RMS stored in the RMS ratio calculation circuit 44090.
According to the ratio RR, the output value R of 34040 is compared with a predetermined threshold value, and the RMS value of the current frame is corrected with the RMS value of the past frame stored in the delay unit 34090 to output a value IR.

In the RMS ratio calculation circuit 34090, the RMS
The ratio of the output value R of the calculation circuit 34040 and the RMS value of the past frame stored in the delay unit 34090 is calculated and output.

RMS quantization vector selection circuit 34060
Then, the RMS correction calculation circuit 3405 is selected from the code vectors stored in advance in the RMS codebook 40000.
A vector close to the output value IR of 0 is selected and output.

According to the above, at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as a feature quantity in the speech coding apparatus according to claim 12. The description of the embodiment relating to "is completed.

An embodiment relating to claim 11 is shown in FIG.

In the present invention, the structure of the proposed RMS extraction circuit of FIG. 24, which is an embodiment related to claim 11, is different from that of the proposed RMS extraction circuit of the present embodiment.
In the present embodiment, this structure will be described with reference to FIG.

In FIG. 25, mode discrimination information is input from the input terminal 35010, an audio signal in frame units is input from the input terminal 35020, and an RMS code vector signal is input from the input terminal 35080.

The RMS calculation circuit 35040 calculates the RMS value R in frame units.

In the RMS correction calculation circuit 35050, the current mode information from the input terminal 35010 and the delay unit 35 are used.
Mode information of the previous frame stored in 030 and the RMS stored in the RMS ratio calculation circuit 45090.
The output value R of 35040 is compared with a predetermined threshold value according to the ratio RR, and a value IR obtained by correcting the RMS value of the current frame is output.

In the RMS ratio calculation circuit 45090, the RMS
The ratio between the output value R of the calculation circuit 35040 and the RMS value of the frame two frames before in the past stored in the delay unit 55090 is calculated and output.

RMS quantization vector selection circuit 35060
Then, the RMS correction calculation circuit 3505 is selected from the code vectors stored in advance in the RMS codebook 40000.
A vector close to the output value IR of 0 is selected and output.

With the above, the description of the embodiment according to claim 11 is completed.

According to a twelfth aspect of the present invention, there is provided the "speech coding apparatus according to the eleventh aspect", wherein at least one of a pitch prediction gain, a short-term prediction gain, a level, and a pitch is included as the feature quantity as the feature quantity. Example of FIG.
6 is shown.

In the present invention, the structure of the proposed RMS extraction circuit of FIG. 25, which is an embodiment relating to claim 11, is different from that of the proposed RMS extraction circuit of the present embodiment.
In this embodiment, this structure will be described with reference to FIG.

In FIG. 26, mode discrimination information is input from the input terminal 36010, an audio signal in frame units is input from the input terminal 36020, and an RMS code vector signal is input from the input terminal 36080.

The RMS calculation circuit 36040 calculates the RMS value R in frame units.

In the RMS correction calculation circuit 36050, the current mode information from the input terminal 36010 and the delay unit 36 are used.
The mode information of the previous frame stored in 030 and the RMS stored in the RMS ratio calculation circuit 46090.
According to the ratio RR, the output value R of 36040 is compared with a predetermined threshold value, and the RMS value of the current frame is stored in the delay unit 36090, and the RMS value of the previous frame of the past is stored.
The value IR corrected with the MS value is output.

In the RMS comparison calculation circuit 46090, the RM
The ratio between the output value R of the S calculation circuit 36040 and the RMS value of the frame two frames before in the past stored in the delay device 56090 is calculated and output.

RMS quantization vector selection circuit 36060
Then, the RMS correction calculation circuit 3605 is selected from the code vectors stored in advance in the RMS codebook 40000.
A vector close to the output value IR of 0 is selected and output.

According to the above, at least one or more of the pitch prediction gain, the short-term prediction gain, the level, and the pitch are included as the feature quantity in the speech coding apparatus according to claim 12. The description of the embodiment relating to "is completed.

[0178]

As described above, according to the present invention,
In the speech coding apparatus, even if the frame length is shortened to 5 ms-10 ms or less in order to achieve low delay, good sound quality is obtained without causing sound quality deterioration due to time variation of mode discrimination, pitch extraction, or level extraction. This is possible and this advantage is extremely large.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is a block diagram of a proposed mode discrimination circuit.

FIG. 3 is a block diagram of a proposed mode discrimination circuit.

FIG. 4 is a configuration diagram of a feature amount calculation circuit B.

FIG. 5 is a configuration diagram of a feature amount calculation circuit C.

FIG. 6 is a configuration diagram of an embodiment of the present invention.

FIG. 7 is a block diagram of a proposed pitch extraction circuit.

FIG. 8 is a configuration diagram of a feature amount D extraction calculation circuit.

FIG. 9 is a configuration diagram of a feature amount calculation circuit C.

FIG. 10 is a block diagram of a proposed mode discrimination circuit.

FIG. 11 is a configuration diagram of a feature quantity calculation circuit B.

FIG. 12 is a block diagram of a proposed mode discrimination circuit.

FIG. 13 is a configuration diagram of a pitch extraction circuit.

FIG. 14 is a configuration diagram of a feature amount D calculation circuit.

FIG. 15 is a configuration diagram of a pitch extraction circuit.

FIG. 16 is a configuration diagram of a feature amount D extraction calculation circuit.

FIG. 17 is a block diagram of a proposed pitch extraction circuit.

FIG. 18 is a configuration diagram of a proposed pitch extraction circuit.

FIG. 19 is a configuration diagram of a feature amount D extraction calculation circuit.

FIG. 20 is a configuration diagram of an embodiment of the present invention.

FIG. 21 is a block diagram of a proposed RMS extraction circuit.

FIG. 22 is a configuration diagram of a proposed RMS extraction circuit.

FIG. 23 is a block diagram of a proposed RMS extraction circuit.

FIG. 24 is a block diagram of a proposed RMS extraction circuit.

FIG. 25 is a block diagram of a proposed RMS extraction circuit.

FIG. 26 is a block diagram of a proposed RMS extraction circuit.

FIG. 27 is a configuration diagram of a feature quantity calculation circuit B.

[Explanation of symbols]

 110 frame division circuit 120 sub-frame division circuit 200 spectrum parameter calculation circuit 210 spectrum parameter quantization circuit 211 LSP codebook 230 weighting circuit 235 subtraction circuit 240 response signal calculation circuit 310 impulse response calculation circuit 350 excitation quantization circuit 351 non-uniform pulse number Type sparse source codebook 355 gain codebook 360 weighting signal calculation circuit 365 gain quantization circuit 400 multiplexer 500, 550 adaptive codebook circuit 2000 proposed mode discrimination circuit 2010 frame-unit input terminal 2020 spectrum parameter input terminal 2030 frame-unit input Feature amount calculation circuit 2040 Feature amount calculation circuit B 2050 Mode discrimination circuit 2060 Mode information storage delay device 2070 mode Information output terminal 3030 Feature amount calculation circuit A 3040 Feature amount calculation circuit B 3050 Feature amount calculation circuit C 3060 Mode determination circuit 3070 Delay device 4000 Mode determination circuit 4020 RMS calculation circuit 4030 Delay device 4040 RMS ratio calculation circuit 5000 Adaptive codebook circuit 5030 Short-term prediction gain calculation circuit 5040 Delay device 5050 Short-term prediction gain ratio calculation circuit 6000 Proposed pitch extraction circuit 6030 Delay device 6040 Feature amount D calculation circuit 6050 Feature amount D correction calculation circuit 7020 Pitch calculation circuit 7030 Delay device 7040 Pitch ratio calculation circuit 8030 Short-term prediction gain calculation circuit 8040 Delay device 8050 Delay device 8080 Short-term prediction gain ratio calculation circuit 9030 Feature amount calculation circuit A 9040 Feature amount calculation circuit B 9050 Feature amount calculation circuit C 906 Mode discriminating circuit 9070 Delay device 10020 RMS calculating circuit 10030 Delay device 10040 RMS ratio calculating circuit 11020 RMS calculating circuit 11030 Delay device 11040 RMS ratio calculating circuit 12030 Feature amount calculating circuit A 12040 Feature amount calculating circuit B 12050 Feature amount calculating circuit C 12060 Mode Discrimination circuit 12070 Delay device 13030 Delay device 13040 Feature amount D calculation circuit 13050 Feature amount D correction calculation circuit 14020 Pitch calculation circuit 14030 Delay device 14040 Pitch ratio calculation circuit 15030 Delay device 15040 Feature amount D calculation circuit 15050 Feature amount D extraction calculation circuit 16020 Pitch calculation circuit 16030 Delay device 16040 Pitch calculation circuit 17020 Pitch calculation circuit 17030 Delay device 17040 Pitch ratio calculation circuit 18030 Delay device 18040 Feature amount D calculation circuit 18050 Feature amount D extraction calculation circuit 19020 Pitch calculation circuit 19030 Delay device 20000 Proposed type mode determination circuit 21030 Delay device 24030 Delay device 26030 Delay device 30000 Proposed RMS extraction circuit 31030 Delay device 31040 RMS calculation circuit 31050 RMS correction calculation circuit 31060 RMS quantization vector selection circuit 32030 delay device 3240 RMS calculation circuit 32050 RMS correction calculation circuit 32060 RMS quantization vector selection circuit 32090 delay device 33030 delay device 33040 RMS calculation circuit 33050 RMS correction calculation circuit 33060 RMS quantization Vector selection circuit 33090 Delay device 34030 Delay device 34040 RMS calculation circuit 34050 RMS correction calculation circuit 3 060 RMS quantization vector selection circuit 34090 delay device 35030 delay device 35040 RMS calculation circuit 35050 RMS correction calculation circuit 35060 RMS quantization vector selection circuit 35090 delay device 36030 delay device 36040 RMS calculation circuit 36050 RMS correction calculation circuit 36060 RMS quantization vector selection Circuit 36090 Delay device 40000 RMS codebook 43090 RMS ratio calculation circuit 44090 RMS ratio calculation circuit 45090 RMS ratio calculation circuit 46090 RMS ratio calculation circuit 55090 Delay device 56090 Delay device

Claims (12)

[Claims] 1. A frame division unit that divides an audio signal into predetermined frame units, a mode determination unit that calculates a feature amount from the audio signal and performs mode determination, and the audio signal is encoded based on the determination result. In the speech coding apparatus, the mode discrimination of the current frame is performed by using at least one kind of feature amount obtained from the present frame and at least one past frame, and the mode discrimination information obtained from at least one past frame. A speech coding apparatus having a function of: 2. The speech encoding apparatus according to claim 1, wherein a temporal change ratio of at least one type of feature quantity is included as the feature quantity as the feature quantity. 3. The feature amount includes a ratio of the two feature amounts as a feature amount with respect to each feature amount of two frames of either the present frame or at least one past frame. The speech coding apparatus according to Item 1. 4. The speech coding apparatus according to claim 1, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as the feature amount as the feature amount. 5. A frame division unit that divides an audio signal into predetermined frame units, a mode determination unit that calculates a feature amount from the audio signal and determines a mode, and a pitch extraction unit that extracts a pitch from the audio signal. In a voice encoding device that encodes the voice signal based on the determination result, it is obtained from at least one type of feature amount obtained from the current frame and at least one past frame and at least one past frame. A speech coding apparatus comprising a pitch extraction unit that corrects a pitch of a current frame using mode discrimination information. 6. The speech coding apparatus according to claim 5, wherein as the feature quantity, a temporal change ratio of at least one kind of feature quantity is included as the feature quantity. 7. The feature amount includes a ratio of the two feature amounts as a feature amount with respect to each feature amount of two frames of at least one of a current frame and at least one past frame. Item 5. The speech encoding device according to item 5. 8. The speech coding apparatus according to claim 5, wherein at least one of pitch prediction gain, short-term prediction gain, level, and pitch is included as the feature amount as the feature amount. 9. A frame division unit that divides an audio signal into predetermined frame units, a mode determination unit that calculates a feature amount from the audio signal and determines a mode, and a level extraction unit that extracts a level from the audio signal. In a voice encoding device that encodes the voice signal based on the determination result, it is obtained from at least one type of feature amount obtained from the current frame and at least one past frame and at least one past frame. A speech coding apparatus comprising a level extraction unit that corrects a level of a current frame using mode discrimination information. 10. The feature amount includes a time change ratio of at least one type of feature amount as the feature amount.
The speech encoding device described. 11. The feature amount includes a ratio of the two feature amounts as a feature amount with respect to each feature amount of two frames of either a current frame or at least one past frame. Item 9. The speech encoding device according to item 9. 12. A pitch prediction gain as the feature quantity,
10. At least one or more of a short-term prediction gain, a level, and a pitch are included as a feature amount, 9.
The audio encoding device according to 0 or 11.