JPH08185199A - Voice coding device - Google Patents

Abstract

PURPOSE: To provide a voice coding device in which good tone quality is obtained even for a low bit rate. CONSTITUTION: In a voice coding device consisting of a frame dividing section 110 dividing a frame with a previously decided frame unit, a spectrum parameter calculating section 200 obtaining a spectrum parameter from a voice signal, an adaptive code book section 500 cutting out a sound source signal of past delay and performing pitch prediction, a sound source quantizing section 350, an adaptive code book section in which delay in an adaptive code book is predicted from a difference quantization value and predicted difference is quantized is provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice coding device for coding a voice signal with high quality at a low bit rate.

[0002]

2. Description of the Related Art As a method for encoding a voice signal with high efficiency, for example, M. Schroederand B. At
“Code-excited linea” by al
rprediction: High quality
speed at very low bit r
ates "(Proc. ICASSP, pp. 937-
940, 1985), a paper (reference 1), and Kl.
"Improved speech" by eijn et al.
quality and efficient ve
ctor quantification in SEL
P "(Proc. ICASSP, pp. 155-15
CELP (Code Excited Lin) described in a paper (reference 2) entitled "8, 1988)."
Ear Predictive Coding) is known. In this conventional example, on the transmitting side, linear prediction (LPC) is performed from the audio signal for each frame (for example, 20 ms).
The analysis is used to extract spectral parameters representative of the spectral characteristics of the audio signal. The frame is further divided into subframes (for example, 5 ms), and parameters (delay parameters and gain parameters corresponding to the pitch period) in the adaptive codebook are extracted for each subframe based on the past excitation signal, and the adaptive codebook is used to extract the parameters. Pitch prediction of a subframe audio signal. For the sound source signal obtained by pitch prediction, the optimum sound source code vector is selected from the sound source codebook (vector quantization codebook) consisting of a noise signal of a predetermined type, and the optimum gain is calculated. Quantize the signal. The sound source code vector is selected so that the error power between the signal synthesized by the selected sound source code vector and the residual signal is minimized. 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 above-mentioned conventional method, the delay parameter is calculated for each subframe in the adaptive codebook and transmitted independently. For example, in the case of voice,
The delay exists in the range of 16-140 samples, but in order to obtain sufficient accuracy for female sounds with a short pitch period,
The delay should be in fractional sample increments rather than integer sample increments. Therefore, at least 8 bits are needed per subframe to represent the delay, and 4 bits per frame
32 per frame if subframes are accommodated
A bit needed. This was 1.6 kb / s in terms of the transmission amount per second when the frame length was 40 ms.

Therefore, when an audio signal is satisfactorily transmitted at 4 kb / s or less, it is necessary to reduce the information required for delay transmission. However, if the number of bits per subframe is simply reduced, there is a problem that the pitch change range is narrowed and the accuracy is insufficient, resulting in a significant deterioration in sound quality.

The present invention solves the above-mentioned problems and enables delay transmission with a small number of bits.
Good coding is possible at b / s or less.

[0006]

According to the first aspect of the present invention, a frame dividing unit that divides a voice signal into predetermined frame units, a spectrum parameter calculating unit that obtains a spectrum parameter from the voice signal, and a delay component past In a speech coding apparatus consisting of an adaptive codebook section for performing pitch prediction by cutting out an excitation signal and an excitation quantization section for quantizing an excitation signal, a delay in the adaptive codebook is predicted from a past difference quantization value, A speech coding apparatus having an adaptive codebook unit for quantizing a difference obtained by prediction is obtained.

According to the second aspect of the invention, a frame dividing section for dividing the voice signal into predetermined frame units, a spectrum parameter calculating section for obtaining a spectrum parameter from the voice signal, and a pitch by extracting a sound source signal in the past by the delay amount. In a speech coding apparatus including an adaptive codebook unit for prediction and an excitation quantization unit for quantizing an excitation signal,
An adaptive codebook unit that predicts a delay in the adaptive codebook from a past difference quantization value, and determines whether to quantize the difference or quantize the delay based on the difference obtained by prediction. A speech coding apparatus characterized by having.

According to the third aspect of the 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 spectrum parameter from the audio signal. In a speech coding apparatus including a spectrum parameter calculation unit that obtains, an adaptive codebook unit that cuts out a past excitation signal for delay and performs pitch prediction, and an excitation quantization unit that quantizes the excitation signal, a predetermined mode In the speech coding apparatus, a delay in the adaptive codebook is predicted from a past difference quantized value, and an adaptive codebook unit that quantizes the predicted difference is obtained.

According to the fourth aspect of the 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 spectrum parameter from the audio signal. In a speech coding apparatus including a spectrum parameter calculation unit that obtains, an adaptive codebook unit that cuts out a past excitation signal for delay and performs pitch prediction, and an excitation quantization unit that quantizes the excitation signal, a predetermined mode In, the adaptive codebook for predicting the delay in the adaptive codebook from the past difference quantization value and determining whether to quantize the difference or quantize the delay based on the difference obtained by prediction A speech coding apparatus having a section is obtained.

[0010]

1 is a block diagram showing an embodiment of a speech coder according to the first invention.

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, 40 ms), and a subframe division circuit 120 divides the voice signal of the frame into shorter than the frame. It is divided into subframes (for example, 8 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 a spectrum parameter is set to a predetermined order (for example, P = 10th order) is calculated. Here, well-known LPC analysis, Burg analysis, or the like can be used for the calculation of the spectrum parameter. here,
Burg analysis will be used. For details of Burg analysis, see "Signal Analysis and System Identification" by Nakamizo.
82-87 in the book titled "Corona Publishing Co., Ltd. 1988"
The description is omitted because it is described in the page (Reference 3) and the like.
Furthermore, in the spectrum parameter calculation unit, the linear prediction coefficient α _i (i = 1, ..., 10) calculated by the Burg method is used.
To LSP parameters 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 4). For example, the first, third, fifth
The linear prediction coefficient obtained by the Burg method in the subframe is converted into an LSP parameter, the LSP of the second and fourth subframes is obtained by linear interpolation, and the LSP of the second and fourth subframe is inversely transformed to obtain the linear prediction coefficient. Return to No. 1-5
Subframe 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 fifth subframe is output to the spectrum parameter quantization circuit 210.

The spectrum parameter quantization circuit 210 efficiently quantizes the LSP parameters of a predetermined subframe. In the following, it is assumed that vector quantization is used as the quantization method, and that the LSP parameter of the fifth 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 5) and Japanese Patent Application Laid-Open No. 4-363000 (Japanese Patent Application 3-26).
1925) (reference 6), JP-A-5-6199 (Japanese Patent Application No. 3-155049) (reference 7) and T.I. No
mura et al. , By "LSPCoding
Using VQ-SVQ With Interp
rotation in 4.075 kbps M-LC
A paper entitled "ELP Speech Coder" (P
rc. Mobile Multimedia Com
communications, pp. B. 2.5,199
3) (Reference 8) and the like can be referred to, so description thereof will be omitted here.

Further, the spectrum parameter quantization circuit 2
In 10, the LSP parameters of the first to fourth subframes are restored based on the LSP parameters quantized in the fifth subframe. Here, the quantized LSP parameter of the fifth subframe of the current frame and the quantized LSP of the fifth subframe of the previous frame are linearly interpolated to obtain the first LSP parameter.
To restore the LSP of the fourth subframe. Here, after selecting one type of code vector that minimizes the error power between the LSP before quantization and the LSP after quantization, the LSPs of the first to fourth subframes can be restored by linear interpolation. In order to further improve the performance, after selecting a plurality of code vectors for minimizing the error power, for each candidate, the cumulative distortion is evaluated, and a combination of the candidate for minimizing the cumulative distortion and the interpolation LSP is determined. Can be selected. For details, refer to, for example, Japanese Patent Application No. 5-8737 (Reference 9).

The linear prediction coefficient _α'il (i = 1, ..., 10, l =) for each subframe of the LSP of the first to fourth subframes and the quantized LSP of the fifth subframe restored as described above.
1, ..., 5), and the impulse response calculation circuit 310
Output to. In addition, the multiplexer 40 uses an index representing the code vector of the quantized LSP of the fifth subframe.
Output to 0.

In the above, instead of linear interpolation, LS
The P interpolation pattern has a predetermined number of bits (for example, 2
Bits), and a set of a code vector and an interpolation pattern that minimizes cumulative distortion by restoring LSPs of 1 to 4 subframes for each of these 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 may be created by learning in advance using LSP data for training, or a predetermined pattern may be stored. As the predetermined pattern, for example, T. Taniguchi et
"Improved CELP speed" by al
ch coding at 4kb / s and be
Low ”(Proc. ICSLP, pp.
41-44, 1992) (Literature 10). 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 coded as 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, the perceptual weighting is performed on the audio signal of the sub-frame based on the reference 1, and the perceptual weighting signal x _w (n) is output.

The response signal calculation circuit 240 receives the linear prediction coefficient α _il for each subframe from the spectrum parameter calculation circuit 200, and quantizes and interpolates and restores the linear prediction coefficient from the spectrum parameter quantization circuit 210. α ′ _il is input for each subframe, a response signal for the input signal d (n) = 0 is calculated for one subframe by using the stored value of the filter memory, and is output to the subtractor 235. Here, the response signal x _z (n) is expressed by the following equation.

[0019]

[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 by the following equation, by subtracting one subframe a response signal from the perceptual weighting signals, x _'w
(N) is output to the adaptive codebook circuit 500.

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

[0023]

[Equation 2]

The configuration of the adaptive codebook circuit 500 is shown in FIG.
Shown in In FIG. 2, the delay calculation unit 510 has a terminal 5
From each of 01, 502, and 503, the past sound source signal v
(N), the output signal x ′ _w (n) of the subtractor 235, and the impulse response h _w (n) are input, and the delay T corresponding to the pitch is input.
Is calculated so as to minimize the following equation.

[0025]

(Equation 3)

Here, y _w (nT) = v (nT) * h _w (n) (5), and the symbol * represents a convolution operation.

The gain β is calculated according to the following equation.

[0028]

[Equation 4]

Here, in order to improve the accuracy of extracting the delay with respect to the female sound and the voice of the child, the delay may be obtained with a decimal sample value instead of an integer sample value. A specific method is described in P. "Pitch" by Kroon et al.
predictors with high temp
Oral resolution ”(Pr
oc. ICASSP, pp. 661-664, 1990
(Year) (Reference 11) and the like can be referred to.

The delay predictor 520 inputs the delay T,
Furthermore, the difference quantized value of the delay of the past subframe is calculated from the subframe delay unit 540 and the prediction coefficient codebook 5
The prediction coefficient is input from 25, and MA (Moving Average) is estimated for the delay of the present sub-frame.
As an example, the case where a quantized value of one past subframe is used for prediction is shown in the following equation.

T _h = ηe _h ^l-1 (7) where η is a fixed prediction coefficient stored in the prediction coefficient codebook. The difference quantization unit 530 calculates the difference according to the following formula.

E ^l = T−T _h (8) The difference value e ^l is represented by a predetermined number of quantization bits and quantized to obtain a quantized value e _h ^l, which is output to the delay restoration unit 550. The quantized value e _h ^l is output to the subframe delay unit 540. Also, the index representing the quantized value e _h ^l is output from the terminal 505.

The delay restoration unit 550 restores and outputs the delay T'according to the following equation.

T ′ = T _h + e _h ^l (9) Further, the pitch prediction section 560 performs pitch prediction according to the following equation, and outputs the adaptive codebook prediction difference signal z (n) from the terminal 504.

Z (n) = x ′ _w (n) −βv (n−T ′) * h _w (n) (10) This completes the description of the adaptive codebook circuit 500.

Excitation quantization circuit 350 shows an example of searching an excitation codebook. Sound source code book 35
The source signal is quantized by searching the code vector stored in 1. The search for the source code vector selects the best source code vector c _j (n) so as to minimize the equation. At this time, one of the best code vectors may be selected, or two or more types of code vectors may be selected and the main selection may be made to one when gain quantization is performed. Here, it is assumed that two or more types of code vectors are selected.

[0037]

(Equation 5)

When the following formula is applied to only some sound source code vectors, a plurality of sound source code vectors are preselected, and the following formula is applied to the preselected sound source code vectors. Can also be applied.

The gain quantization circuit 365 reads the gain code vector from the gain code book 355, and combines the excitation code vector and the gain code vector so as to minimize the following expression with respect to the selected excitation code vector. Select.

[0040]

(Equation 6)

Here, β ′ _k and γ ′ _k are the kth code vector in the two-dimensional gain codebook stored in the gain codebook 355. 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, outputs the driving sound source signal v (n) based on the following equation. Ask.

V (n) = β ′ _k v (n−T) + γ ′ _k c _j (n) (13) Next, the output parameters of the spectrum parameter calculation circuit 200 and the spectrum parameter quantization circuit 210 are The response signal s _w (n) is calculated for each subframe using the following equation and is output to the response signal calculation circuit 240.

[0044]

(Equation 7)

This is the end of the description of the embodiment corresponding to the first invention.

A block diagram showing an embodiment of the second invention is shown in FIG. In the second invention, the operation of the adaptive codebook circuit 600 is different from that of the first invention, so the operation of the adaptive codebook circuit 600 will be described with reference to FIG. Note that, in FIG. 4, the constituent elements given the same numbers as in FIG. 2 perform the same operations as in FIG.

The discriminator 610 inputs the delay predictive value T _h output from the delay predictor 520 and the delay T of the current subframe from the delay calculator 510, and obtains an error by the following equation.

E ^l = T−T _h (15) For example, the absolute value of the error e ^l is compared with a predetermined threshold value, and when it is smaller than the threshold value, prediction is used and When it is larger, a prediction determination signal that no prediction is made is obtained and output to the switches 620 ₁ and 620 ₂ and the terminal 506.

The switch 620 ₁ inputs the prediction judgment signal, and when the prediction is not made, the switch is tilted upward, and when the prediction is made, the switch is tilted downward, and when the prediction is not made, the delay calculation unit 510 outputs the delay judgment signal. The output T is output to the pitch prediction unit 560 as the output T ′ from the delay restoration unit 550 when there is prediction. Switch 620 ₂ receives the prediction discrimination signal, an index corresponding to the delay T when no prediction, when the Yes prediction, and outputs the index of the differential quantization to the terminal 505.

This is the end of the description.

FIG. 5 is a block diagram showing an embodiment of the third invention. In the figure, the components with the same numbers as in FIG. 1 perform the same operations as in FIG. In FIG. 5, the mode discrimination circuit 700 receives a perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, and outputs mode discrimination information. Here, the feature amount of the current frame is used for the mode determination. For example, a pitch prediction gain is used as the feature amount. The pitch prediction gain is calculated using, for example, the following formula.

[0052]

(Equation 8)

Here, T is the optimum delay that maximizes the prediction gain.

The pitch prediction gain is compared with a plurality of predetermined threshold values and classified into a plurality of types of modes. For example, 4 can be used as the number of modes.

The mode discrimination circuit 700 outputs the mode discrimination information to the adaptive codebook circuit 800.

The configuration of the adaptive codebook circuit 800 is shown in FIG.
Shown in In the figure, the components having the same numbers as those in FIGS. 2 and 4 have the same functions as those in FIGS. In FIG. 6, the switches 820 ₁ and 820 ₂ are
Mode discrimination information is input from the terminal 801, and the presence / absence of delay prediction is switched according to the mode.

Further, the operation of the pitch predicting section 860 is changed according to the mode information. For example, the adaptive codebook circuit may not be used only in a predetermined mode (for example, mode 0). To do this,
In the calculation of the pitch prediction unit 860, the gain β may be set to 0 when the expression (9) is executed.

FIG. 7 is a block diagram showing an embodiment of the fourth invention. In the figure, the components having the same numbers as those in FIGS. 1, 3 and 5 perform the same operations, and therefore the description thereof will be omitted. Since the operation of the adaptive codebook circuit 900 is different in FIG. 7, this configuration is shown in FIG. In FIG.
Since the components having the same numbers as those in FIGS. 4 and 6 perform the same operation, the description thereof will be omitted. In FIG. 8, the terminal 9
The mode information is input from 01 and output to the determination unit 910. The discrimination unit 910 discriminates the prediction residual for a predetermined mode, and outputs a discrimination signal with / without prediction to the switches 620 ₁ and 620 ₂ . No prediction is made in modes other than the predetermined mode.

This completes the description of the embodiment of the present invention.

The present invention is not limited to the above embodiment, but various modifications are possible.

In the adaptive codebook circuit, the delay predicting section 520 may perform high-order prediction in which the delay is predicted from the differential quantized values of a plurality of past frames. Let the prediction order be L
Then, the following formula is used as the prediction formula.

[0062]

[Equation 9]

The prediction coefficient codebook may be switched for each mode.

As the configuration of the excitation codebook of the excitation quantization circuit, other well-known configurations such as multistage configuration and sparse configuration can be used.

The excitation codebook in the excitation quantization circuit may be switched using the mode discrimination information.

In the excitation quantization circuit, an example of searching the excitation codebook has been shown, but multipulses having a plurality of positions and different amplitudes may be searched. Here, the amplitude and position of the multi-pulse are set so as to minimize the following equation.

[0067]

[Equation 10]

Here, g _j and m _j respectively indicate the amplitude and position of the j-th multi-pulse. k is the number of multi-pulses.

[0069]

As described above, according to the present invention,
By predicting the delay using the past difference quantized value in the audio encoding device, the number of bits required to express the delay can be reduced from about 8 bits to about 5 bits per subframe. This can be reduced from 1.6 kb / s to 1 kb / s by setting the delay transmission amount per second, so that the coding rate of the entire voice is 4 kb / s.
It is easy to reduce to the following, and even if it is reduced, there is an effect that a better sound quality can be obtained than in the past.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of the first invention.

FIG. 2 is a diagram showing a configuration of an adaptive codebook circuit 500.

FIG. 3 is a diagram showing an embodiment of the second invention.

FIG. 4 is a diagram showing a configuration of an adaptive codebook circuit 600.

FIG. 5 is a diagram showing an embodiment of the third invention.

FIG. 6 is a diagram showing a configuration of an adaptive codebook circuit 800.

FIG. 7 is a diagram showing an embodiment of the fourth invention.

FIG. 8 is a diagram showing a configuration of an adaptive codebook circuit 900.

[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 500, 600, 800, 900 adaptive codebook circuit 310 impulse response calculation Circuit 350 Excitation Quantization Circuit 351 Excitation Codebook 355 Gain Codebook 360 Weighting Signal Calculation Circuit 365 Gain Quantization Circuit 400 Multiplexer 510 Delay Calculation Unit 520 Delay Prediction Unit 525 Prediction Coefficient Codebook 530 Difference Quantization Unit 540 Subframe Delay Unit 550 Delay recovery unit 560,860 Pitch prediction unit 620 ₁ , 620 ₂ , 820 ₁ , 820 ₂ switch circuit 700 mode discrimination circuit

Claims (5)

[Claims] 1. A frame division unit that divides a voice signal into predetermined frame units, a spectrum parameter calculation unit that obtains a spectrum parameter from the voice signal, an adaptive codebook that cuts out a sound source signal that is past the delay amount and performs pitch prediction. In a speech coding apparatus comprising a unit and a source quantization unit for quantizing a source signal, an adaptive code for predicting a delay in the adaptive codebook from a past difference quantization value, and quantizing a difference obtained by prediction A speech coding apparatus having a book section. 2. A frame division unit that divides a voice signal into predetermined frame units, a spectrum parameter calculation unit that obtains a spectrum parameter from the voice signal, an adaptive codebook that cuts out a sound source signal that is a delay past and performs pitch prediction. In a speech coding apparatus comprising a unit and a source quantization unit for quantizing a source signal, a delay in the adaptive codebook is predicted from a past difference quantization value, and the difference is obtained based on the difference obtained by prediction. A speech coding apparatus having an adaptive codebook section for discriminating whether to quantize or quantize the delay. 3. 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 spectrum parameter calculation unit that obtains a spectrum parameter from the audio signal. And a speech coding apparatus comprising an adaptive codebook section for cutting out a past excitation signal by a delay amount for pitch prediction and an excitation quantization section for quantizing the excitation signal, and the adaptive codebook in a predetermined mode. A speech coding apparatus, comprising: an adaptive codebook unit that predicts a delay in 1) from a past difference quantized value, and quantizes a difference obtained by prediction. 4. 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 spectrum parameter calculation unit that obtains a spectrum parameter from the audio signal. And a speech coding apparatus comprising an adaptive codebook section for cutting out a past excitation signal by a delay amount for pitch prediction and an excitation quantization section for quantizing the excitation signal, and the adaptive codebook in a predetermined mode. Is predicted from the past difference quantization value, and has an adaptive codebook unit that determines whether to quantize the difference or to quantize the delay based on the difference obtained by prediction. Speech coding device. 5. The excitation signal is quantized by searching for an excitation codebook composed of a plurality of types of code vectors in the excitation quantization unit, according to claim 1, 2, 3 or 4. Speech coding device.