JP3471542B2 - Audio coding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain excellent speech quality even when the bit rate is lowered by finding the position where predetermined conditions are met by a sound source quantization part, and searching for the best position in the search range of positions of pulses representing a sound source signal on the basis of the found position. SOLUTION: A spectrum parameter calculating circuit 200 finds spectrum parameters from an input speech signal and quantizes them. An adaptive code book circuit 300 finds delay corresponding to a pitch cycle from the speech signal and calculates a pitch prediction signal to predict a pitch. Then a sound source quantizing circuit 350 constitutes a sound source signal of the speech signal with M pulses whose amplitudes are not zero, finds the sample position corresponding to a pulse position where the predetermined conditions are met for the pitch prediction signal, and sets a range of a search for the position of a pulse on the basis of a position which is shifted from the found sample position by a predetermined number of samples, thereby searching for and outputting the best position for the set range.

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 bit rate and a high quality.

[Prior art]

As a method for efficiently encoding a voice signal, for example, "Code-e" by M. Schroeder and B. Atal.
xcited linear prediction: High quality speech at ve
rylow bit rates "(Proc.ICASSP, pp. 937-940, 1985) (Reference 1) and Klejin et al.," Improved
speech quality and efficeint vector quantization
CELP (Code Excited) described in a paper (reference 2) entitled "in SELP" (Proc.ICASSP, pp, 155-158,1988)
Linear Predictive Coding) is known. In this conventional example, the transmission side extracts a spectrum parameter representing a spectrum characteristic of the voice signal by using linear prediction (LPC) analysis from the voice signal for each frame (for example, 20 ms). Subframe (for example, 5ms)
And the parameters in the adaptive codebook (delay parameters and gain parameters corresponding to the pitch period) are extracted based on the past excitation signal for each subframe, and the pitch prediction of the speech signal of the subframe is performed by the adaptive codebook. . For the sound source signal obtained by pitch prediction,
The excitation signal is quantized by selecting an optimal excitation code vector from an excitation codebook (vector quantization codebook) consisting of noise signals of a predetermined type and calculating an optimal gain. The method of selecting the sound source code vector is to minimize the error power between the residual signal and the signal synthesized by the selected noise 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 and transmitted by the multiplexer unit. Since the operation and configuration on the receiving side are well known, description thereof will be omitted.

[0003]

However, the speech coding apparatus described above has a problem that a large amount of calculation is required to select the optimum excitation code vector from the excitation codebook. This is because, in the methods of Documents 1 and 2, in order to select a sound source code vector, filtering or convolution operation must be repeated for each code vector by the number of code vectors stored in the codebook. to cause. For example, when the number of bits of the codebook is B bits and the number of dimensions is N, if the filter or impulse response length in the filtering or convolution calculation is K, the calculation amount is N × K × 2 ^B × Only 8000 / N is required. As an example, if B = 10, N = 40, and K = 10, there is a problem that 819,20,000 times of calculations are required per second, resulting in an extremely large amount of calculations.

Therefore, various methods have been proposed as methods for reducing the amount of calculation required for the search of the sound source codebook. For example, ACELP (Argebraic Code Excited Linear P
rediction) method is, for example, “16 by C. Laflamme et al.
kbps wideband speech coding technique based on al
gebraic CELP "(Proc.ICASSP, pp.13-16,19
91) (Reference 3) and the like. According to the ACELP method, a sound source signal is represented by a plurality of pulses, and the pulse position of each pulse is selected from a candidate of a position predetermined for each pulse, and this is expressed by a predetermined number of bits for transmission. To do. Here, the amplitude of each pulse is +1.0 or −
Since it is limited to 1.0, the calculation amount of pulse search can be significantly reduced.

In the conventional method of Document 3, it is possible to significantly reduce the amount of calculation. However, when the bit rate is reduced, the number of pulses per subframe is rapidly reduced, and the sound quality is significantly deteriorated. There is a problem of doing.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to solve the above problems and provide a voice encoding system with a relatively small amount of calculation even when the bit rate is low and with little deterioration in sound quality.

[0007]

In order to solve the above-mentioned problems, the speech coding method according to the first aspect of the present invention comprises a spectrum parameter calculating unit for obtaining and quantizing a spectrum parameter from an input speech signal, and An adaptive codebook unit for calculating a pitch prediction signal by calculating a delay corresponding to a pitch period from a speech signal and performing pitch prediction; and a sound source signal of the speech signal is composed of a number M of non-zero-amplitude pulses. A range in which the sample position corresponding to the pulse position satisfying a predetermined condition with respect to the prediction signal is obtained, and the position of the pulse is searched based on the position shifted from the obtained sample position by a predetermined number of samples. And a sound source quantizer that searches for and outputs the best position for the set range.

Further, the speech coding apparatus according to the second aspect of the present invention comprises a spectrum parameter calculation section for obtaining and quantizing a spectrum parameter from an input speech signal, and a pitch prediction signal for obtaining a delay corresponding to a pitch period from the speech signal. An adaptive codebook section that calculates
The sound source signal of the audio signal is composed of a number M of pulses having non-zero amplitude, and a sample position satisfying a predetermined condition for the pitch prediction signal is obtained in a section having a length equal to the pitch cycle from the beginning. A sound source quantizing unit that sets a range for searching the position of the pulse based on a position shifted from the position by a predetermined number of samples, and searches for and outputs the best position with respect to the range.

A speech coder according to a third aspect of the present invention calculates a pitch prediction signal by obtaining a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal and a delay corresponding to a pitch period from the speech signal. Then, an adaptive codebook unit for performing pitch prediction and a sound source signal of the speech signal are composed of a number M of pulses of which amplitude is non-zero, and the pitch prediction signal is preliminarily compared with the pitch prediction signal in a section having a length equal to the pitch period from the beginning. A sample position satisfying a predetermined condition is obtained, a pulse position candidate is set while being shifted by the pitch period based on a position shifted by a predetermined number of samples from the position, and the candidate position is searched. And a sound source quantizer that outputs the best position.

Here, the excitation quantizer has a codebook for collectively quantizing the amplitudes or polarities of a plurality of pulses.

A speech coder according to a fourth aspect of the present invention calculates a pitch prediction signal by obtaining a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal and a delay corresponding to a pitch period from the speech signal. Then, an adaptive codebook unit for performing pitch prediction and a sound source signal of the speech signal are composed of a number M of pulses having non-zero amplitude, and a sample position satisfying a predetermined condition for the pitch prediction signal is obtained. A range for searching the position of the pulse is set based on the position after shifting from the position by using each of a plurality of types of shift amounts, and a position is searched for the range, and the best shift amount and pulse And a sound source quantization unit that outputs a combination of positions.

A speech coder according to a fifth aspect of the present invention calculates a pitch prediction signal by obtaining a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal and a delay corresponding to a pitch period from the speech signal. Then, an adaptive codebook unit for performing pitch prediction and a sound source signal of the speech signal are composed of a number M of pulses of which amplitude is non-zero, and the pitch prediction signal is preliminarily compared with the pitch prediction signal in a section having a length equal to the pitch period from the beginning. Obtain a sample position that satisfies the specified conditions, set a range to search the position of the pulse based on the position after shifting from the position using each of a plurality of types of shift amounts, and set the position with respect to the range. And a sound source quantization unit that outputs the best combination of the shift amount and the pulse position.

A speech coding apparatus according to a sixth aspect of the present invention calculates a pitch prediction signal by obtaining a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal and a delay corresponding to a pitch period from the speech signal. Then, an adaptive codebook unit for performing pitch prediction and a sound source signal of the speech signal are composed of a number M of pulses of which amplitude is non-zero, and the pitch prediction signal is preliminarily compared with the pitch prediction signal in a section having a length equal to the pitch period from the beginning. Obtaining a sample position that satisfies the specified condition, based on the position after being shifted from the position using each of a plurality of types of shift amount, further candidates for the position to which the pulse is to be shifted while being shifted by the pitch period. A sound source quantization unit that sets and searches for the position and outputs the best combination of the shift amount and the pulse position.

Here, the excitation quantizer has a codebook for collectively quantizing the amplitudes or polarities of a plurality of pulses.

A speech coding apparatus according to a seventh aspect of the present invention comprises a spectrum parameter calculation section for obtaining and quantizing a spectrum parameter from an input speech signal, and a feature quantity extracted from the input speech signal to discriminate a plurality of modes. And a mode discriminator that outputs the speech signal, an adaptive codebook unit that calculates a pitch prediction signal by obtaining a delay corresponding to a pitch period from the speech signal, and performs pitch prediction,
The amplitude of a non-zero pulse, in the case of a predetermined mode, to obtain a sample position that satisfies a predetermined condition for the pitch prediction signal, based on the position,
A sound source quantizer that sets a range for searching the position of the pulse and searches for and outputs the best for the range.

Here, the feature quantity is an average pitch prediction gain, and the mode discriminating section discriminates a mode based on a result of comparison between the average pitch prediction gain and a plurality of predetermined threshold values. .

A speech coding apparatus according to an eighth aspect of the present invention comprises a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal, and a pitch prediction by obtaining a delay corresponding to a pitch period from the speech signal. An adaptive codebook section that calculates a signal and performs pitch prediction,
A position satisfying a predetermined condition is found with respect to the pitch prediction signal obtained by the adaptive codebook, a search range of positions of a plurality of pulses representing a sound source signal is set based on the obtained position, and this search is performed. And a sound source quantizer for searching for the best position of the plurality of pulses in the range.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a first block diagram of a speech coder according to the present invention.
It is a block diagram showing an embodiment of. In FIG.
An audio signal is input from the input terminal 100, and the frame division circuit 110 outputs the audio signal into a frame (for example, 10 m).
s), and the sub-frame division circuit 120
The frame audio signal is divided into subframes (for example, 5 ms) shorter than the frame.

The spectrum parameter calculation circuit 200 is
For a voice signal of at least one subframe, a voice is cut out by applying a window longer than the subframe length (for example, 24 ms), and a spectrum parameter is calculated in a predetermined order (for example, P = 10th order). Here, the calculation of the spectral parameters includes the well-known LPC analysis and B
Urg analysis or the like can be used. Here, Bur
g analysis will be used. The details of the Burg analysis are described in the book “Signal Analysis and System Identification” by Nakamizo (page 1988), pages 82 to 87 (Corona Publishing Co., Ltd.), and therefore the description thereof is omitted. Furthermore,
The spectrum parameter calculation unit converts the linear prediction coefficient α _i (i = 1, ..., 10) calculated by the Burg method into an LSP parameter suitable for quantization and interpolation. Here, the conversion from linear prediction coefficient to LSP is performed by Sugamura et al., "Speech information compression by line spectrum pair (LSP) speech analysis and synthesis method" (Journal of the Institute of Electronics and Communication Engineers, J.
64-A, pp. 599-606, 1981) (reference 5). For example, the linear prediction coefficient obtained by the Burg method in the second subframe is converted into an LSP parameter, the LSP of the first subframe is obtained by linear interpolation, and the LSP of the first subframe is inversely transformed to obtain the linear prediction coefficient. , The linear prediction coefficient α _il of the first and second subframes, i = 1, ..., 10, l = 1 ,.
2) is output to the perceptual weighting circuit 230. Also, the second
The LSP of the subframe is output to the spectrum parameter quantization circuit 210.

Spectral parameter quantization circuit 210
Efficiently quantizes the LSP parameter of a predetermined subframe using the codebook 220, and outputs a quantized value that reduces the distortion of the following equation.

[Equation 1] Here, LSP (i), QLSP (i), and W (i) are the i-th order LSP and the codebook 22 before quantization, respectively.
It is the j-th code vector and weighting coefficient stored in 0.

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. A well-known method can be used as a method for vector quantization of the LSP parameter. A specific method is, for example, Japanese Patent Laid-Open No. 4-17150.
No. 0 (Japanese Patent Application No. 5-297600) (Reference 6) and Japanese Patent Laid-Open No. 4-363000 (Japanese Patent Application No. 3-261925).
(Japanese Patent Application Laid-Open No. Hei 3-6199)
−155049) (reference 8), T. Nomura et al., “LSP Coding Using VQSVQ with Interpolation in
4.075kbps M-LCELP Speech Coder ”(Pro
c.Mobile Multimedia Communications, pp.B.2.5, 1993)
(Reference 9) and the like can be referred to, so the description is omitted here.

Further, the spectrum parameter quantization circuit 2
10 restores the LSP parameter of the first subframe based on the LSP parameter 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 previous frame are linearly interpolated to restore the LSP of the first subframe. Here, after selecting one type of code vector for quantizing the error power of the LSP before quantization and the LSP after quantization, the LSP of the first subframe can be restored by linear interpolation.

The L of the first subframe restored as described above
A linear prediction coefficient α _il '(i = 1, ..., 10, l =
1, ..., 2), and the impulse response calculation circuit 310
Output to. In addition, the multiplexer 4 calculates the index representing the code vector of the quantized LSP of the second subframe.
Output to 00.

The perceptual weighting circuit 230 inputs the linear prediction coefficient α _ij ′ (i = 1, ..., P) before quantization for each subframe from the spectrum parameter calculation circuit 200, and the above-mentioned reference 1 is used. Based on this, the perceptual weighting is performed on the audio signal of the sub-frame, and the perceptual weighting signal is output.

The response signal calculation circuit 240 receives the linear prediction coefficient α _i for each sub-frame from the spectrum parameter calculation circuit 200, and the spectrum parameter quantization circuit 210 quantizes and interpolates and restores the linear prediction coefficient α _i. Input α _i 'for each subframe, calculate the response signal for one subframe with the input signal being zero d (n) = 0 using the value of the stored filter memory, and output it to the subtractor 235. To do. Here, the response signal x _z (n) is expressed by the following equation.

[Equation 2] However, when n−i ≦ 0

[Equation 3]

[Equation 4] Here, N indicates the subframe length. γ is a weighting coefficient that controls the perceptual weighting amount, and has the same value as the following equation (6). s _w (n) and p (n) respectively represent the output signal of the weighting signal calculation circuit, and the output signal of the denominator term of the filter of the first term on the right side in Expression (6) described later.

The subtractor 235 subtracts the response signal for one subframe from the perceptual weighting signal according to the following equation, and x _w '
(N) is output to the adaptive codebook circuit 300.

[Equation 5]

The impulse response calculation circuit 310 calculates the impulse response h _w (n) of the perceptual weighting filter whose z-transform is expressed by the following equation, by a predetermined score L, and the adaptive codebook circuit 300 and the excitation quantization. Output to the circuit 350.

[Equation 6]

The adaptive codebook circuit 300 receives the sound source signal v (n) from the weighted signal calculation circuit 360, the output signal x _w '(n) from the subtractor 235, and the perceptual weighted impulse from the impulse response calculation circuit 310. Response h
Enter _w (n). The delay T corresponding to the pitch period is calculated so as to minimize the distortion in the following equation, and the index representing the delay is output to the multiplexer 400.

[Equation 7] here,

[Equation 8] Indicates a pitch prediction signal, and the symbol * indicates a convolution operation. The gain β is calculated according to the following formula.

[Equation 9]

Here, in order to improve the extraction ellipticity of the delay with respect to a female sound or a child's voice, the delay may be obtained with a decimal sample number instead of an integer sample number. A concrete method is, for example, “Pitch predictors” by P. Kroon et al.
A paper entitled "with high temporal resolution" (Proc.I
CASSP, pp.661-664, 1990) (Reference 10) and the like can be referred to.

Further, the adaptive codebook circuit 300 performs pitch prediction according to the following equation using the selected delay and gain, and outputs the prediction residual signal z _w (n) to the excitation quantization circuit 35.
Output to 0.

[Equation 10] Further, the pitch prediction signal using the selected delay is output to the excitation quantization circuit 350.

Excitation quantization circuit 350 produces M pulses of non-zero amplitude for a subframe.

A block diagram showing the configuration of the excitation quantization circuit 350 is shown in FIG. The absolute maximum position detection circuit 351
A sample position satisfying a predetermined condition is detected with respect to the pitch prediction signal y _w (n). Here, the condition that “the absolute value of the amplitude is maximum” is used, and the sample position satisfying the condition is detected and output to the position search range setting circuit 352.

The position search range setting circuit 352 sets the search range of the position of each pulse after shifting the input sample position by a predetermined fixed number L of samples in the future or in the past.

For example, let D be the input sample position,
Considering an example in which five pulses are obtained in a 5 ms subframe (40 samples), examples of candidates for positions included in the search range of each pulse are as shown in the table below. First pulse D-L, D-L + 5 ,. ． ． Second pulse D-L + 1, D-L + 6 ,. ． ． Third pulse D-L + 2, D-L + 7 ,. ． ． Fourth pulse D-L + 3, D-L + 8 ,. ． ． Fifth pulse D-L + 4, D-L + 9 ,. ． ．

Next, by inputting z _w (n) and h _w (n), the first correlation function calculating circuit 353 and the second correlation function calculating circuit 354 respectively calculate the first correlation function according to the following equation. Function d
(N), the second correlation function φ is calculated.

[Equation 11]

[Equation 12]

The pulse polarity setting circuit 355 extracts and outputs the polarity of the first correlation function d (n) for the candidate position of each pulse in the search range set by the position search range setting circuit 352.

The pulse position searching circuit 356 calculates the following formula for the combination of candidate positions shown in the above table, and selects the position that maximizes the following formula as the optimum position.

[Equation 13] Here, if the number of pulses is M,

[Equation 14]

[Equation 15] Is. Here, sign (k) indicates the polarity of the k-th pulse, and the one extracted in advance by the pulse polarity setting circuit 355 is used. By the above, the polarities and positions of the M pulses are output to the gain quantization circuit 365.

Further, the pulse position is quantized by a predetermined number of bits, and an index representing the position is output to the multiplexer. In addition, the polarity of the pulse is output to the multiplexer 400.

The gain quantization circuit 365 reads the gain code vector from the gain code book 367, selects the gain code vector that minimizes the following expression for the selected position, and finally minimizes the distortion. Select the combination of amplitude code vector and gain code vector.

Here, the gain β'of the adaptive codebook
And an example of simultaneously vector-quantizing two types of gains of a sound source gain G ′ represented by pulses.

[Equation 16] Here, βt ′ and Gt ′ are the t-th element in the two-dimensional gain code vector stored in the gain codebook 367. The calculation of the above equation is repeated for each gain code vector, and the gain code vector that minimizes the distortion Dt is selected. The index representing the selected gain code vector is output to the multiplexer 400.

The weighting signal calculation circuit 360 inputs each index, reads the code vector corresponding to the index, and first obtains the driving sound source signal v (n) based on the following equation.

[Equation 17] v (n) is output to the adaptive codebook circuit 300.

Next, the spectrum parameter calculation circuit 20
Using the output parameter of 0 and the output parameter of the spectrum parameter quantization circuit 210, the response signal s _w (n) is calculated for each subframe by the following equation and output to the response signal calculation circuit 240.

[Equation 18]

FIG. 3 is a block diagram showing the second embodiment.
Shown in. Here, the operation of the excitation quantization circuit 450 is shown in FIG.
Different from

The structure of the excitation quantization circuit 450 is shown in FIG. The excitation quantization circuit 450 has a prediction signal y _w (n), a prediction residual signal z _w (n), and a perceptual weighting impulse response h _w.
Input not only (n), but the delay T of the adaptive codebook.

The maximum absolute value position calculation circuit 451 inputs the delay T corresponding to the pitch period, and outputs the pitch prediction signal y.
_{For w} (n), the sample position that maximizes the absolute value is detected in the range from the beginning of the subframe to T samples, and the position search range setting circuit 352 is output.

FIG. 5 is a block diagram showing the third embodiment.
Shown in. Here, the operation of the excitation quantization circuit 500 is shown in FIG.
Different from FIG. 6 shows a configuration diagram of the excitation quantization circuit 550.

The position search range setting circuit 552 uses the position shifted in the future or the past by a predetermined fixed number of samples L with respect to the input sample position as a base point, and shifts by the delay T to determine the position of each pulse. The candidates are set and output to the pulse position search circuit 356.

For example, let the input sample position be D,
Considering an example in which 5 pulses are obtained in a 5 ms subframe (40 samples), examples of candidates for the position of each pulse are as shown in the table below. First pulse D-L, D-L + T, ... Second pulse D-L + 1, D-L + T, ... Third pulse D-L + 2, D-L + T, ... Fourth pulse D-L + 3, D-L + T, ... Fifth pulse Pulse D-L + 4, D-L + T, ...

FIG. 7 is a block diagram showing the fourth embodiment.
Shown in. Here, an example in which the amplitude codebook is used in the first embodiment will be described, but the same changes can be made when the amplitude codebook is used in the second and third embodiments.

FIG. 7 is different from FIG. 1 in that the excitation quantization circuit 39
0 and the amplitude codebook 395 are different. The configuration of the excitation quantization circuit 390 is shown in FIG. Amplitude Codebook 3
Quantize the amplitude of the pulse using 95.

After the positions have been obtained for M pulses in the pulse position search circuit 356, the amplitude quantization circuit 3
At 97, the amplitude codevector is selected from the amplitude codebook 395 so as to maximize the following expression and the index is output.

[Formula 19] here,

[Equation 20]

[Equation 21] Is. Here, g _{k, j} is the j-th amplitude code vector of the k-th pulse.

The source quantization circuit 390 uses the multiplexer 4 as an index representing the selected amplitude code vector.
Output to 00. Also, the position value and the amplitude code vector value are output to the gain quantization circuit 400.

Although the amplitude codebook is used in this embodiment, a polarity codebook indicating the polarity of each pulse may be used instead for the search.

FIG. 9 is a block diagram showing the fifth embodiment. In the figure, the operation of the excitation quantization circuit 600 is different from that in FIG. 1, so the configuration will be described using FIG.

FIG. 10 is a block diagram showing the configuration of the excitation quantization circuit 600. The position search range setting circuit 652 is
With respect to the output position of the absolute maximum position detection circuit 351, a search range of each pulse and a set of positions are set with the position shifted by each of a plurality of types (for example, Q types) of shift amounts as a base point, The position setting sets are output to the pulse polarity setting circuit 655 and the pulse position searching circuit 656 by the amount of the shift amount.

The pulse polarity setting circuit 655 extracts the polarity for each of the plural types of candidate positions of the position searching circuit 652 and outputs it to the pulse position searching circuit 656.

The pulse position search circuit 656 searches for a position that maximizes the equation (13) by using the first correlation function, the second correlation function, and the polarity for each of the plurality of types of candidate positions. . This process is repeated Q times, which is a type of shift,
Among the seeds, the position that maximizes the equation (13) is finally selected, and the position of each pulse and the shift amount are output. In addition,
The shift amount is output to the multiplexer 400.

FIG. 11 is a block diagram showing a sixth embodiment. In the figure, the operation of the excitation quantization circuit 650 is different from that in FIG. 3, so the configuration will be described using FIG.

FIG. 12 is a block diagram showing the structure of the excitation quantization circuit 650. The position search range setting circuit 652 is
With respect to the output position of the absolute maximum position detection circuit 451, each pulse position is set with the position shifted by each of the displacement amounts of a plurality of types (for example, Q types) as a base point, and the pulse position is set. It outputs to the pulse polarity setting circuit 655 and the pulse position searching circuit 656 by the amount of the shift amount.

The pulse polarity setting circuit 655 extracts the polarity for each of the plurality of types of candidate positions of the position searching circuit 652 and outputs it to the pulse position searching circuit 656.

The pulse position searching circuit 656 searches for a position that maximizes the equation (13) by using the first correlation function, the second correlation function and the polarity for each of the plurality of types of candidate positions. . This process is repeated Q times, which is the type of shifting,
Among the seeds, the position that maximizes the equation (13) is finally selected, and the position of each pulse and the shift amount are output. In addition,
The shift amount is output to the multiplexer 400.

FIG. 13 is a block diagram showing the seventh embodiment. In the figure, the operation of the excitation quantization circuit 750 is different from that in FIG. 5, so the configuration will be described using FIG.

FIG. 14 is a block diagram showing the structure of the excitation quantization circuit 750. The position search range setting circuit 752
With respect to the output position of the absolute maximum position detecting circuit 451, the position of each pulse is set while further shifting by the delay T, with the position shifted by each of a plurality of types (for example, Q types) of shift amounts as a base point. . In this way, the set of the position of each pulse is output to the Q type pulse polarity setting circuit 655 and the pulse position searching circuit 656.

The pulse polarity setting circuit 655 extracts the polarity for each of the plurality of types of candidate positions of the position searching circuit 652, and extracts it into the pulse position searching circuit 656.

The pulse position search circuit 656 searches for a position that maximizes the equation (13) by using the first correlation function, the second correlation function, and the polarity for each of the plurality of types of candidate positions. . This process is repeated Q times, which is the type of shifting,
Among the seeds, the position that maximizes the equation (13) is finally selected, and the position of each pulse and the shift amount are output. In addition,
The shift amount is output to the multiplexer 400.

FIG. 15 is a block diagram showing the eighth embodiment. Here, an example in which an amplitude codebook for quantizing the pulse amplitude is added to the block diagram showing the form of the fifth embodiment is shown, but it is also possible to add to the sixth and seventh embodiments.

In the figure, the operation of the excitation quantization circuit 850 is different from that of FIG. 7, so the configuration of the speech quantization k circuit 850 will be described with reference to FIG.

FIG. 16 is a block diagram showing the structure of the excitation quantization circuit 85. The position search range setting circuit 652 sets the position of each pulse with respect to the output position of the absolute maximum position detection circuit 351 as the base point of the position shifted by a plurality of types (for example, Q types) of shift amounts. Is set and the pulse position set is output to the pulse polarity setting circuit 655 and the pulse position searching circuit 656 by the amount of the shift amount.

The pulse polarity setting circuit 655 extracts the polarity for each of the plural types of candidate positions of the position searching circuit 652 and outputs it to the pulse position searching circuit 656.

The pulse position search circuit 656 searches for a position that maximizes the equation (13) by using the first correlation function, the second correlation function, and the polarity for each of the plurality of types of candidate positions. . This process is repeated Q times, which is a type of shift,
Among the seeds, the position that maximizes the equation (13) is finally selected, and the position of each pulse and the shift amount are output. In addition,
The shift amount is output to the multiplexer 400. The amplitude quantization circuit 397 performs the same operation as in FIG.

FIG. 17 is a block diagram showing the ninth embodiment. Here, an example based on the first embodiment is shown, but other embodiments can also be used.

The mode discrimination circuit 900 receives the perceptual weighting signal from the perceptual weighting circuit 230 on a frame-by-frame basis, and outputs the mode discrimination information to the adaptive codebook circuit 950, the excitation quantization circuit 960, the gain quantization circuit 965 and the multiplexer 400. . Here, for mode discrimination,
The feature amount of the current frame is used. As the feature amount, for example, a pitch prediction gain averaged in frames is used.
For example, the following formula is used to calculate the pitch prediction gain.

[Equation 22] Here, L is the number of subframes included in the frame. Pi and Ei indicate the speech power and the pitch prediction error power in the i-th subframe, respectively.

[Equation 23]

[Equation 24] Here, T is the optimum delay that maximizes the prediction gain.

The frame average pitch prediction gain G is compared with a plurality of predetermined threshold values and classified into a plurality of types (for example, R types) of modes. As the number of modes R,
For example, 4 can be used.

The adaptive codebook circuit 950 receives the mode information, and in the case of a predetermined mode, performs the same operation as the adaptive codebook circuit 300 of FIG. 1, delay, adaptive codebook prediction signal, prediction residual error. Output a signal. For other modes, the input signal from the subtractor 235 is output as it is.

The source quantization circuit 960 receives the mode information and performs the same operation as the source quantization circuit 350 of FIG. 1 in the predetermined mode.

The gain quantization circuit 965 inputs the mode information, switches a plurality of types of gain codebooks 367 ₁ to 367 _R designed for each mode, and uses them for gain quantization. The operation of gain quantization is the same as that of the gain quantization circuit 365 shown in FIG.

The present invention is not limited to the embodiment described above, but various modifications are possible. For example, a codebook for quantizing the amplitudes of a plurality of pulses can be learned and stored in advance using a voice signal. The codebook learning method is
For example, Linde et al. “An algorithm for vector
Paper entitled "quantization design" (IEEE Trans.Commu
n., pp. 84-95, Januay, 1980) (Reference 11) and the like.

Instead of the amplitude codebook, it is possible to have a polarity codebook in which a combination of the polarities of each pulse is prepared by the number of bits equal to the number of pulses.

[0079]

As described above, according to the present invention,
In the excitation quantizer, a position satisfying a predetermined condition is obtained for the pitch prediction signal obtained by the adaptive codebook, and a search range of positions of a plurality of pulses representing the excitation signal is set based on the position. Then, search for the best position within this range. As a result, since the search range of the pulse position can be synchronized with the pitch waveform and the sound source signal for expressing the pitch waveform can be expressed satisfactorily, even if the bit rate is reduced, better sound quality than the conventional method can be obtained. can get.

Further, according to the present invention, the feature quantity is extracted from the input speech to discriminate a plurality of modes, and the sound source quantizing section performs the above-described processing in a predetermined mode to thereby obtain the periodicity of the speech. It can improve the sound quality for the strong mode part.

[Brief description of drawings]

FIG. 1 is a configuration block diagram showing a first embodiment of a speech encoding apparatus according to the present invention.

FIG. 2 is an excitation quantization circuit 35 according to the first embodiment.
It is a figure which shows the structure of 0.

FIG. 3 is a configuration block diagram showing a second embodiment of a speech encoding device according to the present invention.

FIG. 4 is an excitation quantization circuit 45 according to the second embodiment.
It is a figure which shows the structure of 0.

FIG. 5 is a configuration block diagram showing a third embodiment of a speech encoding apparatus according to the present invention.

FIG. 6 is a sound source quantization circuit 55 according to the third embodiment.
It is a figure which shows the structure of 0.

FIG. 7 is a configuration block diagram showing a fourth embodiment of a speech encoding apparatus according to the present invention.

FIG. 8 is a sound source quantization circuit 39 according to the fourth embodiment.
It is a figure which shows the structure of 0.

[Fig. 9] Fig. 9 is a configuration block diagram showing a fifth embodiment of a speech encoding device according to the present invention.

FIG. 10 is an excitation quantization circuit 6 according to the fifth embodiment.
It is a figure which shows the structure of 00.

FIG. 11 is a configuration block diagram showing a sixth embodiment of a speech coding apparatus according to the present invention.

FIG. 12 is an excitation quantization circuit 6 according to the sixth embodiment.
It is a figure which shows the structure of 50.

[Fig. 13] Fig. 13 is a configuration block diagram showing a seventh embodiment of a speech encoding device according to the present invention.

FIG. 14 is an excitation quantization circuit 7 according to the seventh embodiment.
It is a figure which shows the structure of 50.

[Fig. 15] Fig. 15 is a configuration block diagram showing an eighth embodiment of a speech encoding device according to the present invention.

FIG. 16 is an excitation quantization circuit 8 according to the eighth embodiment.
It is a figure which shows the structure of 50.

FIG. 17 is a configuration block diagram showing a ninth embodiment of a speech encoding device according to the present invention.

[Explanation of symbols]

110 frame division circuit 120 sub-frame division circuit 200 spectrum parameter calculation circuit 210 spectrum parameter quantization circuit 220 codebook 230 perceptual weighting circuit 235 subtraction circuit 240 response signal calculation circuit 310 impulse response calculation circuit 350, 390, 450, 500, 600, 650, 7
50, 850, 960 Excitation quantization circuit 360 Weighting signal calculation circuit 365, 965 Gain quantization circuit 395 Amplitude codebook 367 Gain codebook 400 Multiplexer 900 Mode discrimination circuit

Claims (12)

(57) [Claims] 1. A spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input voice signal, and an adaptive codebook unit for obtaining a pitch prediction signal by obtaining a delay corresponding to a pitch period from the voice signal and calculating a pitch prediction signal. And the sound source signal of the audio signal is composed of a number M of pulses of which amplitude is non-zero, the sample position corresponding to the pulse position satisfying a predetermined condition for the pitch prediction signal is obtained, and the obtained sample And a sound source quantizer that sets a range for searching the position of the pulse based on a position shifted by a predetermined number of samples from the position and searches for and outputs the best position for the set range. Speech coding device. 2. A spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input voice signal, an adaptive codebook unit for obtaining a pitch prediction signal by obtaining a delay corresponding to a pitch period from the voice signal, and an adaptive codebook unit, The sound source signal of the voice signal is composed of M pulses of non-zero amplitude, and a sample position satisfying a predetermined condition for the pitch prediction signal is obtained in a section having a length equal to the pitch period from the beginning and the position A speech encoding apparatus having a sound source quantization unit that sets a range for searching a pulse position based on a position shifted by a predetermined number of samples from, and searches for and outputs the best position with respect to the range. . 3. A spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input voice signal, an adaptive codebook unit for obtaining a pitch prediction signal by obtaining a delay corresponding to a pitch period from the voice signal, and an adaptive codebook unit, A sound source signal of a voice signal is composed of a number M of pulses having non-zero amplitude, and a sample position satisfying a predetermined condition with respect to the pitch prediction signal is obtained in a section having a length equal to a pitch cycle from the beginning, Based on the position shifted by a predetermined number of samples from the position, the pulse position candidates are set while being shifted by the pitch period, and a sound source quantizer that searches the candidate position and outputs the best position is used. Speech coding apparatus having. 4. The speech coding apparatus according to claim 1, wherein the excitation quantizer has a codebook for collectively quantizing the amplitudes or polarities of a plurality of pulses. 5. A spectrum parameter calculation unit that obtains and quantizes a spectrum parameter from an input speech signal, an adaptive codebook unit that obtains a delay corresponding to a pitch period from the speech signal, calculates a pitch prediction signal, and performs pitch prediction. A sound source signal is composed of a number M of non-zero-amplitude pulses, a sample position satisfying a predetermined condition for the pitch prediction signal is obtained, and the position is determined using each of a plurality of types of shift amounts. A sound source quantization unit that sets a range for searching the position of the pulse based on the position after the shift and searches for a position with respect to the range, and outputs a combination of the best shift amount and the position of the pulse. A speech encoding apparatus having. 6. A spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input voice signal, an adaptive codebook unit for obtaining a pitch corresponding to a pitch period from the voice signal, calculating a pitch prediction signal and performing pitch prediction, A sound source signal of a voice signal is composed of a number M of pulses of which amplitude is non-zero, and a sample position satisfying a predetermined condition for the pitch prediction signal is obtained in a section having a length equal to a pitch cycle from the beginning, Using each of the shift amounts of the seed, set a range for searching the position of the pulse based on the position after shifting from the position, and search the position with respect to the range, and the best shift amount and pulse A speech coding apparatus having a sound source quantization unit that outputs a combination of positions. 7. A spectrum parameter calculating section for obtaining and quantizing a spectrum parameter from an input speech signal, an adaptive codebook section for obtaining a pitch corresponding to a pitch period from the speech signal, calculating a pitch prediction signal and performing pitch prediction, A sound source signal of a voice signal is composed of a number M of pulses having non-zero amplitude, and a sample position satisfying a predetermined condition with respect to the pitch prediction signal is obtained in a section having a length equal to a pitch cycle from the beginning, Based on the position after shifting from the position by using each of the shift amounts of the seeds, the position of the pulse is set while shifting the pitch period, and the position is searched for. A speech encoding apparatus having an excitation quantizer that outputs a combination of a shift amount and a pulse position. 8. The speech coding apparatus according to claim 5, wherein the excitation quantizer has a codebook for collectively quantizing the amplitudes or polarities of a plurality of pulses. 9. A spectrum parameter calculation unit that obtains and quantizes a spectrum parameter from an input voice signal, a mode determination unit that extracts a feature amount from the input voice signal and determines and outputs a plurality of modes, and a voice signal from the voice signal. An adaptive codebook unit that calculates a pitch prediction signal by calculating a delay corresponding to a pitch period and a pitch prediction signal, and a sound source signal of the voice signal is composed of a number M of pulses whose amplitude is non-zero, and in the case of a predetermined mode. , Obtaining a sample position satisfying a predetermined condition with respect to the pitch prediction signal, setting a range for searching the position of the pulse based on the position, searching for and outputting the best for the range A speech coding apparatus comprising: a sound source quantization unit. 10. The speech coding apparatus according to claim 9, wherein the feature amount is an average pitch prediction gain. 11. The speech coding apparatus according to claim 9, wherein the mode discrimination unit discriminates a mode based on a result of comparison between the average pitch prediction gain and a plurality of predetermined threshold values. 12. A spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal, an adaptive codebook for performing pitch estimation by calculating a pitch prediction signal by obtaining a delay corresponding to a pitch period from the speech signal. And a position for satisfying a predetermined condition with respect to the pitch prediction signal obtained by the adaptive codebook, and setting a search range of a plurality of pulse positions representing a sound source signal based on the obtained position. A speech coding apparatus, comprising: a sound source quantization unit that searches for the best positions of the plurality of pulses within the search range.