JP3360545B2 - Audio coding device - Google Patents

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 at a low bit rate with high quality.

[0002]

2. Description of the Related Art As a method for encoding a speech signal with high efficiency, for example, M. Schroeder and B. Atal's "Code
-excited linear prediciton: High quality speech at
verylow bit rates ”(Proc. ICASSP, pp. 937-940,19
1985), and by Kleijn et al.
Improved speech quality and efficeint vector quant
ization in SELP ”(Proc. ICASSP, pp. 155-158, 1988
CE) described in a paper (Reference 2) titled
LP (Code Excited Linear Predictive Coding) is known. In this conventional example, on the transmission side, linear prediction (LPC) is performed from a speech signal every frame (for example, 20 ms).
The analysis is used to extract spectral parameters that represent the spectral characteristics of the audio signal. The frame is further divided into subframes (for example, 5 ms), and parameters (a delay parameter and a gain parameter corresponding to a pitch period) in the adaptive codebook are extracted for each subframe based on a past sound source signal. Pitch prediction is performed on the audio signal of the subframe. For an excitation signal obtained by pitch prediction, an optimal excitation code vector is selected from an excitation codebook (vector quantization codebook) composed of predetermined types of noise signals, and an optimal gain is calculated. , Quantize the sound source signal. 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 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. Description on the receiving side is omitted.

[0003]

The conventional method has a problem that a large amount of calculation is required to select an optimal excitation code vector from an 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 is once performed on each code vector, and this operation is operated by the number of code vectors stored in the code book. Due to return. For example, when the number of bits in the codebook is B and the number of dimensions is N, the amount of operation is N × K × 2 per second, where K is the filter or impulse response length in the filtering or convolution operation.
Only B × 8000 / N is required. As an example, B = 1
Assuming 0, N = 40 and k = 10, 81,9 per second
20,000 operations are required, which is extremely large.

[0004] Various methods have been proposed as a method for reducing the amount of calculation required for searching the sound source codebook. For example, ACELP (Argebraic Code Excited Linear Pred
iction) method has been proposed. This is, for example, C.La
“16 kbps wideband speech coding tec” by flamme et al.
hnique based on algebraic CELP ”(Pro
c. ICASSP, pp. 13-16, 1991) (Reference 3). According to the method of Document 3, the sound source signal is represented by a plurality of pulses, and the position of each pulse is represented by a predetermined number of bits and transmitted. Here, the amplitude of each pulse is +
Since it is limited to 1.0 or -1.0, the calculation amount of the pulse search can be greatly reduced.

In the conventional method of Reference 3, the amount of calculation can be greatly reduced, but there is a problem that the sound quality is not sufficient. The reason for this is that each pulse has only positive and negative polarities and the absolute value amplitude is always 1.0 irrespective of the pulse position, so that the amplitude is quantized extremely coarsely. The sound quality had deteriorated.

Further, according to the methods of the above-mentioned references 1-3, when searching for a sound source codebook or a pulse, the search is performed on the assumption that the gain multiplied by the sound source signal is constant.
Therefore, when the bit rate is reduced and the size of the sound source codebook is small, or when the number of pulses is small, the performance is reduced.

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

[0008]

According to a first aspect of the present invention, there is provided a speech coding apparatus comprising: a spectrum parameter calculating section for obtaining a spectrum parameter from an input speech signal and quantizing the spectrum parameter; consists zero pulse, said excitation quantization section searches and outputs the position of the pulse while changing a gain in every two or more number rather smaller than M when searching the location of the pulse with the spectrum parameters And characterized in that:

A speech encoding apparatus according to a second aspect of the present invention is characterized in that, in the sound source quantization section according to the first aspect of the present invention, a code book is provided for quantizing the amplitudes or polarities of a plurality of pulses collectively. .

A speech encoding apparatus according to a third aspect of the present invention comprises a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter from an input speech signal, and wherein the sound source signal of the speech signal is composed of M non-zero pulses. a first excitation quantization unit that searches and outputs the position of the pulse while changing the gain for each small rather more number M1 than M when searching the location of the pulse using the spectral parameters, the A second sound source quantization unit that searches for and outputs a predetermined number of pulse positions using spectral parameters, and uses outputs of the first sound source quantization unit and the second sound source quantization unit. And calculating a distortion with the voice and selecting a sound source quantization unit having a small distortion.

A speech encoding apparatus according to a fourth aspect of the present invention is characterized in that, in the sound source quantization section according to the third aspect of the present invention, a code book is provided for quantizing the amplitudes or polarities of a plurality of pulses collectively. .

According to a fifth aspect of the present invention, in the speech encoding apparatus according to the third or fourth aspect, a feature amount is obtained from an input speech signal, a plurality of types of modes are determined from the feature amount, and mode information is output. Using the first sound source quantization unit according to the mode information or the second
It is characterized in that whether or not to use the sound source quantization unit is switched.

[0013]

[0014]

FIG. 1 is a block diagram showing a first embodiment of a speech coding apparatus according to the present invention. In the figure, an audio signal is input from an input terminal 100, and a frame dividing circuit 110 converts the audio signal into a frame (eg,
s), and the subframe division circuit 120 divides the audio signal of the frame into subframes (for example, 5 ms) shorter than the frame.

The spectrum parameter calculation circuit 200 cuts out the voice signal of at least one sub-frame by applying a window (for example, 24 ms) longer than the sub-frame length, and sets the spectrum parameter to a predetermined order (for example, , P = 10th order). Here, a well-known LPC analysis, Burg analysis, or the like can be used for calculating the spectrum parameters. here,
Burg analysis will be used. For details of Burg analysis, see "Signal Analysis and System Identification" by Nakamizo.
82-87 in a book entitled Corona Publishing Co., 1988
The description is omitted because it is described on page (Literature 4) and the like.

Further, the spectrum parameter calculation section calculates the linear prediction coefficient αi (i
= 1,..., 10) into LSP parameters suitable for quantization and interpolation. Here, the conversion from the linear prediction coefficient to the LSP is performed by a paper titled “Speech Information Compression by Line Spectrum Pair (LSP) Speech Analysis / Synthesis Method” by Sugamura et al. (Transactions of the Institute of Electronics and Communication Engineers, J64-A, pp. 599) −606,
1981) (Reference 5). For example, the linear prediction coefficient obtained by the Burg method in the second sub-frame is converted into LSP parameters, the LSP of the first sub-frame is obtained by linear interpolation, and the LSP of the first sub-frame is inversely converted to obtain a linear prediction coefficient. , And the linear prediction coefficients αiι (i = 1,..., 10,
ι = 1,..., 2) are output to the audibility weighting circuit 230. Further, it outputs the LSP of the second subframe to spectrum parameter quantization circuit 210.

The spectrum parameter quantization circuit 210 efficiently quantizes LSP parameters of a predetermined subframe using the codebook 220, and outputs a quantization value for minimizing distortion of the following equation.

(Equation 1)

Here, LSP (i), QLSP
(I) _j and W (i) are the i-th L before quantization.
SP, the j-th code vector stored in the code book 220, and a weighting coefficient.

In the following, it is assumed that vector quantization is used as a quantization method, and that the LSP parameter of the second subframe is quantized. A well-known method can be used for the method of vector quantization of LSP parameters.
Specific methods are described in, for example, JP-A-4-171500 (Japanese Patent Application No. 2-297600) (Reference 6) and JP-A-4-171500.
No. 363000 (Japanese Patent Application No. 3-261925) (Patent Document 7) and Japanese Patent Application Laid-Open No. 5-6199 (Japanese Patent Application No. 3-15).
No. 5049) (Reference 8) and T. Nomura et al.,
LSP Coding Using VQ-SVQWith Interpolation in 4.075
kbps M-LCELPSpeech Coder ”(Proc. Mob
ile Multimedia Communications, pp. B.2.5, 1993) (Reference 9) and the like can be referred to, and the description is omitted here.

The spectrum parameter quantization circuit 2
At 10, the LSP parameters of the first sub-frame are restored based on the LSP parameters quantized in the second sub-frame. Here, the LSP of the first subframe is restored by linearly interpolating the quantized LSP parameter of the second subframe of the current frame and the quantized LSP of the second subframe of the previous frame. Here, LSP before quantization
After selecting one type of code vector that minimizes the error power between the LSP and the quantized LSP, the LSP of the first subframe can be restored by linear interpolation.

The L of the first subframe restored as described above
The SP and the quantized LSP of the second subframe are assigned to a linear prediction coefficient αiι ′ (i = 1,..., 10, ι = 1,
, 2), and outputs the result to the impulse response calculation circuit 310. Further, an index representing the code vector of the quantized LSP of the second subframe is output to the multiplexer 400.

The perceptual weighting circuit 230 inputs the linear prediction coefficients αi (i = 1,..., P) before quantization from the spectrum parameter calculation circuit 200 for each subframe,
Based on Document 1, perceptual weighting is performed on the audio signal of the subframe, and a perceptual weighting signal is output.

The response signal calculation circuit 240 receives the linear prediction coefficient αi for each subframe from the spectrum parameter calculation circuit 200, and quantizes, interpolates and restores the linear prediction coefficient αi from the spectrum parameter quantization circuit 210. 'Is input for each sub-frame, a response signal with the input signal set to zero d (n) = 0 is calculated for one sub-frame using the value of the filter memory held, and output to the subtractor 235. Here, the response signal χz (n) is represented by the following equation.

(Equation 2) However, when ni ≦ 0,

(Equation 3)

(Equation 4) Here, N indicates the subframe length. γ is a weight coefficient for controlling the perceptual weighting amount, and is the same value as the following equation (6). s _w (n) and p (n) denote the output signal of the weighting signal calculation circuit and the output signal of the denominator term of the filter on the right-hand side first term in equation (6) described later, respectively.

The subtractor 235 subtracts the response signal by one subframe from the auditory sensation weighting signal by the following equation, and calculates xw '
(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 number L, and the adaptive code book circuit 300 and the sound source quantization Output to the circuit 350.

(Equation 6)

In the adaptive codebook circuit 300, the past sound source signal v (n) from the weighting signal calculation circuit 360 is
The output signal x _w ′ (n) from the subtractor 235 is output from the impulse response calculation circuit 310 to the perceptual weighting impulse response h _w
Enter (n). The delay T corresponding to the pitch is determined so as to minimize the distortion of the following expression, and an index representing the delay is output to the multiplexer 400.

(Equation 7) here,

(Equation 8) And the symbol * represents a convolution operation. The gain β is obtained according to the following equation.

(Equation 9)

Here, in order to improve the accuracy of delay extraction for a female sound or a child's voice, the delay may be determined by a decimal sample value instead of an integer sample. A specific method is described in, for example, “Pitch predictors” by P. Kroon et al.
with high temporal resolution ”(Proc.
ICASSP, pp. 661-664, 1990) (Literature 10).

Further, the adaptive codebook circuit 300 performs pitch prediction in accordance with the following equation to obtain a prediction residual signal z.
_w (n) is output to the sound source quantization circuit 350.

(Equation 10)

The sound source quantization circuit 350 generates M pulses. FIG. 2 is a flowchart showing the processing flow of the sound source quantization circuit 350. Here, the pulse is decomposed into two stages,
A method of searching for a pulse position while multiplying different gains at each stage will be described. The number of stages can take any value. At this time, the sound source signal c (n) can be expressed as follows.

[Equation 11] Here, M ₁ , M ₂ , sign (k), mk, G ₁ , and G ₂ are the number of first-stage pulses, the number of second-stage pulses, the polarity of the K-th pulse, and the first-stage pulse, respectively. Eye pulse gain,
Each shows the gain of the second stage pulse. Also, M ₁ +
M ₂ = M.

In FIG. 2, in the first block, z
_w (n) and h _w (n) are input, and a first correlation function d (n) and a second correlation function φ are calculated according to the following equations.

(Equation 12)

(Equation 13)

In the next block, the positions of the number M ₁ (M ₁ ≦ M) of non-zero amplitude pulses (first pulses) are calculated using two types of correlation functions. For this purpose, as in Reference 3, an optimum pulse position is searched for a predetermined position candidate for each pulse.

For example, when the subframe length is N = 40 and the number of pulses is M ₁ = 5, an example of a candidate for the position of each pulse can be expressed as shown in the following table.

[Table 1]

For each pulse, each position candidate is checked, and the position that maximizes the following equation is selected as the optimum position.

[Equation 14] here,

(Equation 15)

(Equation 16) It is. Position of M ₁ pulses are outputted.

Next, using the calculated M ₁ pulse positions, the amplitude corrects the correlation function d (n) as polarity. It is shown in the following equation.

[Equation 17]

(Equation 18)

Next, the positions of M2 pulses (second pulses) are calculated using d ′ (n) and φ. Here, in equation (15), d ′ (n) may be used instead of d (n), and the number of pulses may be M ₂ .

The polarities and positions of M pulses in total are determined,
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. Further, the polarity of the pulse is output to the multiplexer 400.

The gain quantization circuit 365 reads out a gain code vector from the gain code book 355, selects a gain code vector for minimizing the following equation for a selected position, and finally minimizes distortion. The combination of the amplitude code vector and the gain code vector to be selected.

Here, an example in which three kinds of gains of the adaptive codebook gain and the sound source gains G1 and G2 expressed by pulses are simultaneously vector-quantized will be described.

[Equation 19] Here, β _t ′, G _1t ′, and G _2t ′ are gain codebook 3
The k-th element in the three-dimensional gain code vector stored in 55. The above equation is repeated for each of the gain code vectors, and a gain code vector that minimizes the distortion D _t is selected.

An index representing the selected gain code vector is output to the multiplexer 400.

The weighting signal calculation circuit 360 receives the respective indexes, reads out the corresponding code vectors from the indexes, and obtains the driving sound source signal v (n) based on the following equation.

(Equation 20) 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 is output to the response signal calculation circuit 240.

(Equation 21) This concludes the description of the first embodiment.

FIG. 3 is a block diagram showing the second embodiment.
Shown in Here, the operation of the sound source quantization circuit 450 is shown in FIG.
And different. The sound source quantization circuit 450 uses the amplitude codebook 451 to quantize the pulse amplitude.

After the positions are determined for M ₁ pulses, Q (Q ≧ 1) amplitude code vector candidates are output so as to maximize the following equation.

(Equation 22) here,

(Equation 23)

(Equation 24) It is. Here, g _kj ′ is the j-th amplitude code vector of the k-th pulse.

For each of the selected Q amplitude code vectors, the correlation function is modified by the following equation.

(Equation 25)

Next, for each of the Q d ′ (n),
The amplitude codebook 451 for the remaining M2 pulses
From the search for the amplitude code vector. Select the one that maximizes the following equation.

(Equation 26) here,

[Equation 27]

[Equation 28]

The above processing is repeated for Q d '(n), and a combination that maximizes the accumulated value according to the following equation is selected.

(Equation 29)

The index representing the selected amplitude code vector is output to the multiplexer 400. Further, the position value and the value of the amplitude code vector are converted to a gain quantization circuit 46.
Output to 0.

The gain quantization circuit 460 selects a gain code vector that minimizes the following equation from the gain code book 355.

[Equation 30] In this embodiment, the amplitude codebook is used.
Alternatively, the search may be performed using a polarity codebook indicating the polarity of each pulse. This concludes the description of the second embodiment.

FIG. 4 is a block diagram showing the third embodiment. Here, the first sound source quantization circuit 500
The pulse is decomposed into, for example, two stages, and the position of the pulse is searched for while changing the gain multiplied by the pulse in the first stage and the second stage. Here, the number of stages is not limited to two and can take any value. The pulse search method itself uses the same method as the sound source quantization circuit 350 in FIG. The sound source signal at this time is represented by c ₁ (n) in the following equation.

(Equation 31) After the search of pulses, it calculates the distortion D ₁ of the first sound from the following equation.

(Equation 32)

The following equation can be used instead of the above equation.

[Equation 33] Here, the values after the pulse search is completed are used for C _j , C _i , E _j , and E _i .

The second sound source quantization circuit 510 searches for pulses without decomposing the pulses while using a constant gain for all M pulses. Here, M> [M ₁ +
M ₂ ]. Further, the second sound source signal c ₂ (n) can be expressed as the following expression.

(Equation 34) Here, G is a gain applied to the entire M pulses.

Further, the distortion D ₂ due to the second sound source is calculated from the following equation.

(Equation 35)

Alternatively, it can be calculated from the following equation.

[Equation 36] Here, the values after the search for the pulse in the second sound source quantization circuit is completed are used for Cι and Eι.

The discriminating circuit 520 outputs the first sound source signal c
₁ (n), the second sound source signal c ₂ (n), and the distortions D ₁ and D ₂ due to them are input, the magnitude is discriminated, and the sound source signal with the smaller distortion is output to the gain quantization circuit, and discrimination is performed. The code is output to the gain quantization circuit 530 and the multiplexer 400. Further, a code representing the position and polarity of the pulse of the sound source signal having the smaller distortion is output to the multiplexer 400.

Gain quantization circuit 530 receives the discrimination code and performs the same operation as gain quantization circuit 365 when the first excitation signal is used. On the other hand, when the second sound source signal is used, a two-dimensional gain code vector is read from the gain code book 540, and a code vector that minimizes the following equation is searched.

(37) An index representing the selected gain code vector is output to the multiplexer 400. This is the end of the description of the third embodiment.

FIG. 5 is a block diagram showing a fourth embodiment. Here, the operations of the first sound source quantization circuit 600 and the second sound source quantization circuit 610 are different from those in FIG.

The first sound source quantization circuit 600 quantizes the pulse amplitude using the amplitude codebook 451, as in the case of the sound source quantization circuit 450.

After the positions of the M ₁ pulses are determined, Q (Q ≧ 1) amplitude code vector candidates are output so as to maximize the following equation.

(38) here,

[Equation 39]

(Equation 40) It is. Here, g _kj ′ is the j-th amplitude code vector of the k-th pulse.

For each of the selected Q amplitude code vectors, the correlation function is modified by the following equation.

[Equation 41]

Next, for each of the Q d ′ (n),
Amplitude codebook 451 for the rest of the M ₂ pulses
From the search for the amplitude code vector. Select the one that maximizes the following equation.

(Equation 42) here,

[Equation 43]

[Equation 44]

The above processing is repeated for Q d ′ (n), and a combination that maximizes the accumulated value according to the following equation is selected.

[Equation 45]

Further, the first sound source signal is obtained by the following equation.

[Equation 46]

Further, the distortion D _{1 due} to the _first sound source is calculated from the following equation, and is output to the discrimination circuit 520.

[Equation 47]

The second sound source quantization circuit 610 searches for an amplitude code vector that maximizes the following equation.

[Equation 48] here,

[Equation 49]

[Equation 50]

Further, a second sound source signal is obtained by the following equation.

(Equation 51)

Further, the distortion D ₂ due to the second sound source is calculated from the following equation, and is output to the discriminating circuit 520.

(Equation 52)

Alternatively, it may be obtained by the following equation.

(Equation 53) Here, Cι and Eι are correlation values after the second sound source pulse has been obtained.

The discrimination circuit 520 outputs the first sound source signal c ₁ ′
(N), the second sound source signal c ₂ ′ (n) and the distortions D ₁ ′ and D ₂ ′ due to them are input, the magnitude is discriminated, and the sound source signal with the smaller distortion is output to the gain quantization circuit. , And outputs the discrimination code to the gain quantization circuit 530 and the multiplexer 400.

FIG. 6 is a block diagram showing a fifth embodiment. Here, an example based on the third embodiment will be described, but it is also possible to use the fourth embodiment.

The mode discrimination circuit 900 receives the perceptual weighting signal from the perceptual weighting circuit 230 in frame units, and outputs mode discrimination information to the sound source quantization circuit 600. Here, the feature amount of the current frame is used for mode determination. As the characteristic amount, for example, a pitch prediction gain averaged in a frame is used. For example, the following equation is used to calculate the pitch prediction gain.

(Equation 54) Here, L is the number of subframes included in the frame. Pi and Ei respectively represent the voice power and the pitch prediction error power in the i-th subframe.

[Equation 55]

[Equation 56] Here, T is an optimal delay for maximizing the prediction gain.

The frame average pitch prediction gain G is compared with a plurality of predetermined thresholds, and classified into a plurality of types of modes. As the number of modes, for example, 4 can be used. The mode determination circuit 900 outputs the mode information to the sound source quantization circuit 700 and the multiplexer 400.

The sound source quantization circuit 700 receives the mode information, performs the same operation as the first sound source quantization circuit 500 in a predetermined mode, and sends the first sound source signal to the gain quantization circuit 750. Output. Further, a code representing the position and polarity of the pulse is output to the multiplexer 400.
In modes other than the above, the second sound source quantization circuit 51
The same operation as that of 0 is performed, the second excitation signal is output to the gain quantization circuit 750, and a code representing the position and polarity of the pulse is output to the multiplexer 400.

Gain quantization circuit 750 receives the mode information and performs the same operation as gain quantization circuit 365 in a predetermined mode. In other modes, the same operation as that of gain quantization circuit 530 is performed.
This concludes the description of the fifth embodiment.

The present invention is not limited to the above-described embodiment, and various modifications are possible. A code book for quantizing the amplitudes of a plurality of pulses can be learned and stored in advance using an audio signal. Codebook learning methods are, for example, Lind
e et al. “An algorithm for vector quantization
design, ”(IEEE Trans. Commun., pp.84-
95, January, 1980) (Reference 11).

Instead of the amplitude codebook, a polarity codebook in which combinations of the polarity of each pulse are prepared by the number of bits equal to the number of pulses may be provided.

The number of pulses when obtaining pulses while changing the gain can take any value.

The adaptive code book circuit and the gain code book can be switched using the mode information.

When quantizing the pulse amplitude, the amplitude code book 35 is used for each group of L pulses.
One or more code vectors may be preliminarily selected, and the pulse amplitude may be quantized using the preselected code vector. This processing can reduce the amount of calculation required for amplitude quantization.

The following is an example of the preselection method. Equation (5
9) Alternatively, a plurality of types of amplitude code vectors are preliminarily selected in the order of maximizing Expression (60) and output to the excitation quantization circuit.

[Equation 57]

[Equation 58]

[0080]

As described above, according to the present invention,
In the sound source quantization unit, the sound source is composed of M amplitude non-zero pulses, and when searching for the position of the pulse, the position of the pulse is changed while changing the gain multiplied by the pulse for each number smaller than M. By performing the search, the accuracy of the sound source can be increased more than before, and the performance can be improved.

According to the present invention, the first sound source quantization for searching for the position of the pulse while changing the gain multiplied by the pulse for each number smaller than M, and the second for searching for the pulse with the gain being constant. Source quantization of 2
By discriminating both distortions and selecting the better one, it is possible to always use a good sound source in accordance with the time change of the characteristics of the audio signal, so that the performance can be improved as compared with the related art.

Further, according to the present invention, a mode is determined by extracting a feature from an input voice, and a pulse is obtained by switching between the first sound source quantization and the second sound source quantization according to the mode. Therefore, it is possible to always use a good sound source in accordance with the time change of the feature of the audio signal with a smaller amount of calculation, so that the performance can be improved as compared with the related art.

[Brief description of the drawings]

FIG. 1 is a diagram showing a first embodiment.

FIG. 2 is a diagram showing a processing flow of a sound source quantization circuit 350.

FIG. 3 is a diagram showing a second embodiment.

FIG. 4 is a diagram showing a third embodiment.

FIG. 5 is a diagram showing a fourth embodiment.

FIG. 6 is a diagram showing a fifth embodiment.

[Explanation of symbols]

Reference Signs List 110 frame division circuit 120 subframe division circuit 200 spectrum parameter calculation circuit 210 spectrum parameter quantization circuit 220 LSP codebook 230 auditory sensation weighting circuit 235 subtraction circuit 240 response signal calculation circuit 310 impulse response calculation circuit 350, 450, 500, 600, 700 Sound source quantization circuit 451 Amplitude codebook 355 Gain codebook 360 Weighted signal calculation circuit 365, 460, 530, 750 Gain quantization circuit 400 Multiplexer 520 Discrimination circuit 89900 Mode discrimination circuit

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-2-310600 (JP, A) JP-A-5-6200 (JP, A) JP-A-60-97399 (JP, A) JP-A-3-310 53300 (JP, A) JP-A-6-202699 (JP, A) JP-A-6-222797 (JP, A) JP-A-9-146599 (JP, A) JP-A 9-281998 (JP, A) WO 95/30222 (WO, A1) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 19/00-19/14 H03M 7/30 H04B 14/04

Claims (5)

(57) [Claims] A spectrum parameter calculator for calculating and quantizing a spectrum parameter from an input speech signal;
Sound source signal of the speech signal is composed of non-zero pulse number M, while changing the gain for each small rather more number than M when searching the location of the pulse by using the spectral parameter pulse And a sound source quantization unit for searching for and outputting the position of the sound source. 2. The speech encoding apparatus according to claim 1, wherein the excitation quantization unit has a code book for quantizing the amplitudes or polarities of the plurality of pulses collectively. 3. A spectrum parameter calculator for obtaining and quantizing a spectrum parameter from an input speech signal,
The sound source signal of the audio signal is composed of non-zero pulse number M, while changing the gain for each small rather more number M1 than M when searching the location of the pulse by using the spectral parameter pulse A first sound source quantization unit that searches for and outputs the position of a second sound source, a second sound source quantization unit that searches for and outputs the position of a predetermined number of pulses using the spectrum parameters, A speech encoding apparatus comprising: calculating a distortion from the speech using an output of a sound source quantization unit and the output of the second sound source quantization unit; and selecting a sound source quantization unit having a small distortion. 4. The speech coding apparatus according to claim 3, wherein the excitation quantization unit has a code book for quantizing the amplitudes or polarities of the plurality of pulses collectively. 5. A mode discriminating circuit for obtaining a feature amount from an input audio signal, discriminating a plurality of modes from the feature amount and outputting mode information, and a first sound source quantization unit according to the mode information. The speech encoding apparatus according to claim 3, wherein switching is made between use and use of the second excitation quantization unit.