JPH0844398A - Voice encoding device - Google Patents

Abstract

PURPOSE:To provide a voice encoding system capable of obtaining satisfactory sound quality even with a low bit rate. CONSTITUTION:Delays corresponding to pitch periods of a voice signal are calculated at every sub-frame in an adaptive code book circuit 500. At this time, the delay is expressed with a difference in at least one sub-frame among plural pieces of sub-frames included in one frame. Plural kinds of bit allotting patterns are inputted from a pattern storage circuit 510 and at least one part of the number of bits to be allotted to the position and the delay of a sub-frame whose delay is expressed with the difference is determined at every frame and delays are calculated at every sub-frame and then they are outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice coding system for coding a voice signal with a low bit rate, especially at a quality of 4.8 kb / s or less.

[0002]

2. Description of the Related Art As a method for encoding a voice signal at a low bit rate of 4.8 kb / s or less, for example, "Code-excited linear prediction: High" by M. Schroeder and BA Atal. quality sp
eech at low bit rates ”(Proc.ICASSP, pp.937 ~ 940,
1985) (referred to as "Reference 1") and "Improved speech quality" by Kleijn et al.
and efficient vector quantization in SELP ”(Proc.
CELP (Code Excited L) described in a paper (referred to as "reference 2") entitled "ICASSP, pp.155-158, 1988).
PC Coding) is known.

In this system, on the transmitting side, a linear prediction (LP) is performed for each frame (for example, 20 ms) from a voice signal.
C) Using analysis, the spectral parameters representing the spectral characteristics of the speech signal are extracted, the frame is further divided into subframes (for example, 5 ms), and the parameters in the adaptive codebook based on the past excitation signal for each subframe. (Delay parameter and gain parameter) are extracted, pitch prediction of the voice signal of the subframe is performed by an adaptive codebook, and a sound source composed of a noise signal of a predetermined type for the residual signal obtained by pitch prediction The optimum source code vector is selected from the code book (vector quantization code book) and the optimum gain is calculated.

The method of selecting the sound source code vector is such that the error power between the signal synthesized by the selected noise signal 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.

The receiving-side decoding apparatus synthesizes a voice signal based on the transmission code of the code vector index, gain, spectrum parameter, etc. transmitted from the transmitting-side encoding apparatus. Note that the configuration of the decoding device is not directly related to the subject of the present invention, and therefore its description is omitted.

[0006]

In the conventional methods of the above-mentioned Documents 1 and 2, when the number of bits of the sound source codebook is reduced in order to reduce the bit rate, the sound quality of female sound is particularly deteriorated. was there.

As one of the methods for solving this problem, the number of bits of the excitation codebook is reduced as much as possible, and the delay of the adaptive codebook is expressed as a difference to reduce the number of bits for expressing the delay of the adaptive codebook. It is known how to do it.

To express the difference, the difference between the delay of the previous (previous) subframe and the delay of the current subframe is expressed by a predetermined small number of bits.

For example, when the frame length is 40 ms and the subframe length is 8 ms, the delay of the first subframe is 8
Expressed in bits, the delay in the second to fifth subframes is represented by 5 bits as a differential representation with respect to the previous subframe, and is represented by 28 bits in the entire frame.

According to this method, the number of bits is reduced by 30% as compared with the conventional method of allocating 8 bits to each subframe, which requires 40 bits per frame.

For details of the configuration in the case of differentially encoding the delay, see, for example, "Techni by Gerson et al.
ques for improving the performance of CELP type sp
eechcoders ”, IEEE J. Sel. Areas in Commun, pp.858
~ 865, 1992 (referred to as "Reference 3") and the like can be referred to.

However, in the method of Reference 3, in the stationary vowel part, the delay value has a strong temporal correlation between the subframes, so even if it is expressed as a differential expression, there is little deterioration, but the transient part of the voice, Even in the vowel part, where the pitch period of the voice changes comparatively greatly, such as in the transition part of the phoneme, it is not possible to satisfactorily express the temporal change of the pitch by using the differential expression, and the reproduced sound There is a problem in that the sound quality of the reproduced voice is deteriorated due to unclearness or the inclusion of noise.

Further, this problem becomes noticeable when the bit rate is lowered, especially for a female speaker or a speaker whose pitch changes with time.

Therefore, an object of the present invention is to solve the above-mentioned problems and to obtain a good sound quality at a low bit rate of, for example, 4.8 kb / s or less by a relatively small amount of calculation and memory. To provide.

[0015]

According to the first aspect of the present invention, the object is to input an audio signal, divide the frame into frames of a predetermined time length, and convert the audio signal of the frame to a time longer than the frame. Corresponding to the pitch period of the voice signal for each subframe, a frame division unit for dividing into a plurality of subframes that are relatively short, a spectrum parameter calculation unit for obtaining and quantizing a spectrum parameter representing the spectral characteristics of the voice signal. An adaptive codebook unit for obtaining a delay, an excitation quantization unit for quantizing an excitation signal, the spectrum parameter calculation unit, the adaptive codebook unit,
In a speech coding apparatus having a multiplexer unit for combining and outputting respective output signals of the excitation quantizer unit, a delay of the adaptive codebook unit in at least one subframe of a frame of a past subframe. The present invention is achieved by a voice encoding device characterized by determining at least one of a position of a subframe represented by a difference and a number of bits for representing a delay, for each frame, which is represented by a difference from a delay.

Further, in the second aspect of the present invention, the audio signal is input, divided into frames of a predetermined time length, and the audio signal of the frame is divided into a plurality of subframes which are shorter in time than the frame. A frame division part to divide,
A spectrum parameter calculation unit that obtains and quantizes a spectrum parameter that represents a spectral characteristic of the voice signal, and a plurality of types of modes in which the feature amount is calculated from the voice signal and the voice signal is predetermined for each frame or subframe. A mode classification unit for classifying into one, an adaptive codebook unit for obtaining a delay corresponding to a pitch period of a voice signal for each subframe, a sound source quantization unit for quantizing a sound source signal, and the spectrum parameter calculation unit. A mode classification unit, an adaptive codebook unit, and a sound source quantization unit, and a multiplexer unit that outputs the combined output signals, and the adaptive code in at least one subframe in a frame. The delay of the book part is expressed by the difference with the delay of the past subframes, and the subframe is expressed by the difference. To provide a speech coding apparatus characterized by determining in accordance with said mode at least one of the number of bits to represent the position or delay arm.

In a first aspect, the present invention is characterized by comprising a pattern accumulating section for accumulating and accumulating a pattern which predefines a position of a subframe which represents a delay by a difference and a bit number which represents the delay.

In the first aspect of the present invention, the adaptive codebook unit reads (a) a pattern defining a position of a subframe representing a delay by a difference and a bit number representing the delay from a pattern storage unit, and b) a delay search range is set based on the number of bits corresponding to the pattern in each subframe, and (c) at least one delay is searched for in the order of minimizing pitch prediction distortion in the delay search range for each subframe. d) Calculate the cumulative distortion by accumulating the pitch prediction distortion over a predetermined number (a plurality) of subframes, and after repeating the steps (a) to (d) by the type of pattern, (e) While selecting the pattern that minimizes the cumulative distortion, calculate the delay candidate in each subframe,
(f) It is characterized by including the above-mentioned steps of performing a delayed search by a closed loop search.

In a second aspect of the present invention, the mode classification unit preferably calculates, as the feature amount of the voice signal, cumulative distortion in which pitch prediction distortion is accumulated in the entire frame,
A mode is determined based on the magnitude of the cumulative distortion.

Further, in a second aspect of the present invention, the adaptive codebook section (a) based on the output signal from the mode classification section, the position of a subframe representing the delay as a difference and the bit representing the delay. Determine the pattern that defines the number,
(b) sets a delay search range based on the number of bits corresponding to the pattern in each subframe, and (c) searches at least one delay for each subframe in the order of minimizing pitch prediction distortion in the delay search range, (d) A cumulative distortion in which the pitch prediction distortion is added is calculated over a predetermined number (a plurality of) of subframes, and after the steps (b) to (d) are repeated for each pattern type, (e) ) It is characterized by including each of the above steps of selecting a pattern that minimizes the cumulative distortion, calculating a delay candidate in each subframe, and (f) performing a delay search by a closed loop search.

[0021]

The principle and operation of the speech coding apparatus according to the present invention will be described below.

In the speech coding apparatus according to the first aspect of the present invention, the speech signal is divided into frames (for example, 40 ms) and further divided into subframes (8 ms). Spectral parameters representing the spectral characteristics of speech are calculated and quantized for each frame.

The adaptive codebook unit calculates the delay corresponding to the pitch period of the voice for each subframe.

Here, an example will be described in which the position of the subframe, which represents the delay by the difference, is determined for each frame.

M types of patterns representing the number of delay bits of subframes in a frame are created in advance. For simplicity, M = 2 will be described below.

When M = 2, the pattern is, for example, (8,
5, 8, 5, 5) or (8, 5, 5, 8, 5). Where 5
It is assumed that the delay is represented by the subframe in the bit subframe, and the differential representation is not performed in the 8-bit subframe.

Therefore, in the above example, in the first pattern (8,5,8,5,5), the second, fourth, and fifth subframes are divided into the second pattern (8,5,5,8,5). ) Then, the second,
The 3 and 5 subframes are represented by the difference. It should be noted that one frame (40 ms) is composed of 5 subframes (8 ms).

In the adaptive codebook section, in preselection of delay by open loop or closed loop, first, in each subframe, L types (L ≧ 1) of delays that minimize the following expression (1) are selected.

[0029]

[Equation 1]

However, in the above equation (1), x _w (n) is the output signal of the perceptual weighting circuit 230, T is the delay, and j is the subframe number.

The selection of the delay by the closed loop in the adaptive codebook section means that the past sound source signal is filtered to generate a synthetic signal, the error power between the synthetic signal and the speech signal is calculated, and One or a plurality of delay candidates are selected in the order in which the calculated error power is the minimum. On the other hand, the delay selection by the open loop is to use the past audio signal, and the filtering process is performed in the search. It is unnecessary and the amount of calculation is reduced.

Next, in the sub-frame to be differentially encoded, the delay search range (T ₃ ≤T ≤T ₄ , where T ₃ ≤T ≤T ₄ is set based on the delay of the preceding sub-frame, but
In T ₁ ≤T ₃ <T ₄ ≤T ₂ , the delay that minimizes the above equation (1) is selected. That is, in a subframe in which the delay is not expressed differentially, for example, T ₁ = 20 and T ₂ = 147
If there are decimal point delays in increments, there are 256 types of delay search ranges, which are represented by 8 bits, and in a subframe that expresses a difference, the delay search range is T ₃ ≦ T ≦ T ₄ .

Here, the specific method of the differential expression can be referred to the above-mentioned document 3 and the like.

These processes are performed by the number of subframes in the frame, and the cumulative distortion accumulated over a plurality of subframes is calculated by the following equation (2).

[0035]

[Equation 2]

In the above equation, S is the number of subframes in which distortion is accumulated. For example, the value of S can be the number of subframes of a frame.

The above processing is repeated for the number of delay candidates obtained in the subframe, and one kind of combination of delay subframes that minimizes the cumulative distortion (the above equation (2)) is determined.

Further, the above processing is repeated for each of the two types of patterns, and the pattern with the smaller cumulative distortion is selected.

Next, in the second aspect of the present invention, the feature amount is calculated from the voice signal for each frame, and the feature amount is used to set the voice signal to one of a plurality of predetermined modes. Classify.

In the following, there are four types of modes.
Also, as the feature amount, the cumulative distortion obtained by accumulating the above equation (1), which is the open-loop pitch prediction distortion obtained for each subframe, by the equation (2) is used.

Here, the S value in the above equation (2) is 5. The value of the accumulated strain, as compared with the threshold value TH ₁ to TH ₃ of predetermined three predetermined, to determine the mode. The mode is determined, for example, as follows.

G> TH ₁ mode 0 TH ₂ <G ≦ TH ₁ mode 1 TH ₃ <G ≦ TH ₂ mode 2 G ≦ TH ₃ mode 3

That is, when the value of the cumulative distortion G is larger than the threshold value TH ₁ , the mode 0 is set, when the value is less than the threshold value TH _{1 and} larger than TH _{2, the} mode 1 is set, and when the value is less than the threshold value TH ₂ and larger than TH ₃ . When it is less than the threshold value TH ₃ , mode 2 is selected. However, the _{TH 3 <TH 2 <TH 1} .

Next, the adaptive codebook section determines the position of the subframe in which the delay is represented by a difference and the number of bits for representing the delay, according to the mode. For example, the correspondence between the modes and the number of bits is as follows.

Mode 0 (0,0,0,0,0) Mode 1 (8,5,5,8,5) Mode 2 (8,5,8,5,5) Mode 3 (8,5,5) , 5, 5)

In the mode 0, the number of bits is all 0 and the adaptive codebook is not used. Further, in the above correspondence between the mode and the bit number, the differential expression of the delay is performed in the subframe having the bit number of 5 bits, and the differential expression is not performed in the subframe of the 8-bit number.

According to the present invention, when the delay is calculated in the adaptive codebook section, the delay is represented by a difference in at least one subframe within the frame, and the position and delay of the subframe represented by the difference are represented. Since it is configured to determine at least one of the number of bits for each frame, the information to be transmitted in the adaptive codebook unit is reduced compared to the conventional method.
For this reason, not only the bit rate can be reduced, but also the reproduced sound with less deterioration can be provided in the transitional part of the sound or the like even if the delay corresponding to the pitch cycle changes with time.

Further, in the second aspect of the present invention,
The sound of a frame is classified into multiple types of modes, and the position of a subframe that expresses a difference according to the mode, or
Since it is configured to determine the number of allocated bits in the differential representation, the information to be transmitted in the adaptive codebook unit is reduced as compared with the conventional method, and not only the bit rate can be reduced but also the voice It is possible to provide a reproduced voice with little deterioration even if the delay corresponding to the pit period changes with time in the transition part of the above.

[0049]

Embodiments of the present invention will be described below with reference to the drawings.

[0050]

[Embodiment 1] FIG. 1 is a block diagram showing an embodiment of a voice encoding system according to the present invention.

In FIG. 1, a voice signal is input from the input terminal 100, the frame division circuit 110 divides the voice signal into frames (for example, 40 ms), and the subframe division circuit 12
At 0, the audio signal of the frame is divided into subframes (for example, 8 ms) shorter than the frame.

In the spectrum parameter calculation circuit 200,
A speech signal is cut out by applying a window (for example, 24 ms) longer than a subframe length to a speech signal of at least one subframe, and a spectrum parameter is calculated in a predetermined order (for example, P = 10th order). .

Since the spectral parameters change greatly in time especially in the transitional section between consonants and vowels, it is desirable to analyze every short time, but doing so increases the amount of calculation required for analysis. , Where any L (L> 1) sub-frames (eg L
= 3, the spectral parameters are calculated for the first, third, and fifth frames).

Then, in the sub-frames not analyzed (here, the second and fourth frames), the spectrum parameters of the first and third sub-frames and the third and fifth sub-frames are respectively set to LSP (line spectrum pair described later). ) Use the above linear interpolation as the spectral parameter.

Here, the known LPC analysis (linear predictive coding) or Burg is used for the calculation of the spectrum parameter.
Analysis or the like can be used.

In this case, Burg analysis is used. For details of Burg analysis, which is a spectrum estimation method based on the maximum entropy method (MEM), see pages 82 to 87 of a book entitled "Signal Analysis and System Identification" by Nakamizo (Corona Publishing Co., Ltd., 1988). 4 ")
The explanation is omitted here.

Further, in the spectrum parameter calculation section,
Linear prediction coefficient α _i (i = 1) calculated by the Burg method
10 to 10) are converted into LSP parameters suitable for quantization and interpolation. Here, the conversion from the linear prediction coefficient to LSP is
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, J64-A, pp.599-606, 1981) (referred to as "reference 5") Can be referred to. The LSP is to increase the quantization efficiency on the frequency axis by obtaining the spectrum with a line spectrum pair.

In this embodiment, the linear prediction coefficient obtained by the Burg method in the first, third and fifth subframes is LSP.
Converted to parameters, the LSP of the 2nd and 4th subframes is obtained by linear interpolation, and the LSP of the 2nd and 4th subframes is obtained.
Is inversely transformed into a linear prediction coefficient, and the linear prediction coefficient α _il (i = 1 to 10, l = 1 to 5) of the first to fifth subframes is output to the perceptual weighting circuit 230.

Also, the LSP of the first to fifth subframes 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 the LSP parameter of the fifth subframe is quantized. A well-known method can be used as the method of vector quantization of the LSP parameter.
A specific method is, for example, a series of inventions by the present inventor,
That is, JP-A-4-171500 (Japanese Patent Application No. 2-297600)
(Referred to as "reference 6"), Japanese Patent Application Laid-Open No. 4-363000 (Japanese Patent Application No.
3-261925) (referred to as "reference 7"), and JP-A-5-6199.
Publication (Japanese Patent Application No. 3-155949) (referred to as “reference 8”) or “LSP Coding Using VQ-SVQ W” by T. Nomura et al.
ith Interpolationin 4.075 kbps M-LCELP Speech Code
r ”entitled“ Proc. Mobile Multimedia Communica
tions, pp.B.2.5, 1993) (referred to as "reference 9"), etc., and therefore the explanation is omitted here.

Further, the spectrum parameter quantization circuit 21
At 0, 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 fifth LSP of the previous frame are used.
The quantized LSP of the subframe is linearly interpolated to restore the LSP of the first to fourth subframes.

Here, LSP before quantization and LSP after quantization
After selecting one type of code vector that minimizes the error power with the SP, the LSP 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 that minimize the error power, the cumulative distortion is evaluated for each candidate, and the candidate and the interpolation LSP that minimize the cumulative distortion are evaluated. Can be selected. Details are described in, for example, Japanese Patent Application No. 5-8737 (referred to as “reference 10”) filed by the present inventor.

The LSP of the first to fourth sub-frames and the quantized LSP of the fifth sub-frame restored by the above are linear prediction coefficients _α'il (i = 1 to 10, l = 1) for each sub-frame.
˜5) and output to impulse response calculation circuit 310. Also, the index representing the code vector of the quantized LSP of the fifth subframe is output to the multiplexer 400.

In the above, instead of linear interpolation, LS
The P interpolation pattern has a predetermined number of bits (for example, 2
(Bits), and the LSPs of the first to fourth subframes are restored for each of these patterns to select a pair of a code vector and an interpolation pattern that minimizes cumulative distortion. .

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 preliminarily learned and created by using SP data for training.
A predetermined pattern may be stored. The predetermined pattern is, for example, “Improved CELP speech coding at 4kb / s and b” by T. Taniguchi et al.
Paper entitled "elow" (Proc. ICSLP, pp.41-44, 1992)
The pattern described in (Reference 11) or the like can be used.

In order to further improve the performance, after selecting an interpolation pattern, an error signal between the true value of LSP and the interpolation value of LSP is obtained in a predetermined subframe, and the error signal is calculated. Further, it may be represented by an error code book. For details, reference can be made to the aforementioned reference 9.

The perceptual weighting circuit 230 inputs the linear prediction coefficient α _il (i = 1 to 10, l = 1 to 5) before quantization from the spectrum parameter calculation circuit 200 for each subframe, and the reference 10 Based on the above, the perceptual weighting signal x _w (n) is output for the sub-frame audio signal.

The response signal calculation circuit 240 inputs the linear prediction coefficient α _il for each subframe from the spectrum parameter calculation circuit 200, and the spectrum parameter quantization circuit 21
0, the quantization linear prediction coefficient alpha _'il restored by interpolating
Is input for each subframe, a response signal for the input signal d (n) = 0 is calculated for one subframe by using the value of the stored filter memory, and is output to the subtractor 250. Here, the response signal x _z (n) is expressed by the following equation (3).

[0073]

(Equation 3)

Here, γ is a weighting coefficient for controlling the perceptual weighting amount, and has the same value as γ in the following equation (5).

The subtracter 250 subtracts the response signal x _z (n) for one subframe from the perceptual weighting signal x _w (n) by the following equation (4), and x ′ _w (n) is applied to the adaptive codebook circuit. Output to 500.

X ′ _w (n) = x _w (n) −x _z (n) (4)

The impulse response calculation circuit 310 calculates the impulse response h _w (n) of the weighting filter, the transfer function of which is represented by the z-transform expression of the following equation (5), by a predetermined point L, and the adaptive code It outputs to the book circuit 500 and the excitation quantization circuit 350.

[0078]

[Equation 4]

Adaptive codebook circuit 500 determines the pitch parameter.

Here, an example will be described in which the position of the subframe in which the delay is represented by a difference and the number of bits assigned to the delay are determined for each frame. M types of patterns representing the number of delay bits of subframes in a frame are created in advance.

In the following, for simplification of the description, M = 2, and as the pattern, for example, (8, 5, 8, 5, 5) and (8, 5, 5, , 8, 5). In these patterns, the delay is differentially expressed in the 5-bit subframe and is not differentially expressed in the 8-bit subframe.

FIG. 2 shows a processing flow when the delay calculation is realized by using a microprocessor or the like.

Referring to FIG. 2, in step 501, M types of bit allocation patterns stored in advance in pattern storage circuit 510 are read.

In step 502, the delay search range is set in each subframe according to the number of bits indicated by the pattern read in step 501.

Here, the delay search range is set to T ₁ ≦ T ≦ T ₂ in a subframe that does not express a difference. As an example, if T ₁ = 20 and T ₂ = 147 and the decimal delay is ½, the search range is 256 types and can be represented by 8 bits.

Further, in the sub-frame representing the difference, the delay search range is T ₃ ≤T ≤T _4, and T ₁ ≤T ₃ <T ₄ ≤T ₂
And

Further, with respect to the delay value T _{j-1 of} the preceding subframe, T ₃ = T _j-1 -15 · Δ and T ₄ = T _j-1 + 16 · Δ are set. Where Δ is the step size of the delay,
For example, 1/2 step.

In step 503, the delay is searched for each subframe within the delay search range set for each subframe, the distortion G _j is calculated by the above equation (1), and the order of minimizing G _j is set. , L types (L ≧ 1) of delay candidates are selected.

At step 504, the cumulative distortion G _j obtained by accumulating the distortion G _{j found} in each subframe over S subframes is calculated. S can be the number of subframes included in the frame. In step 504, the above processing is repeated for the number of delay candidates L to select a delay combination that minimizes the cumulative distortion G.

Then, as shown in FIG. 2, steps 501 to 5
Repeat the process of 04 for each pattern type.

Next, in step 505, the cumulative distortion G is compared for each pattern to select the pattern that minimizes the cumulative distortion, and the delay in each subframe is output.

In step 506, the bit allocation pattern and the delay value in each subframe are received, the search range is set in each subframe, and the optimum delay is calculated by the closed loop method. Here, for the delay calculation by the closed loop method, for example, the above-mentioned Document 2 can be referred to.

Then, the index corresponding to the calculated delay value for each subframe is output to the multiplexer 400.
In addition, the index indicating the selected pattern is output to the multiplexer 400.

Further, adaptive codebook circuit 500 performs pitch prediction according to the following equation (6) and outputs adaptive codebook prediction residual signal z (n).

Z (n) = x ' _w (n) -b (n) (6)

Here, b (n) is an adaptive codebook pitch prediction signal and is given by the following equation (7).

B (n) = βv (nT) * h _w (n) (7)

Here, β and T indicate the gain and delay of the adaptive codebook, respectively. v (n) is an adaptive code vector. Operation symbol * is a convolution operation
Is represented.

Referring again to FIG. 1, the excitation quantization circuit 35
At 0, the best sound source code vector c _j (n) is selected so as to minimize the following expression (8) for all or some of the sound source code vectors stored in the sound source code book 340.

At this time, one kind of the best code vector may be selected, or two or more kinds of code vectors may be preliminarily selected and one kind may be finally selected at the time of gain quantization. Good. Here, it is assumed that two or more types of code vectors are selected.

[0101]

(Equation 5)

When the above equation (8) is applied only to some sound source code vectors, a plurality of sound source code vectors are preselected in advance and the preselected sound source code vectors are selected. The above equation (8) can also be applied.

The gain quantization circuit 380 reads the gain code vector from the gain code book 370, and for the selected excitation code vector, the excitation code vector and the gain code vector are set so as to minimize the following expression (9). Select a vector combination.

[0104]

(Equation 6)

Here, β ′ _k and γ ′ _k are the kth code vector in the two-dimensional gain codebook stored in the gain codebook 370.

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, and first, based on the following equation (10), the driving sound source signal v (n ).

V (n) = β ′ _k · v (nT) + γ ′ _k · c _j (n) (10)

Next, the spectrum parameter calculation circuit 200
Output parameter, spectrum parameter quantization circuit 21
The weighted signal s _w (n) is calculated for each subframe by the following equation (11) using the output parameter of 0, and is output to the response signal calculation circuit 240.

[0110]

(Equation 7)

[0111]

[Embodiment 2] FIG. 3 is a block diagram showing a second embodiment of the speech coding apparatus according to the present invention, and corresponds to the second viewpoint of the present invention. In FIG. 3, components denoted by the same reference numerals as those in FIG. 1 have the same functions as those in FIG. 1, and therefore description thereof will be omitted and only differences from FIG. 1 will be described below. In this embodiment, the feature amount is calculated from the audio signal for each frame,
This is used to classify into one of a plurality of predetermined modes.

Referring to FIG. 3, in mode classification circuit 600, a feature amount is extracted from the audio signal of the frame based on the output from frame division circuit 110 and classified into any of a plurality of modes for each frame. .

In the following description, there are four types of modes, and the feature quantity is the cumulative distortion G accumulated in the entire frame.
(See the above equation (2)), the cumulative strain G is determined by the method described in the section of the action, for example, by comparing it with _three predetermined thresholds TH _{1 to} TH _3. .

The mode classification circuit 600 outputs the mode information to the adaptive codebook circuit 550. The mode information is also output to the multiplexer 400.

FIG. 4 shows a processing flow of the adaptive codebook circuit 550 in this embodiment. 4, each step with the same number as in FIG. 2 performs the same process as in FIG.

Referring to FIG. 4, in step 555, the mode information is input, and the position of the subframe in which the delay is represented by a difference and the number of bits for representing the delay are determined. A specific example is as described in the section of the action.

Similar to the first embodiment, steps
In step 502, a delay search range is set in each subframe, and in step 503, the distortion G _j is calculated for each subframe by the equation (1), and L types of delay candidates (in the order of minimizing G _j ) are calculated. L ≧ 1) is selected, and in step 504, the cumulative distortion G _j obtained by accumulating the distortion G _j obtained in each subframe over S subframes is calculated. S can be the number of subframes included in the frame. Step 504
Then, the above processing is repeated for the number of delay candidates L to select a delay combination that minimizes the cumulative distortion G.

Then, the processes of steps 502 to 504 are repeated for the pattern determined according to the mode in step 555.

In step 505, a pattern that minimizes the cumulative distortion is selected, and candidates for delay in each subframe are output. In step 506, the bit allocation pattern and the delay value in each subframe are received. , In each subframe, the search range is set and the optimum delay is calculated by the closed loop method.

Although the present invention has been described with reference to the above embodiments, various modifications can be made within the scope (scope) of the present invention other than the above-described embodiments.

For example, the bit allocation pattern in the adaptive codebook circuit can be selected as desired.

Regarding the bit allocation pattern, in the above embodiment, the best pattern is selected by using the open loop search, but it may be selected by using the closed loop search.

Further, in the above embodiment, the position of the subframe in which the differential expression of the delay is expressed and the number of bits allocated to the delay are simultaneously expressed by using M kinds of bit allocation patterns. It is also possible that the position of B is represented by B ₁ bit and the number of allocated bits in the differential representation is represented by another B ₂ bit.

Furthermore, in the second aspect of the present invention,
Depending on the mode, the position of the subframe for differential expression,
The number of bits, or the number of allocated bits when expressing the difference may be changed.

Furthermore, the spectral parameter is LS
Other known parameters than P can be used.

Then, the spectrum parameter calculation circuit measures the change in RMS or the change in power between the previous subframe and the current subframe when calculating the spectrum parameter in at least one subframe in the frame, The spectrum parameter may be calculated for a plurality of subframes in which these changes are large. In this way, the spectrum parameter is always analyzed at the voice change point, and even if the number of subframes to be analyzed is reduced, it is possible to prevent performance degradation.

In the present invention, well-known methods such as vector quantization, scalar quantization, vector-scalar quantization, etc. can be used for the quantization of the spectrum parameter.

In the present invention, another well-known distance measure can be used for selecting the interpolation pattern in the spectrum parameter quantization circuit.

Further, in the above embodiment, the case where the codebook has one stage in the excitation quantization circuit has been described, but in the present invention, the codebook has a two-stage or multistage configuration in the excitation quantization circuit. You can also

Furthermore, other known scales can be used as the distance scale at the time of searching and learning the sound source codebook, or as the learning method.

For the gain codebook, a codebook having a size several times larger than the total number of transmission bits is previously learned, and a partial area of the codebook is assigned as a use area for each predetermined mode. In addition, when encoding, it is also possible to switch and use the use area according to the mode.

[0132]

As described above, according to the present invention,
When calculating the delay in the adaptive codebook unit, at least one subframe in the frame represents the delay as a difference, and at least one of the position of the subframe represented by the difference and the number of bits for representing the delay. Since it is determined for each frame, the information to be transmitted in the adaptive codebook unit is reduced as compared with the conventional method. Therefore, according to the present invention, not only is it possible to reduce the bit rate, but it is possible to provide a reproduced voice that is less deteriorated even in the transitional portion of the voice even if the delay corresponding to the pitch period changes with time. is there.

Further, in the second aspect of the present invention,
The sound of a frame is classified into multiple types of modes, and the position of a subframe that expresses a difference according to the mode, or
Since the number of allocated bits in the differential representation is determined, the information to be transmitted in the adaptive codebook can be reduced as compared with the conventional method, so that not only the bit rate can be reduced, but also the voice transient part. As a result, there is an effect that it is possible to provide a reproduced voice with little deterioration even if the delay corresponding to the pit period changes with time.

Further, according to the present invention, the adaptive codebook unit preferably includes the processing steps defined in claim 4 and has a relatively small amount of calculation and memory, so that the microcomputer or the like can be used. The present invention provides a speech coding apparatus which is suitable for mounting, realizes reduction of the amount of transmitted information, and can obtain good sound quality even at a low bit rate.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a processing flow of an adaptive codebook circuit 500.

FIG. 3 is a block diagram showing the configuration of a second embodiment (corresponding to the second aspect of the present invention) of the audio encoding device according to the present invention.

FIG. 4 is a flowchart showing a processing flow of an adaptive codebook circuit 550.

[Explanation of symbols]

 110 frame division circuit 120 sub-frame division circuit 200 spectrum parameter calculation circuit 210 spectrum parameter quantization circuit 220 LSP codebook 230 weighting circuit 240 response signal calculation circuit 250 subtraction circuit 310 impulse response calculation circuit 340 excitation codebook 350 excitation quantization circuit 360 Weighting signal calculation circuit 370 Gain codebook 380 Gain quantization circuit 400 Multiplexer 500, 550 Adaptive codebook circuit

Claims (6)

[Claims] 1. A frame division unit for inputting a voice signal, dividing the voice signal into frames of a predetermined time length, and dividing the voice signal of the frame into a plurality of subframes that are shorter in time than the frame, and the voice signal. A spectrum parameter calculation unit that finds and quantizes the spectral parameters that represent the spectral characteristics of the, an adaptive codebook unit that finds the delay corresponding to the pitch period of the speech signal for each subframe, and a source quantization that quantizes the source signal. At least one of a frame, a spectrum parameter calculation unit, an adaptive codebook unit, and a multiplexer unit that outputs combined output signals of the excitation quantization unit. In one subframe, the delay of the adaptive codebook part is compared with the delay of the past subframe. With expressed in minutes, speech coding apparatus characterized by determining at least one of the number of bits to represent the location and the delay of the sub-frame for each frame representing the difference. 2. A frame dividing section for inputting a voice signal, dividing the frame into frames of a predetermined time length, and dividing the voice signal of the frame into a plurality of subframes which are shorter in time than the frame, and the voice signal. A spectral parameter calculation unit that obtains and quantizes a spectral parameter that represents the spectral characteristics of the voice signal, and calculates one of the plurality of types of modes in which the feature amount is calculated from the voice signal and the voice signal is predetermined for each frame or subframe. A mode classification unit that classifies into a sub-frame, an adaptive codebook unit that obtains a delay corresponding to a pitch period of a speech signal for each subframe, a sound source quantization unit that quantizes a sound source signal, the spectrum parameter calculation unit, and the mode A classification unit, the adaptive codebook unit, the excitation quantization unit,
And a multiplexer unit that outputs the combined output signals of, and represents the delay of the adaptive codebook unit in at least one subframe in the frame by the difference from the delay of the past subframe and by the difference. A speech coding apparatus, characterized in that at least one of the number of bits for indicating the position or delay of a subframe is determined according to the mode. 3. A pattern accumulating section for accumulating and accumulating a pattern which predefines a position of a subframe which represents a delay by a difference and a bit number which represents the delay.
The speech encoding device described. 4. The adaptive codebook unit reads (a) a pattern defining a position of a subframe that represents a delay as a difference and a bit number that represents the delay from the pattern accumulating unit, and (b) in each subframe. A delay search range is set based on the number of bits corresponding to the pattern, (c) at least one delay is searched for in each subframe in the order of minimizing pitch prediction distortion in the delay search range, and (d) predetermined The cumulative distortion is calculated by accumulating the pitch prediction distortion over a number of sub-frames, and after repeating the steps (a) to (d) by the type of pattern, (e) a pattern that minimizes the cumulative distortion. 4. The speech code according to claim 3, further comprising: each step of selecting a delay candidate in each subframe and calculating (f) a delay search by a closed loop search. Apparatus. 5. The mode classification unit calculates, as the feature quantity of the audio signal, cumulative distortion in which pitch prediction distortion is accumulated over the entire frame, and determines a mode based on the magnitude of the cumulative distortion. The audio encoding device according to claim 2. 6. The adaptive codebook section determines (a) a pattern that defines a position of a subframe that represents a delay as a difference and a bit number that represents the delay based on an output from the mode classification section, and (b) ) In each subframe, a delay search range is set based on the number of bits corresponding to the pattern, (c) For each subframe, at least one delay is searched in the delay search range in the order of minimizing pitch prediction distortion, and (d) ) Calculate the cumulative distortion by adding the pitch prediction distortion over a predetermined number of subframes, and after repeating the steps (b) to (d) only for the type of pattern, (e) the cumulative distortion. 6. The method according to claim 4, further comprising: selecting the minimum pattern, calculating a delay candidate in each subframe, and (f) performing a delay search by a closed loop search. Speech coding apparatus.