JP3092436B2 - 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 encoding apparatus for encoding a speech signal at a low bit rate, especially at a high quality of about 8 to 4.8 kb / s.

[0002]

2. Description of the Related Art As a method of encoding an audio signal at a low bit rate of about 8 to 4.8 kb / s, for example,
M. Schroeder and B.S. Atal's "Code-excited linear pre
dictation: High quality speech
chat at low low bit rates ”
(Proc. ICASSP, pp. 937-940, 1
985) (Reference 1) and Kleijn et al.'S "Improved speech qual
city and efficiency vector
Quantization in SELP "(ICASSP, pp. 155-158, 1988).
CELP (Code) described in
Excited LPC Coding)
B. "A new model of by Atal et al.
LPC exit for produ
sing natural-sounding at
low bit rates ", (Proc. ICAS
SP, pp. 614-617, 1982) is known.

In the methods disclosed in Documents 1 and 2, the transmitting side extracts a spectrum parameter representing a spectrum characteristic of an audio signal from an audio signal for each frame (for example, 20 ms), and further divides the frame into small section subframes (for example, 5 ms). An adaptive codebook that represents a long-time correlation (pitch correlation) so as to minimize a weighted square error between a reproduced signal reproduced based on a past sound source signal and the sound source signal for each subframe. A pitch parameter is extracted, a speech signal of the subframe is long-term predicted by the pitch parameter, and a residual signal obtained by the long-term prediction is a signal selected from a codebook including a predetermined type of noise signal. In addition to selecting one type of noise signal so as to minimize the weighted square error between the signal synthesized by Calculate the gain. And the index and gain representing the type of the selected noise signal, and
The spectrum parameter and the pitch parameter are transmitted. Description on the receiving side is omitted.

[0004]

In the conventional system of the above-mentioned reference 1, when searching for a codebook composed of multi-pulses or noise signals, the number of bits of the codebook for each subframe is fixed. When obtaining multipulses, the number of multipulses in a subframe or a frame was fixed.

However, since the power of an audio signal changes greatly with time, it is difficult to satisfactorily encode a signal at a portion where the power of the signal changes with time at a fixed number of bits. There is a problem that the deterioration becomes remarkable. In addition, this problem is remarkable especially when the bit rate is reduced to reduce the size of the codebook or the number of multi-pulses is reduced.

[0006]

According to the first aspect of the present invention, an input discrete audio signal is divided into frames of a predetermined time length, and a spectrum parameter representing a spectrum envelope of the audio signal is obtained. An output spectrum parameter calculator, dividing the frame into small sections of a predetermined time length, and adjusting the pitch so that a signal reproduced based on an adaptive codebook including past sound source signals is close to the audio signal. An adaptive codebook unit for obtaining parameters;
A sound source search unit that outputs a sound source signal of the audio signal by expressing the source signal by a codebook or a multi-pulse composed of a plurality of types of code vectors configured in advance, and outputting the audio signal. A masking threshold value calculation unit for calculating a masking threshold value, and a bit allocation unit for determining the number of bits of a codebook or the number of multi-pulses in a subframe based on the threshold value. A speech coding device is obtained.

According to a second invention, in the speech coding apparatus according to the first invention, a band dividing section for dividing a band of an input speech signal, and the masking threshold value for the band-divided signal is set. A speech coding apparatus characterized by having a bit allocating unit that determines the number of bits of a codebook or the number of multi-pulses when coding a small section and the band-divided signal is obtained.

Further, according to a third aspect of the present invention, there is provided the speech encoding device according to the first invention, wherein the speech encoding device has a codebook in which the impulse response of the band division filter is convolved in advance. can get.

[0009]

In the first invention, the number of bits of the excitation codebook is allocated to an adaptive number for each subframe obtained by dividing a frame, or the number of multipulses is adaptively allocated in multipulse calculation. I do.

In the following, a case where a sound source codebook is used will be described.

[0011] A masking threshold is determined for an input voice signal based on the masking characteristics of the auditory sense, and is converted into a power spectrum T _ij on the frequency axis based on the threshold.
Here, i indicates the i-th critical band on the frequency axis, and j indicates the j-th subframe in the frame. Here, in order to obtain the masking threshold, for example, an audio signal is subjected to FFT to obtain a power spectrum | X (k) | ² , and this is analyzed by a critical band filter or an auditory model, and each critical band is analyzed. The power or RMS is calculated, and a masking threshold in each critical band is obtained from these values. As a method of obtaining a masking threshold, for example, a method using a value obtained by an auditory psychology experiment is known.
"Transform co by Johnston
dinof audio signals usin
g perceptual noise writer
ia "(IEEE J. Sel. Areas on C
ommun. Pp. 314-323, 1988) (Reference 4); Drago de Iac
Ovo et al., "Vector Quantizat
ion and perceptual criter
ia in SVD based CELP cede
rs "(ICASSP, pp. 33-3)
6, 1990) (Literature 5). For the critical band filter or critical band analysis, see, for example, "Foundation by Tobias
of modern audition theor
y "in Chapter 5 of the book (Reference 6) and the like. For the auditory model, for example, Seneff
"A computational mode
for the peripheral audio
ry system: Application to
speech recognition research
ch ”(Proc. ICASSP, pp.
1983-1986, 1986) (Reference 7).

For all subframes in the frame, T _ij (i = 1... L, j = 1.
The signal-to-masking threshold ratio SMR _ij is obtained for each subframe and band according to the following equation.

SMR _ij = P _ij / T _ij (1) where i is a critical band signal and j is a subframe number. P _ij and T _ij indicate the power and masking threshold of the input voice in the i-th critical band in the j-th subframe.
Adaptive bit allocation between subframes follows the equation below.

[0014]

(Equation 1)

Here, R _j , R, M and L are j
It indicates the number of bits allocated to the third subframe, the average number of bits of the excitation codebook, the number of critical bands, and the number of subframes in the frame.

As another method of adaptive bit allocation,
The following equation can also be used.

[0017]

(Equation 2)

In the second invention, the input signal is divided into a plurality of bands in advance. Here, Q is
MF (Quadrature Mirror Filter)
er). For details on the QMF filter,
P. "Multi by Vaidyanathan et al.
rate digital filters, filter
er banks, polyphase network
ks, and applications: A tu
trial "(Proc. IEEE, pp. 56-9)
3, 1990) can be referred to. Here, if it is divided into W bands, the bit allocation for each subframe and each band is obtained by modifying the above equation (2).

[0019]

(Equation 3)

Where R _kj is the j-th subframe,
Indicates the k-th band. However, j = 1. . . L, k =
1. . . W. SMR _kj = P _kj / T _kj (5), which is the power of the input signal for each divided band of the j-th subframe and the masking threshold value for each divided band of the j-th subframe.

Furthermore, in the third invention, there are W excitation codebooks, and the impulse response of the band division filter is pre-convolved in all code vectors of each excitation codebook. When searching for such a sound source codebook, a search is performed in the same manner as in the first invention after adding the sound source code vectors of the respective sound source codebooks according to the following equation.

[0022]

(Equation 4)

Bit allocation for each subframe and each band is as follows:
According to the formula (2), (3) or (4).

[0024]

FIG. 1 is a block diagram showing an embodiment of a speech coding apparatus according to the first invention. Here, for the sake of simplicity, in the search for the sound source codebook, an example will be described in which the number of bits of the codebook is allocated based on the masking threshold. However, the present invention can be extended to the adaptive codebook and other codebooks. it can.

In the figure, on the transmitting side, an input terminal 100
And input a voice signal from the device for one frame (for example, 20 ms)
Is stored in the buffer memory 110.

The LPC analysis circuit 130 uses the LSP as a parameter representing the spectral characteristic of the audio signal of the frame.
The parameters are calculated from the audio signal of the frame using a well-known LPC.
The analysis is performed and calculation is performed for a predetermined order L.

Next, the LSP quantization circuit 140 quantizes the LSP parameter with a predetermined number of quantization bits,
The obtained code l _k is output to the multiplexer 290, which is decoded and further subjected to linear prediction coefficients a _i ′ (i =
1 to P), and outputs the result to the impulse response calculation circuit 170 and the synthesis filter 295. The method of encoding LSP parameters and converting between LSP parameters and linear prediction coefficients is described in "Quantiz" by Sugamura et al.
er design in LSP speech a
analysis-synthesis "(IEEE J. Sel. Areas Commun.,
pp. 432-440, 1988) (Reference 9). In addition, in order to quantize the LSP parameters more efficiently, vector-scalar quantization or other well-known vector quantization methods can be used. For LSP vector-scalar quantization, see Mo.
"Transform Codin by Riya
g of Speech using a Weigh
ed Vector Quantizer, "(IEEE J. Sel. Areas, Comm.
un. Pp. 425-431, 1988) (Reference 1)
0) etc. can be referred to.

The sub-frame division circuit 150 divides the audio signal of the frame into sub-frames. Here, for example, the subframe length is 5 ms.

The masking threshold value calculation circuit 205 performs N-point FET exchange on the input audio signal x (n) to obtain a spectrum X (k) (k = 0 to N-1). X (k) | ² is obtained and analyzed by a critical band filter or an auditory model, and the power or RMS for each critical band is calculated. Here, the power is calculated according to the following equation.

[0030]

(Equation 5)

Here, bl _i and bh _i indicate the lower limit frequency and the upper limit frequency of the i-th critical band, respectively. R is the number of critical bands included in the audio signal band. Reference 6 can be referred to for the critical band.

Next, the scatter function is convolved with the critical band spectrum according to the following equation.

[0033]

(Equation 6)

Here, sprd (j, i) is a scatter function, and the specific value can be referred to the aforementioned reference 4. Also, b
_max is the number of critical bands included up to the angular frequency π. Next, a masking threshold spectrum Th _i is calculated according to the following equation.

T ′ _i = C _i T _i (9) where T _i = 10 ^{− (0i / 10)} (10) 0 _i = α (14.5 + i) + (1−α) 5.5 (11) α = Min [(NG / R), 1.0] (12)

[0036]

(Equation 7)

[0037] Here, k _i is K parameter of the i following th obtained by converting by methods known from the linear prediction coefficients input from the LPC analysis circuit 130. M is the order of the linear prediction analysis.

The masking threshold spectrum is given by the following equation by considering the absolute threshold.

T ″ _i = max [T _i , absth _i ] (14) Here, absth _i is an absolute threshold value in the critical band i, and can be referred to the aforementioned reference 5.

Next, the masking threshold spectrum T ·
For i (i = 1... b _max ), the power spectrum P _m (f) obtained by converting the frequency axis from the Bark axis to the Hertz axis is obtained, and these are subjected to inverse FFT to obtain an autocorrelation function r (j ) (J = 0 ... N-1). Next, a filter coefficient b _i (i = 1... P) is calculated by performing well-known linear prediction analysis on the autocorrelation function.

The perceptual weighting circuit 220 performs perceptual weighting through a filter having a transfer characteristic which is determined by using the filter coefficient b _i (14) wherein the weighting signal x
get _wm (n).

[0042]

(Equation 8)

Here, γ ₁ and γ ₂ are constants for controlling the weighting amounts, and are usually selected to be 0 <γ ₂ <γ ₁ <1.

In the impulse response calculation circuit 170, (1
Impulse response h _{wm of the} filter having the transfer characteristic of equation 5)
(N) is obtained up to a predetermined length and output.

A _w (z) = H _wm (z) · [1 / A (z)] (15)

[0046]

(Equation 9)

A _i ′ is the LSP quantization circuit 140
Output from

The subtractor 190 subtracts the output of the synthesis filter 295 from the weighted signal and outputs the result.

The adaptive codebook 210 receives the weighted impulse response h from the impulse response calculation circuit 170.
_w (n), a weighted signal is input from the subtractor 190, pitch prediction is performed based on long-term correlation, and delay M and gain β are calculated as pitch parameters. In the following description, the prediction order of the adaptive codebook is assumed to be 1, but may be higher than or equal to second order. The calculation of the delay M in the adaptive codebook can be performed by referring to Documents 1 and 2 described above. Further, a gain β is obtained, an adaptive code vector x _z (n) is obtained by the following equation, and the adaptive code vector x _z (n) is subtracted from the output of 190.

X _z (n) = x _wm (n) −β · v (n−M) * h _wm (n) (17) where x _wm (n) is the output signal of the subtractor 190, v
(N) is output from the past synthesis filter drive signal, and h _wm (n) is output from the impulse response calculation circuit 170. The symbol * represents convolution.

The bit allocation circuit 215 inputs the masking threshold spectrum T _i, T ′ _i, or T ″ _i for each subframe from the masking threshold calculation circuit 205, and obtains the equation (2) or (3). Bit allocation is performed according to the following equation, provided that the number of bits allocated to the subframe does not exceed the lower limit bit number and the upper limit bit number so that the number of bits in the entire frame becomes a predetermined value as in the following equation. Next, the number of bits is adjusted.

[0052]

(Equation 10)

R_min<R_j<R_max  Where R_j, R_T, R_min, R_maxIs the j-th
The number of bits allocated to the first subframe, the total
Total number of bits, lower limit number of bits of subframe, subframe
Indicates the upper limit bit number of the system. L is the support within the frame.
This is the number of subframes. As a result of the above processing,
This information is output to the multiplexer 290.

The excitation codebook search circuit 230 has codebooks (250 ₁ to 250 _N ) having different numbers of bits, and inputs the allocated number of bits for each subframe.
Codebook 250 ₁ to 250 _N depending on the number of bits
Switch. Then, a sound source code vector is selected so as to minimize the following expression.

[0055]

[Equation 11]

Where γ _k is the code vector c
_k (n) (j = 0 ... 2 ^B -1; B is the number of bits of the sound source codebook). h _wm (n)
Is an impulse response obtained by the impulse response calculation circuit 170.

The sound source code book may be composed of Gaussian random numbers as in Document 1, for example, or may be constructed by learning in advance. A method of constructing a codebook by learning is described in, for example, “An Algor
ism for Vector Quantizat
The paper entitled “Ion Design” (IEEE Tr
ans. COM-28, pp. 84-95, 1980
Year) (Reference 11).

The gain codebook search circuit 260 searches for and outputs a gain codevector using the selected sound source codevector and the gain codebook 270 so as to minimize the following equation.

[0059]

(Equation 12)

Here, g _1k and g _2k are k-th two-dimensional gain code vectors. Index of selected adaptive code vector, index of sound source code vector,
Outputs the gain code vector index.

The multiplexer 290 combines and outputs the output of the LSP quantization circuit 140, the output of the bit allocation circuit 215, and the output of the gain codebook search circuit 260.

The synthesis filter circuit 295 obtains a weighted reproduction signal using the output of the gain codebook search circuit 260 and outputs it to the subtractor 190.

FIG. 2 is a block diagram showing an embodiment of the speech encoding apparatus according to the second invention. In the figure, FIG.
The components denoted by the same reference numerals perform the same operations as those in FIG. 1, and a description thereof will not be repeated.

The band dividing circuit 300 divides the audio signal into a predetermined number (for example, W) of bands.
The bandwidth of each band is set in advance. A QMF filter bank is used for band division. For the method of configuring the QMF filter bank, reference can be made to the aforementioned reference 8.

The masking threshold value calculation circuit 410 calculates the masking threshold value of each critical band, similarly to the masking threshold value calculation circuit 205 of FIG. Then, using the masking threshold included in each band divided by the band division circuit 300, the SMR _kj is obtained according to the equation (5), and is output to the bit allocation circuit 420. Further, from the masking threshold included in each band, the filter coefficient b _i calculated in the same manner as the masking threshold calculating circuit 205 in FIG. 1, and outputs to the audio coding circuit 400 ₁ -400 _W.

The bit allocation circuit 420 receives the SMR _kj (j =
1. . . L, k = 1. . . Using W), the number of bits is assigned to each subframe and each band in accordance with the above equation (4), and output to the speech coding circuits 400 ₁ -400 _W.

FIG. 3 shows a speech encoding circuit 400 ₁ -400 _W
FIG. 3 is a block diagram showing the configuration of FIG. Since 400 _{1 to} 400 _W all perform the same operation, FIG. 3 shows an example of a speech encoding circuit 400 _l for the l-th band. In the figure, components having the same reference numerals as those in FIG. 1 perform the same operations as those in FIG.

[0068] Perceptual weighting circuit 220 inputs the filter coefficients b _i for perceptual weighting, performs the same operation as perceptual weighting circuit 220 of FIG.

The excitation codebook search circuit 230 receives the bit allocation value R _kj for each subframe and each band, and switches the number of bits of the _excitation codebook.

FIG. 4 is a block diagram showing an embodiment of the third invention. In the figure, components having the same reference numerals as those in FIG. 1 or FIG. 2 perform the same operations as those in FIG. 1 or FIG.

The excitation codebook search circuit 230 receives the bit allocation value for each subframe and each band from the bit allocation circuit 420, and switches the excitation codebook for each band and each subframe according to the bit allocation value. Each band has N types of codebooks having different numbers of bits. For example, in band 1, codebooks 500 ₁₁ to 50
Up to 0 _1N . Similarly, the band W is set to 50 _W1 to 50 _W1.
It has up to 0 _WN . Further, for each band, the impulse response of the corresponding band division filter is convolved in advance in all the code vectors in the codebook.
For example, in the band 1, the impulse response of the band division filter corresponding to the band 1 is calculated in advance by the above-mentioned reference 8, and this is previously convolved with all the code vectors of the N codebooks of all the bands 1. is there.

For each subframe, a bit allocation value for each band is input, a code book corresponding to the number of bits is read out, and the code vectors for the entire band (here, W bits) are obtained according to the above equation (6). Addition and new code vector c (n)
Is created, and a code vector that minimizes the expression (18) is selected. At this time, if a search is performed for the total combination of the codebooks in each band in all the bands, an enormous amount of calculation is required. Therefore, the adaptive codebook output signal is divided into bands, and the corresponding code A plurality of code vectors with small distortion are selected from the book, and for each of the total combinations between the candidates in all the bands, the code vectors of all the bands are restored using the equation (6).
A code vector that minimizes distortion may be selected. In this way, the amount of calculation required for searching the codebook can be significantly reduced.

In the above embodiment, the bit allocation is determined by clustering SMRs in advance and designing a bit allocation codebook in which the SMR of each cluster and the number of allocated bits are tabulated for a predetermined number of bits (for example, B bits). In addition, this can be used when calculating bit allocation in the bit and allocation circuit. With such a configuration, the bit allocation information to be transmitted may be B bits per frame, so that the transmission information for bit allocation can be reduced.

In the second and third aspects of the present invention, the following equation can be used for the bit allocation for each subframe and each band other than the equation (4).

[0075]

(Equation 13)

Here, Q _k is the number of critical bands included in the k-th divided band.

In the above embodiment, an example has been described in which the number of bits of the sound source codebook is adaptively allocated. However, not only the sound source codebook but also any one of the LSP codebook, the adaptive codebook, and the gain codebook The present invention is applicable to assignment.

As a bit allocation method in the bit allocation circuits 215 and 420, a bit number is once allocated according to the equation (2) or (3), and then a sound source codebook based on the actually allocated bit number is used. It is also possible to perform quantization, measure the quantization noise and adjust the bit allocation to maximize the following equation.

[0079]

[Equation 14]

Here, σ _nj ² is the quantization noise measured in the j-th subframe.

Further, as a method of calculating the masking threshold spectrum, other known methods can be used.

The masking threshold value calculating circuit 20
In 5, 410, in order to reduce the amount of calculation, a band division filter group can be used instead of the Fourier transform.

The auditory weighting circuit 220 can also perform conventional auditory weighting as described in Documents 1 and 2.
The transfer characteristic of the weighting filter at this time is given by the following equation.

[0084]

(Equation 15)

Here, a _i is a linear prediction coefficient obtained by the LPC analysis circuit 130.

[0086]

As described above, according to the present invention, the auditory masking threshold is obtained from the audio signal, and based on this, the number of bits of the codebook or the number of multi-pulses in a small section of the audio signal is determined. Is assigned, so that there is an effect that a good sound quality can always be maintained aurally by following the temporal change of the characteristic of the audio signal. Furthermore, it is possible to allocate the number of bits or the number of multi-pulses in the codebook in consideration of not only changes in the characteristics in the time direction but also changes in the characteristics in the frequency direction. There is an effect that encoded speech can be obtained.

[Brief description of the drawings]

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

FIG. 2 is a block diagram showing an embodiment of the second invention.

FIG. 3 is a block diagram illustrating a configuration of a speech encoding circuit in FIG. 2;

FIG. 4 is a block diagram showing an embodiment of the third invention.

[Explanation of symbols]

 Reference Signs List 110 buffer memory 130 LPC analysis circuit 140 LSP quantization circuit 150 subframe division circuit 170 impulse response circuit 190 subtractor 205, 410 masking threshold calculation circuit 210 adaptive codebook 215, 420 bit allocation circuit 220 auditory weighting circuit 230, 530 Sound source codebook search circuit 250, 50 codebook 260 Gain codebook search circuit 270 Gain codebook 290, 430 Multiplexer 295 Synthesis filter 300 Band division circuit 400 Speech coding circuit

Continuation of front page (56) References JP-A-2-294700 (JP, A) JP-A-4-302532 (JP, A) JP-A-5-232998 (JP, A) JP-A-1-207799 (JP) , A) Drago et. al. "Vector or Quantization And Criteria In SVD Based CELP Coders", ICASP-90, Vol. 1, pp33-36 (1990) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 11/00-21/06 H03M 7/30 H04B 14/04 JICST file (JOIS)

Claims (3)

(57) [Claims] 1. A spectrum parameter calculator for dividing an input discrete voice signal into frames of a predetermined time length, obtaining a spectrum parameter representing a spectrum envelope of the voice signal, and outputting the spectrum parameter. An adaptive codebook section that divides the audio signal into small sections of a predetermined time length and calculates a pitch parameter so that a signal reproduced based on an adaptive codebook including a past sound source signal is close to the audio signal; And a sound source search unit that outputs the sound source signal in a form of a codebook or a multi-pulse composed of a plurality of types of code vectors pre-configured, and outputs the result. A masking threshold value calculation unit for obtaining a threshold value; Speech encoding apparatus; and a bit allocation unit for determining the number of bits or the number of multi-pulses Dobukku. 2. A band splitting unit for splitting an input audio signal into bands, and a code for encoding a small section and the band-divided signal of the band-divided signal based on the masking threshold. 2. The speech coding apparatus according to claim 1, further comprising a bit allocating unit that determines the number of bits of the book or the number of multi-pulses. 3. The speech coding apparatus according to claim 1, wherein the codebook includes a codebook in which an impulse response of a band division filter is convolved in advance.