JPH0377999B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a multi-pulse speech code decoding device. The input audio signal is analyzed, the spectral envelope information and sound source information that constitute the audio information of this input audio signal are extracted on the analysis side, and these audio information are sent to the synthesis side via the transmission path to generate the input audio signal. Vocoders that play .

The above-mentioned spectral envelope information represents the spectral distribution information of the vocal tract system that generates the input speech signal, and usually includes a number of LPC coefficients corresponding to the analysis order obtained by LPC analysis, such as α parameters and κ parameters. , LSP parameters, etc., and the sound source information indicates the fine structure of the spectral envelope, and is known as the so-called residual signal, which is obtained by removing the spectral distribution information from the input audio signal. It is also well known that information regarding intensity, pitch period, and speech/silence is included, and that this information is usually extracted via autocorrelation coefficients for each analysis frame of the input speech signal.

Now, when spectral envelope information is synthesized on the synthesis side of a vocoder, it is usually used as coefficients of an LPC synthesizer that uses an all-pole digital filter to form an approximate vocal tract system, and the sound source information is It is used as a driving sound source for this digital filter, and input audio signals are synthesized by this digital filter.

Although the conventional LPC vocoder obtained in this way is capable of synthesizing speech even at low bit rates of about 4Kb (kilobits) or less and is widely used, it has the disadvantage that high-quality speech synthesis is difficult even at high bit rates. has. The reason for this is that when modeling sound source information, for a voiced sound, the pitch period corresponding to its content is extracted and approximately represented by a single impulse train corresponding to this pitch period, while unvoiced sounds with a random period are is assumed to be a simple modeling process in which it is approximated by white noise, so it does not faithfully extract the sound source information of the input audio signal, and therefore the input audio signal contained in the sound source information This is due to the fact that analysis and synthesis of waveform information has not been carried out.

A multi-pulse audio code/decoder has recently become well known as a type of audio code/decoder that performs waveform transmission and synthesizes input audio signals in order to improve the problem caused by non-transmission of waveforms. It is something that is growing.

FIG. 1 is a block diagram showing the basic configuration of a conventional multi-pulse vocoder.

The LPC synthesizer 1 is equipped with an all-pole digital filter that simulates the vocal tract, and its coefficients are input to the input terminal.
The input audio signal x(n)(n
= 1, 2, 3...n) by the LPC analyzer 2 for each analysis frame. The sound source pulse generator 3 obtains a driving sound source sequence V(n) consisting of a plurality of impulse sequences, that is, multipulses, from the sound source information of the input audio signal,
This is supplied as a driving sound source to the LPC synthesizer 1.

The LPC synthesizer 1 inputs the LPC coefficients in this way,
This is usually the coefficient of a synthesis filter that uses an all-pole digital filter, and outputs a synthesis signal x〓(n) driven by a multi-pulse as a driving sound source. In this case, the multi-pulse includes waveform information of the input audio signal, and the LPC synthesizer 1 synthesizes the input audio signal including the waveform information.

Now, the composite signal x output from LPC synthesizer 1
(n) is then subtracted from the input audio signal x(n) by a subtracter 4 to obtain an error e(n), which is sent to the auditory weighter 5.

The perceptual weighting device 5 applies perceptual weighting to the error e(n) using a weighting filter having a characteristic W(Z) shown in the following equation (1), and
These are sent to the square error minimizer 6.

W (Z) = [1- _P 〓 ^k=1 a _k Z ^-k ] / [1- _P 〓 ^k=1 a _k γ ^k Z ^-k ] ...(1) In equation (1), a _k is LPC The LPC coefficient to be used as the coefficient of the all-pole digital filter of the synthesizer 1, p is its order and therefore the LPC analysis order, γ is the weighting coefficient, and Z is the transfer function H( Z = exp(jλ) at Z ^-1 )
, where λ=2πΔTf, ΔT is the sampling period of the analysis frame, and is the frequency.

In addition, in equation (1), the weighting coefficient γ is 0<γ<
It is set in the range of 1.

W(Z) shown in equation (1) is 1 for γ=1, γ
= 0, it changes in the range W(Z) = 1-p(Z), and the value of γ corresponds to the degree to which an excessive level appearing in the formant region in the frequency spectrum of error e(n) is suppressed. It is set within the above-mentioned range and plays the role of perceptual weighting of the signals to be synthesized, and its optimal value is usually selected in advance by an optimal auditory test.

The error e(n) weighted in this way is
In order to determine the driving sound source sequence V(n) output from the sound source pulse generator 3, that is, the optimal time position and amplitude of the multi-pulse, it is sent to the square error reducer 6, and is calculated according to the following equation (2). A squared error ε is calculated, and a driving sound source sequence V(n) is selected so as to minimize ε.

ε= _N 〓 ⁿ⁼¹ [e(n)〓w(n)] ² ...(2) In equation (2), the symbol 〓 is the convolution integral by the weighting filter of the auditory weighter 5, and N is the multipulse calculation Indicates the length of the interval.

The above-mentioned process is repeated for each multi-pulse, and synthesis by analysis is performed for each multi-pulse, which is the so-called Analysis-by-Synthesis method (hereinafter abbreviated as A-b-S method).
As is clear from the above, this A-b-S method is an effective method in the low bit rate region because it is performed in a loop of pulse generation, square error calculation, and pulse position/amplitude adjustment for each multipulse. However, the disadvantage is that the amount of calculation required is extremely large.

Regarding this A-b-S method, BS
Atal et al, “A New Model of LPC Ex−
citation for Producing Natural−Sounding
Speech at Low Bit Rates”，Proc，
It is described in ICASSP82, pp614-617 (1982), etc.

In order to address these shortcomings in the conventional A-b-S method, the following arithmetic processing algorithm has recently been introduced which efficiently calculates optimal multi-pulses based on correlation calculations.

That is, the input audio signal x(n) is divided into processing frames every N samples, and multipulses are comprehensively calculated for each frame.

Assume that there are k sound source pulses in one analysis frame, and that the i-th pulse is at the time position mi from the frame end and its amplitude is gi, then the drive sound source d(n )
is expressed by the following equation (3).

d(n) = _K 〓 ⁱ⁼¹ gi・δn, mi ...(3) In equation (3), δn, mi are Kronetzker's delta functions, and δn, mi=1 (n=mi), δn, mi =0
(n≒mi).

The LPC synthesis filter is driven by this drive sound source d(n) and outputs a synthesis signal x〓(m).

As an LPC synthesis filter, let us consider, for example, an all-pole digital filter, and its transfer function is expressed by an impulse response k(n) (0≦n≦M _-1 ), then the synthesized signal (n) is as follows. It is expressed by equation (4).

(n)= _M-1 ^〓l=0 d(1)・h(n-1)...(4) In equation (4), d(1) represents the driving sound source. Next, the weighted error, which has been audibly corrected for the error between the input audio signal x(n) and the composite signal (n), is expressed as e _w
(n), e _w (n) is expressed by the following five equations.

e _w (n)={x(n)−(n)}〓w(n)……(5
) Furthermore, the squared error can be derived from equation (5) and expressed as the following equation (6).

_M 〓 ⁿ⁼¹ e ² _w (n)= _M 〓 ⁿ⁼¹ [{x(n)−(n)}〓w(n)] ² ...(6) In equation (6), M minimizes the error The number of samples in the section to be analyzed is selected, for example, as one analysis frame length. Multipulse as the optimal sound source pulse train is
Obtained by obtaining gi that minimizes equation (6),
This gi is derived from the above-mentioned equations (3), (4), and (6) as shown in the following equation (7).

gi(mi)= _M 〓 ⁿ⁼¹ x _w (n)・h _w (n−mi) − _i=1 〓 ⁱ⁼¹ [ge _M 〓 ^M=1 hw(n−me)・hw(n−mi )]/ _M 〓 ^M=1 hw(n-mi)hw(n
−mi) ...(7) In equation (7), xw(n) is x(n)〓w(n), hw
(n) indicates h(n)〓w(n). The first term in the numerator on the right side of equation (7) is the time lag between xw(n) and hw(n).
It shows the cross-correlation function hx(mi) of mi,
Also, the second term _P 〓 ^k=1 hw(n-me)・hw(n-mi)
is the covariance function hh(me, mi) of hw(n) (1≦
me, mi≦M). Covariance function hh(me, mi)
is equal to the autocorrelation function Rhh (|me−mi|), and therefore, equation (7) can be expressed as the following equation (8).

According to equation (8), if a pulse is generated at time position mi, the amplitude gi(mi) can be determined as the optimal one. Note that in equation (8), 1≦
mi≦M.

In other words, by focusing on a certain sound source pulse and calculating its amplitude using equation (8) at various time positions, the one that maximizes the absolute value of the amplitude minimizes the squared error shown in equation (6). A plurality of sound source pulses can be obtained by repeating this procedure.

Regarding the calculation algorithm mentioned above,
Ozawa, Araseki, and Ono, ``Study of multi-pulse driven speech coding method'', detailed at the Institute of Electronics and Communication Engineers communication system study group, March 1983.

Generating multipulses based on such calculation algorithms makes it possible to calculate optimal multipulses from cross-correlation functions, autocorrelation functions, and maximum value calculations, which greatly simplifies the configuration. As a result, it is possible to realize a multi-pulse speech code decoding device that can significantly reduce the amount of calculation.

However, even the multi-pulse speech code/decoder improved in this manner still has the following drawbacks.

In other words, calculate the LPC coefficient for each frame,
Since multi-pulses are determined and synthesized, the filter coefficients of the speech synthesis filter using LPC coefficients on the synthesis side are changed on a frame-by-frame basis. As a result, the synthesis filter synthesizes an abnormal waveform that is significantly affected by the dynamic characteristics of the filter and impairs the naturalness of the voice. If the LPC coefficients are interpolated and used on the synthesis side to alleviate the influence of the abnormal waveform, the problem of the abnormal waveform can be solved, but on the other hand, the interpolation used to determine the multipulse on the analysis side The previous LPC coefficients and the interpolated LPC coefficients used on the synthesis side are different, and naturally the synthesized speech will be different from the input speech.

An object of the present invention is to eliminate the above-mentioned drawbacks, and to calculate the same coefficients on the analysis side as the interpolation LPC coefficients used on the synthesis side in a multi-pulse speech code decoding device, and further calculate the calculated interpolation LPC coefficients on the analysis side. An object of the present invention is to provide a multi-pulse type speech encoding/decoding device having a simple configuration, which greatly improves the deterioration of the S/N between an input speech signal and a synthesized speech signal by providing means for determining multi-pulses using the multi-pulse method. .

The multi-pulse speech code/decoding device of the present invention performs LPC analysis on an input speech signal for each analysis frame and uses the extracted LPC coefficients as spectral envelope information, and the sound source constitutes the speech information of the input speech signal together with this spectral envelope information. A multipulse speech code that expresses information in each analysis frame as a plurality of impulse sequences (multipulses) having generation time positions and amplitudes corresponding to the characteristics of the sound source information, and analyzes and synthesizes the input speech signal. In the decoding device, means for interpolating LPC coefficients extracted for each analysis frame of the input audio signal on an analysis side to calculate interpolated LPC coefficients;
and means for determining multi-pulses using LPC coefficients.

Next, the present invention will be explained in detail with reference to the drawings.
FIG. 2 is a block diagram showing an embodiment of the analysis side of the multi-pulse speech code/decoder according to the present invention;
FIG. 3 is a block diagram showing an embodiment of the synthesis side of the multi-pulse speech code/decoder according to the present invention.

The analysis side of the multi-pulse speech code decoding device according to the present invention shown in FIG. relational number calculator 11,
It is comprised of a sound source pulse generator 12, an encoder (2) 13, an interpolator 14, a multiplexer 15, and an impulse response calculator 16.

Audio signals input via input terminal 7001 are supplied to LPC analyzer 7 and standard digital filter 9, respectively.

The LPC analyzer 7 samples the input audio signal at 8 kHz for each analysis frame, quantizes it as a digital quantity with a preset number of bits, performs LPC analysis on this quantized audio signal, and calculates the p-order κ parameter ( (partial autocorrelation coefficient) is extracted and supplied to the encoder (1) 10 via an output line 701. In this example, the analysis frame is 20mSEC
It is set to .

The encoder (1) 10 quantizes and encodes the input LPC coefficients, and then outputs the output line 1001.
Output line 1002 to multipulse 15 via
The signals are sent to the interpolator 14 via the respective signals.

The interpolator 14 converts the quantized LPC coefficients into
5mSEC (equivalent to 4-point interpolation in this example),
Alternatively, linear interpolation is performed at 125 μS (equivalent to 160-point interpolation in this embodiment), and an interpolated LPC coefficient is calculated. The interpolator 14 further sends the interpolated LPC coefficients to the impulse response calculator 16 via an output line 1401 and to the standard digital filter 9 via an output line 1402, respectively.

The impulse response calculator 16 calculates the impulse response h(n) (0≦n≦M−1) from the interpolated LPC coefficients, and outputs it to the cross-correlation coefficient calculator 8 and the autocorrelation function calculator via output lines 1602 and 1601. 11. Note that the impulse response to be calculated is basically composed of impulse response waveform sequences whose number is equivalent to the frame period (160 types in this embodiment).

The standard digital filter 9 receives the input audio signal.
It is standardized at 8 kHz and quantized as a digital quantity with a preset number of bits, and perceptual weighting is applied to this quantized audio signal. This is implemented using standard digital filters. Although the explanation is complicated, the standard digital filter 9 is an interpolation LPC supplied via the output line 1402.
The coefficient (in this example, the κ parameter after interpolation) is α
Convert it to a parameter and use it as αk in equation (1).
Also, γ is selected to be 0.8, for example. The standard digital filter 9 outputs the perceptually weighted input audio signal to the cross-correlation coefficient calculator 8 via an output line 901.

The cross-correlation coefficient calculator 8 calculates a correlation coefficient hx using the perceptually weighted input audio signal and a plurality of (160 types in this example) impulse responses h(n), and It is sent to the sound source pulse generator 12 via an output line 801.

Further, the autocorrelation coefficient calculator 11 calculates a plurality of autocorrelation coefficients Rhh in response to each of the plurality of input impulse responses h(n), and sends these to the sound source pulse calculator 12 via an output line 1101. Send to.

The sound source pulse calculator 12 executes the calculation of equation (8) using the cross-correlation coefficient hx and the plurality of autocorrelation coefficients Rhx for each analysis frame thus inputted, and obtains a predetermined number of sound source pulse sequences. , the amplitude and position information of these pulses are sent to the encoder (2) 13 via the output line 1201, where they are quantized and encoded, and then output to the output line 1301.
The signal is sent to multiplexer 15 via.

In this way, the LPC coefficients and multipulse data that are quantized and encoded and sent to the multiplexer 15 are time-divided in a predetermined manner via the multiplexer 15 as data representing the spectral envelope and sound source information of the input audio signal. The signal is transmitted from the analysis side shown in FIG. 2 to the synthesis side shown in FIG. 3 via a transmission path 1501.

The synthesis side shown in FIG. 3 synthesizes input audio signals based on data transmitted from the synthesis side via a transmission path 1501, and includes a demultiplexer 20, decoders (1) 17, decoding (2) 18, interpolator 19, LPC synthesizer 21 and LPF (Low
Pass Filter) 22, etc.

The demultiplexer 20 restores various data input via the transmission line 1501 to the state before conversion by the time division transmission format of the multiplexer 15,
The LPC coefficient data is supplied to the decoder (1) 17 via the output line 201, and the multipulse data is supplied to the decoder (2) 18 via the output line 202. After decoding the output lines 171 and 1, respectively,
81.

The interpolator 19 interpolates the LPC coefficients with the same convention as the interpolator shown in FIG.
Output to.

The LPC synthesizer 21 uses the input multi-pulses as sound source information as the driving sound source of the p-order all-pole digital filter, and also uses the p-order interpolated LPC coefficient data input via the output line 191. This LPC synthesis filter is controlled as a coefficient of the all-pole digital filter to synthesize the input audio signal, and this is sent via the output line 211.
The signal is sent to the LPF 22, subjected to predetermined low-pass filtering, and sent to the output line 221 as an analog synthesized voice.

In the embodiment of the present invention shown in FIGS. 2 and 3, the κ parameter is used as the LPC coefficient before interpolation, but this is done before and after interpolation, and other LPC coefficients, such as LSP parameters, are used. May also be used as an encoder and multiplexer,
It is clear that the decoder and demultiplexer can also be implemented in the same way by integrating them, and the LPC synthesis filter can be replaced with a non-polar digital filter other than the all-pole type. It is also clear that it can be implemented in much the same way.

In addition, a standard digital filter 9 shown in FIG.
It is also obvious that if perceptual weighting is not performed, the A/D conversion function will remain and become unnecessary.

Furthermore, the autocorrelation coefficient calculator 11 does not necessarily require a direct impulse response waveform as input, but uses the Levinson method, which is well known as an efficient solution method for LPC analysis, rather than interpolated LPC coefficients. It is also obvious that , can be calculated directly.

As explained above, according to the present invention, the same coefficients as the interpolation LPC coefficients used on the synthesis side are calculated on the analysis side in the multi-pulse speech code decoding device, and the calculated interpolation LPC coefficients are further used. By providing means for determining multi-pulses based on the input signal, there is an effect that the S/N ratio between the input audio signal and the synthesized audio signal can be improved.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the basic configuration of a conventional multi-pulse speech code decoding device, and FIG. 2 is a block diagram showing an embodiment of the analysis side of the multi-pulse speech code decoding device according to the present invention. FIG. 3 is a block diagram showing an embodiment of the synthesis side of the multi-pulse code decoding apparatus according to the present invention. 1...LPC synthesizer, 2...LPC analyzer, 3...
... Sound source pulse generator, 4 ... Subtractor, 5 ... Auditory weighting device, 6 ... Square error minimizer, 7 ...
LPC analyzer, 8... Cross correlation coefficient calculator, 9...
...Standard digital filter, 10...Encoder
(1), 11... Autocorrelation coefficient calculator, 12... Sound source pulse generator, 13... Encoder (2), 14... Interpolator, 15... Multiplexer, 16... Impulse response calculator , 17... decoder (1), 18...
Decoder (2), 19... interpolator, 20... demultiplexer, 21... LPC synthesizer, 22... LPF.

Claims (1)

[Claims] 1. LPC coefficients extracted by LPC analysis for each analysis frame of an input audio signal are used as spectral envelope information, and together with this spectral envelope information, sound source information constituting the audio information of the input audio signal is analyzed for each analysis frame. A multi-pulse type in which the input audio signal is analyzed and synthesized by expressing it as a plurality of predetermined impulse sequences (multi-pulses) having occurrence time positions and amplitudes corresponding to the characteristics of the sound source information. In the speech decoding device, the analysis side has a first means for interpolating the LPC coefficients extracted by LPC analysis for each analysis frame to obtain interpolated LPC coefficients, and the impulse response waveform obtained from the interpolated LPC coefficients and the above-mentioned a second means for obtaining a cross-correlation coefficient sequence with the input audio signal; a third means for obtaining an auto-correlation coefficient sequence of the impulse response waveform; 1. A multi-pulse speech code decoding device, comprising a fourth means for determining pulses on the analysis side. 2. The multi-pulse speech code decoding device according to claim 1, wherein the third means is means for directly calculating the autocorrelation coefficient sequence from the interpolated LPC coefficients. 3 In claim 1, the interpolation
The input audio signal is applied to a standard digital filter whose transfer function is controlled by an LPC coefficient, and the output signal of the filter is used as the new input audio signal. .