JPS6041100A - Multipulse type vocoder - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a multi-pulse vocoder. 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 is usually L
The number of L corresponding to the analysis order obtained by PC analysis
The sound source information is expressed by PC coefficients, such as the α parameter, and other parameters, 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 sound signal. It is also well known that it contains information about the source strength, pitch period and voiced/unvoiced of the return signal, and that this information is usually extracted via the autocorrelation coefficient for each analysis frame of the input audio 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 7 for this digital filter, and input audio signals are synthesized by this digital filter.

The conventional LPC vocoder thus obtained is approximately 4
Although it is possible to synthesize speech even at a low bit rate of Kb (kilobits) or less, and it is often used, it has the disadvantage that pipe-quality speech synthesis is difficult even at a high pit rate. The reason for this is that when modeling sound source information, the pitch period corresponding to the pre-voiced content is extracted and approximately expressed by a single impulse train corresponding to this pitch period, and the unvoiced sound with a random period is In the seventh step, which assumes a simple modeling process, it is approximated by white noise, and it is not possible to faithfully extract the sharpness and false information of the input audio signal. f! r, 'light't
? Analysis of waveform and Wei Chen of input back door Ig included in IJ 2
This is because the synthesis has not been performed.

The multi-pulse vocoder has recently become well known as one of the vocoders that can effectively synthesize 1hC input voice signals while transmitting waveforms in order to improve the problem caused by non-transmission of A2 and I signals. 1 and 9.

Figure 1 shows the conventional multi-pulse 1 (vocoder portion Jji
2 is a block diagram showing the basic configuration.

I, l) The C synthesizer l) is equipped with a full-frame digital filter, and its coefficients are inputted via the input 91 connector 2001.
(n) (n=1.2.3...n) L P C
l, i' analyzed for each analysis frame using force analysis valley 2
The number of C yarns is supplied t3.

The sound source pulse generator 3 is a heka sound-fi, = Hakugo draft source 1
, 1 report.

A drive-silence sequence V (r+) consisting of a plurality of impulse sequences, 7jゎchi multipulses, is supplied at 14 as a driving sound source of the total LPC synthesizer 1.

The LPC synthesizer 1 uses the input LPC coefficients as coefficients of a synthesis filter that normally uses an all-pole digital filter, and generates a synthesized signal driven by a multi-pulse as a driving sound source. Output (n). In this case, the multi-pulse contains the waveform information of the input audio signal 7), and the Ll)C synthesizer 1 synthesizes the input audio No. 40 including the waveform information.

Now, the synthesized signal ma(n) output from the LPG synthesizer 1 is then subtracted by the subtracter 4 to take the difference from the human voice signal x(n),
The error e(n) is obtained and sent to the perceptual weighting device 5.

The auditory weighting unit 5 calculates the following (1
) The weighting having the characteristic w(z) shown in equation (2) is performed by applying perceptual weighting using a filter and then sending these weights to the 2' power error minimizer 6.

・・・・・・・・・・・・・・・ Q) In equation (1), ak is the LPC coefficient that should be the coefficient of the all-pole digital filter of LPC synthesizer 1, p is its order, and therefore LP
C analysis θζ number, rti constant is the coefficient, 2 is the transfer function H (z = exp (jλ) in z-representation of the all-pole digital filter, where λ = 2πΔT/
and ΔT is the sampling period of the analysis frame,
f indicates the frequency.

In addition, in equation (1), the weighting factor Oil r is set within the range J'1. Ru.

w (z) shown in equation (1) is r □ = l f (: whereas for f41.r-〇, w (z) = 1-1) (
The value of r varies in the range of fii2uil in the range of error e (n), and the value of r corresponds to the degree to which the oversignal level appearing in the normalant region in the frequency spectrum of the error e (n) is suppressed. 48's auditory 1 cage to be set and synthesized, (= 1 sea 1's performance).

It is usually the best J! The optimum value is selected by the auditory test.

The error e weighted in this way (+ is the driving sound source series V (n) output from the sound source pulse generator 3
In other words, in order to determine the optimal time position and amplitude of the multi-pulse, the following (2) is applied to the squared error minimizer 6.
The squared error ε is calculated according to the formula, and the drive sound source sequence V (n) is selected so as to minimize ε. 2 In Equation (2), the symbol -X- indicates the auditory weighting is the convolution integral by the filter, and N indicates the interval length for calculating the multipulse.

The above process is repeated for each pulse of the multi-pulse,
Synthesis by analysis is carried out for each
So-called Analysis -by -8yntl〕e
sis method (hereinafter referred to as A-1)-8+ method), in this A, -b-8 method, as is clear from the contents of the double series, the pulses for each multi-pulse are generation, error meter q, and pulse position/amplitude g
1.1 Low bitrate i because it is done in his loop
Although it is an effective method in the region, its story 1
The disadvantage is that the amount of data becomes extremely large.

Regarding this A-b-8 method, see 13. S.

Atal et al ANewModel of L
PC Excit-ation for Produci
ng Natural -8oundingSpeec
h at Low Bit Aates”, Pro
c, ICASSP82, pp 614-617. (1
982)"; 萱C14 consecutive A.

The disadvantage of such conventional A-b=8 method (1(
On the other hand, the following arithmetic processing algorithm has recently been introduced that efficiently calculates optimal multi-pulses based on correlation calculations.

In other words, the human voice signal x (n) is divided by Ns/Glugoto processing functions, and the multipulses are comprehensively summed for each claim.

Now, it is assumed that there are 4·n pulses in the 1-minute V frame, and the iJ-th pulse is at the time position m, from the root.
, and its amplitude is g, then the LPC synthesis filter's drive 1! I sound source +1 (n) is the next (3)
It is shown by the formula.

d(n)=Σgl・δn,In, ・・・・・・・・・
...(3) I: +I11 In equation (3), δn, InI are Kronesoker's delta functions, and δ(r, m, = 1 (n = ml)
, δn , In 1 = 0(n'=,m,).

LPG91 Yamifiruri V Yoko no Kakupuukuriki sound source d (n
) and outputs 1, R number 7C (rn).

As an LPG synthesis filter, for example, all hfi iv!
Assume that the digital filter can be omitted, and its transfer function is defined as the impulse response k (,1)' (0≦n≦1!v1.
−1) k, the composite signal '5r
(n) is expressed by the following equation (4).

t-.

In equation (4), d (t)'',! ii17 Represents a moving sound source. Next, for the error between human voice h44 bow>: (, 1) and the synthesized name (n), listen 11 & target: Ni 111 correct?
The applied weighting is eW(
n) is the following (5):A. It is indicated by.

Cw(n) = ( x (n) − x (n) l
-X-w (n) (5) The squared ratio can be expressed by the following equation (6) by deriving from equation (5).

・・・・・・・・・・・・(6) In equation (6), Mid error kAJ is reduced.
The number of samples is 1 minute t11゛frame long (y) ζ, f. .

The multi-pulse as the optimal sound source pulse train minimizes equation (6) g. 'l: is/!・Did you get caught? -〇gl is obtained from the above-mentioned equations (3), (4), and (6) as shown in the following equation (7).

1-1 ・・・・・・・・・・・・(7) In equation (7), xw(n) is x (n) −X−w
(nhah7(n) is h (n) %w (n) 's:
show. The first term in the numerator on the right side of equation (7) is the cross-correlation function of the time delay 111I between x-(n) and hW(n).
hx(mθ, and the covariance function of M is 1'hh(rnztml) (1≦111
1+171+4M). Covariance function mbh (
mtt "θ is the autocorrelation function 几hh (1mz ml)
Therefore, the following equation (7) can be expressed as the following equation (8).

・・・・・・・・・・・・(8) According to equation (8), if a pulse is generated at a time of about 1 m, the amplitude g I (m + ) will be determined to be optimal. becomes. Note that in equation (8), 1=
m, ≦M.

In other words, looking at 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). By repeating this procedure, a plurality of sound source pulses can be obtained.

Regarding the above-mentioned calculation algorithm, please refer to Ozawa, Hikai, and Ono's study of multi-pulse driven speech coding method.
It was detailed in the Communication Method Study Group of the Institute of Electronics and Communication Engineers, 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. Therefore, it is possible to realize a multi-pulse vocoder that can significantly reduce the amount of calculation.

However, even the multi-pulse vocoder improved in this way still has the following drawbacks.

That is, according to Ozawa et al.'s algorithm, the time position and amplitude of the multipulse are determined by the following procedure. First, calculate 0hx (mI). The waveform in Fig. 2 (A) is the actual measured value of φ), , (mθ) of the voice uttered by a certain male speaker. The position of the first pulse constituting the multi-pulse is determined as the position (m, = 72) where the absolute value of the waveform in Fig. 2 (A) is maximum, and the amplitude of the pulse is m, =
Da at 72 (value of mρ < 12 + (72) = -5
, 3).

Next, remove the influence of the first pulse from l (mθ. This operation is done from the waveform of FIG.
This means that the waveform in FIG. 3 is subtracted by multiplying it by (-5, 3) with 72 as the center. The waveform in Figure 2(B) is the same as that in Figure 2(A).
) shows the result of removing the influence of the first pulse from the waveform. The position and amplitude of the second pulse are determined in a manner similar to the method used to determine the position and amplitude of the first pulse for the waveform of FIG. 2(B). Next, the influence of the second pulse is removed from the waveform of FIG. 2(B).

(The results are shown in Figure 2 (C)), repeat the above operations, and then proceed to step 3. 4th.・・・・・・・・・No. t・・・・・・・・・
The position and amplitude of the second pulse are determined. As mentioned above, Ozawa et al.'s algorithm searches for Inl with the maximum absolute value in each waveform in the $2 diagram, and then searches for Inl with the maximum absolute value in m, or the value of G after removing the influence of the pulse, 121 (mI).
Furthermore, the above rnl and g (ml) are determined as the position and amplitude of the pulse. However, the shape of l (mI) near m1 is not necessarily similar to the shape of hh. For example, m in Figure 2 (F) = 159
The shape of the nearby waveform is significantly different from that shown in FIG. As a result, the waveform in Figure 2 (G) is (m, - compared to one).
Increasing the value of l near 163 in FIG. 2 (I) is a contributing factor to the pulse reaching m = 163.

As mentioned above, Ozawa et al.'s algorithm is 9! $(In,
) or the influence of the pulse is removed (1 = (ml where the absolute value of mθ is maximum and the corresponding φ(mI)'t are determined as the temporal position and amplitude of the pulse, respectively, so especially φ
When the shapes of (m I) and 几, h are significantly different, ψ
(m + ) does not necessarily decrease optimally, and the number of pulses increases unnecessarily, resulting in a decrease in encoding efficiency.

The object of the present invention is to eliminate the above-mentioned drawbacks and provide a multi-pulse vocoder with X, , (11) (i.e.
Convolution x (
n) *w (n)), (n) (i.e. I,
The cross-correlation coefficient φ (m?) between the PC synthesizer 1 and the perceptual weighting device 5 (mutual impulse response) and the hW (
n) By determining the position and amplitude of the pulse in consideration of the autocorrelation coefficient and the similarity with the multi-pulse, the shortcomings of Ozawa et al.'s algorithm can be completely removed and the efficiency of multi-pulse encoding can be improved. Its purpose is to provide a type vocoder.

The multi-pulse vocoder of the present invention performs LPC analysis on an input audio signal for each analysis frame, extracts the first LPC coefficient, and uses the spectral envelope information as spectral envelope information together with the sound source information constituting the audio information of the input audio signal. In a multi-pulse vocoder that analyzes and synthesizes the input audio signal by representing each analysis frame with a plurality of impulse sequences (multi-pulses) having generation time positions and amplitudes corresponding to the characteristics of the preceding source information. , means for calculating a cross-correlation coefficient sequence between the input speech signal and the impulse response of the speech synthesis filter; means for calculating the entire autocorrelation coefficient sequence of the impulse response; The analysis side includes a means for calculating the degree of similarity, and a means for searching for the maximum value of the degree of similarity and calculating the amplitude and position of the impulse sequence (multipulse) in a forward manner. Ru.

Next, the present invention will be described in detail with reference to the drawings. Figures 3 and 4 are block diagrams showing one embodiment of the analysis side of the multi-pulse type vocoder according to the present invention, and supplementary 51 is a block diagram showing an embodiment of the analysis side of the multi-pulse type vocoder according to the present invention. FIG. 2 is a block diagram showing an example of a synthesis side of a vocoder.

Analysis side of a multi-pulse vocoder according to the present invention shown in FIG. 4, LPC analyzer 7. Mutual) 11] Bar] %V゛ζ Calculator 8. Encoder (1) 9. The autocorrelation 13 is composed of a J0 calculator 10, a similarity anti-calculator 11, an encoder (2) 12, and a multiplexer 13.

Input audio No. 45 input via the input terminal 7001 is supplied to the LPC analyzer 7 and the cross-correlation function calculator 8.

The LPC analyzer 7 analyzes the input audio signal for each analysis frame.
Quantize it as a digital quantity with a preset number of bits,
This digitalized audio signal is analyzed by PC to obtain the following 1) parameter (partial autocorrelation coefficient) as an LPC coefficient.
The encoder (1
) 9. In this example, there are two analysis frames.
It is set to 0rnSEC.

The encoder (1) 9 quantizes and encodes the input fcLPC coefficients, and then sends them to the multiplexer 13 via an output line 901.

The LPC analyzer 7 also calculates an impulse response h(+ri (0≦n≦ta−1)) from the LPC coefficients and transmits the impulse response h(+ri (0≦n≦ta−1) The signal is supplied to the first turn function calculator 8 and the autocorrelation function calculator 10.

The cross-correlation function lti calculator 8 calculates a cross-correlation function hx using the input audio signal and the impulse response h(n), and sends it to the p-similarity calculator 11 via the output line 801. .

Further, the autocorrelation function calculator lO calculates the autocorrelation interval 'e'''hl+ of the input impulse response h(n) by dl,
9, analogous to this via output line 1001, )(t
Send to HF output 11.

The similarity calculator 11 calculates the cross-correlation function DA, X, and self-phase 1 for each analysis frame input as E2. I! :l function 1(
,1,.

Using this method, we obtain a precursor pulse train with a similarity factor (it Atf;
It is sent to the encoder (2) 12 via the encoder (2) 101, and after being converted into a W child and encoded by kneading, it is output to the output line 12.
01? r to the multiplexer 13.

In this way, the LPC movable and multipulse data signals that are quantized and converted into signals and sent to the multiplexer 13 are passed through the MA-multiplexer 13 as data representing the spectral envelope and sound source information of the input audio signal. , are time-divided in a predetermined manner and transmitted via a transmission path 1301 from the analysis side shown in FIG. 2 to the synthesis side shown in FIG.

The synthesis side shown in FIG. 5 synthesizes input audio signals based on data transmitted from the analysis side via a transmission path 1301, and includes demultiplexers 14. Decoder (
1)15. Decoder (2) 16° LPC synthesizer 17 and L P F (Low Pa5sFiltcr) 18
Constructed with all the necessary features.

The demultiplexer 14 restores various data input via the transmission path 1301 to the state before conversion by the time division transmission format of the multiplexer 13, and the LPC coefficient data is sent to the decoder (1) 15 via the output line 141. , the multi-pulse data is sent to the decoder (2) via output line 142.
16 respectively, the data is decoded by these decoders, and the output line 1
51,161.

The LPG synthesizer 17 uses the input multi-pulses in this way as source information for the driving sound source of the p-order all-pole digital filter, and also uses the p-order LPC coefficient data input via the output 2-in 151. is used as the coefficient of the all-pole digital filter to control this LPC synthesis filter to synthesize the input audio signal, send it to the LPF 18 via the output line 211, perform predetermined low-pass filtering, and synthesize the analog signal. It is sent to the output line 181 as audio.

Next, the similarity calculator 11 will be explained in detail with reference to the drawings. FIG. 6 is a block diagram showing one embodiment of the similarity calculator 11.

The cross-correlation function φhx is stored in the cross-correlation coefficient memory 19 via the transmission line 801. The autocorrelation function Rhh is supplied to the autocorrelation normalizer 20 via the transmission line 3.001f. The autocorrelation normalizer 20 calculates the normalization coefficient a corresponding to the power when the R, .
Calculated using the formula.

NB a −Rhh <o) + 2 Σ 几hh Cs)
``...''''・ (9) s=Madashi hh hh (X) indicates the 几hh component of the delay X.

Also, Nntj: the above-mentioned impulse response hW(n
) indicates the practical duration of Furthermore, autocorrelation normalizer 2
゜ normalizes each element of 几bh (X) using a, and outputs the result as a normalized autocorrelation coefficient R'hh to the autocorrelation coefficient memory 21 via the transmission line 201i. Product sum calculator 2
2 is the cross-correlation function 'h supplied via the transmission line 191
The delay m of x, nl for all centers, (the elements for the back NB,
The sum of products, .

R b,,,l=Σ 'hx (m, 10s)・"'hh
(S) 8 Tomi-NB (10) The sum-of-products calculator 22 calculates the interval (240 in this embodiment) in which the cross-correlation function f'hx is defined, that is, ml = 1”'fl-
240, and calculate bm one after another and send the results to the transmission line 22.
1 to the maximum value searcher 23. The maximum value searcher 23 searches for the one having the maximum absolute value among the 110 columns, determines the delay time 2□ (corresponding to the time position of the first pulse) and the amplitude Jv, Furthermore, the above z1
tJ1 is output to the cross-correlation corrector 24 and multipulse memory 25 via transmission lines 231 and 232. The cross-correlation corrector 24 is connected to the transmission line 19 from the cross-correlation coefficient memory 19.
The following (11
) is corrected by the formula.

1'Hx (z4+t)=l'H, (zt+[) -bz
l""hh(t) (11) However, t is a correction interval and is set to 18 to 10S. The cross-correlation corrector 24 further supplies the result of equation (11) above to the cross-correlation coefficient memory 19 via the transmission line 241.

The above processing is repeated until the number of multipulses required is reached, and the results are stored one after another in the multipulse memory 25. After the repetition is completed, the multi-pulse memory 25 outputs the time position and amplitude of the multi-pulse to the transmission line 1101.

Next, FIG. 7 shows an actual measurement example of the 'hx and corrected goue obtained by the configuration of FIG. 6 (the audio sample is the same as the example of FIG. 2). Figure 7 (A) shows the cross-correlation coefficient 'b
The time position and amplitude of the first pulse determined as x are shown. (B) shows the time position and amplitude of the second pulse determined as the A cross-correlation coefficient ghx corrected by 0(fil) by the first pulse.Similarly, (C) to ( K
) shows the time position and full width of the pulse corrected and determined as fclbx.

The present invention differs from Tezawa et al.'s algorithm in that the cross-correlation coefficient is 1.
'hx and normalized autocorrelation coefficient 9'5. Searching for the maximum value of similarity with. As a result, φ as shown in Figure 2 (F)
A pulse is not determined at a time position where the shapes of hx and hb are significantly different, and therefore the number of pulses does not increase unnecessarily.

Next, input audio according to the present invention (audio signal at terminal 7001)
S/ of the output audio (audio signal at terminal 181) based on
The N ratio was measured by varying the number of multi-pulses. All data examples are shown in FIG. 8 in comparison with the S/N ratio measured by the Tezawa et al. algorithm using the same method. In FIG. 8, X is the S/N ratio according to the conventional method, and .
Shows the N ratio.

As is clear from FIG. 8, the code efficiency of the present invention is improved over the algorithm of Tezawa et al.

Above, as an example of the degree of similarity, we have shown the sum of products of the cross-correlation coefficient 1bx or 'hx modified in consideration of the influence of pulses and the normalized autocorrelation coefficient Rhh. It is limited to the sum of products and is not an illusion.

For example, 'hx shown in the following equation (12) and old 55. Cml with maximum magnitude gold at lag mI with
, and then search for the ml with the minimum magnitude at each delay and the maximum product similarity.

N.

C1i,,=min Σ l f' h, (m, +
S) Cm+8=-NB Rhh (S) l ・・・・・・・・・・・・(12
) When magnitude is used as the similarity, the autocorrelation normalizer 20 is not necessarily required. Also, a sum of products calculator 22. It is obvious that by replacing M1 with a minimum magnitude estimator and a minimum value searcher such as the maximum value searcher 23, the similarity cost can be reduced by 4:1.

Also, in the above explanation, it is shown in equation (1)! Although the skewer is based on the assumption that the perceptual weighting is performed, it is not necessarily necessary to perform the perceptual weighting. (In formula (I), γ=
1.0), and when perceptual weighting is not performed, the present invention sets the cross-correlation function between the input speech waveform and the impulse response of the speech synthesis filter to 'hx, and the autocorrelation function of the impulse response of the speech synthesis filter to can be performed as ``hb''.

In the embodiments of the present invention shown in FIGS. 4 and 5, the 1st parameter is used as the LPG coefficient, but other LPC coefficients such as the α parameter may also be used. The encoder and multiplexer, and the decoder and demultiplexer, respectively, integrate these.
It is clear that the present invention can be implemented in the same manner even if the LPC synthesis filter has such a configuration, and it is also clear that the LPC synthesis filter can be implemented in almost the same way even if the LPC synthesis filter is replaced with a non-polar type digital filter or the like other than the all-pole type.

As described above, according to the present invention, a multipulse vocoder includes means for calculating a cross-correlation coefficient between an input speech signal and an impulse response of a speech synthesis filter, and means for calculating a sequence of Nij-Iffine pulse response autocorrelation coefficients. , How to calculate the similarity between the cross-correlation coefficient sequence and the autocorrelation coefficient sequence!・Equip the analysis side, and further search for the maximum value of the similarity and calculate the impulse sequence (multipulse) +! g, i, (and having a means for calculating the positions of l'j in a forward manner), it is also possible to efficiently reduce the influence of multi-pulses from the cross-correlation coefficient sequence.
This has the effect of making it possible to improve the efficiency of multi-pulse encoding.

[Brief explanation of the drawing]

Fig. 1 shows the basic configuration of a conventional multi-pulse 4 (vocoder -!Block diagram, 2nd) + (A) - (1<) are mutual correlations in the conventional method? The waveform diagram showing the relationship with the multi-pulse vocoder such as J, l bx) and the autocorrelation coefficient 1t, -, z of the impulse response of the third I A block diagram showing an embodiment of the analysis side of the vocoder, 51'&l; Synthesis of multi-pulse vocoder according to the present invention 6! FIG. 6 is a block diagram showing one embodiment of the similarity field Li2 unit 11.
(IN)~(K)l-j: Cross-correlation coefficient according to the present invention"
Relationship between hx and multipulse determiner 1: iFj? The waveform diagram shown in Part 8 is the effect of the present invention on arching efficiency ihJ? What is the S/N compared to the conventional method? It is a ``a1'' rated hourglass figure. 1...L, PC synthesizer, 2...LPC
Separator, 3...Back source pulse generator, 4...
...Subtractor, 5... Auditory weighting is device, 6...
...Squared error minimizer, 7...LPC analyzer, 8...Cross correlation l! I number calculator, 9...
... Encoder (1), 10 ... Autocorrelation function calculator, 11 ... Similarity calculator, 12 ...
...Encoder (2), 13...Multiplexer 114...Demultiplexer, 15...
... Temporary encoder (1), 16... Temporary encoder (2)
, 17...1. PC synthesizer, 18...
LPF. 19... Mutual A-memory coefficient memory, 20...
... self); inter-eye normalizer, 21 ... autocorrelation coefficient memory, 22 ... sum of products header)
...Maximum value detector, 24...Cross-correlation peak corrector, 25...Multi-pulse memory. 1st Vertex -30-20-/D O#+ 26. 30 things 3j y Figure 4' Mine 2 times B l θrノ0/left 2ρ2 left C +) shift, teach) False ~ + (N ~ zu force Ka ×

Claims (1)

[Claims] The input audio signal is subjected to LPC (Linea) for each analysis frame.
rPrediction Coefficient (Linear Prediction Coefficient) The LPC coefficients analyzed and extracted are used as spectral envelope information, and together with this spectral envelope information, sound source information that constitutes the audio information of the input audio signal is generated corresponding to the characteristics of this sound source information for each analysis frame. Multiple impulse sequences (multipulses) with time positions and amplitudes
Analysis and synthesis of the input audio signal 2
In the multi-pulse vocoder, means for calculating a cross-correlation coefficient sequence between the input speech signal and an impulse response of a speech synthesis filter; The analysis side includes means for calculating the degree of similarity f with the autocorrelation coefficient sequence, and further includes means for searching for the maximum value of the degree of similarity and calculating the amplitude and position of the impulse sequence (multipulse) in a forward manner. A multi-pulse type vocoder having an analysis side.