JPS6337723A - Coding transmitter - Google Patents

Abstract

PURPOSE:To obtain a reproducing signal which has almost as good quality as an original signal has by selecting a signal which has minimum reproducing signal distortion among the reproducing signals of sequential adaption predictive reproducing devices which differ in characteristics and informing it to the decoding side of that. CONSTITUTION:A residual signal X after pitch prediction is supplied to a sequential adaption predictive filter 1 to select the closest signal E' by a matching part 16 through the matching of a code book 15 from a residual signal E after proximity prediction by sequential adaptive prediction. Further, an optimum reproducing device determination part 2 operates plural sequential adaptive reproduction parts ADF1 - ADFm for the signal E', thereby obtaining reproducing signals X1' - Xm'. The S/Ns of those playback signals to the residual signal X are found, frame by frame, and the results are evaluated to determine and send out the number (n) of the sequential adaption predictive reproduction part ADFn having the best S/N. A decoding side performs the reverse operation of the adaptive prediction according to the transmitted predicting device number. Consequently, reproduced sound distortion due to matching distortion can be reduced and the pit rate is decreased to perform high-efficiency coding transmission.

Description

[Detailed Description of the Invention] [Overview] In order to reduce signal distortion caused in the encoding process, the present application sequentially selects one that minimizes reproduced signal distortion from among sequential adaptive predictive regenerators with different characteristics. This invention discloses a 71 encoding transmission apparatus that corrects the distortion of the reproduced signal by informing the decoding side of the error, thereby making it possible to obtain a reproduced signal 13 with a quality close to that of the original signal.

[Industrial application field]

The present invention relates to a coding/transmission device, and particularly to a highly efficient coding/transmission device for audio or image signals.

With the introduction of digital lines, there is a demand for highly efficient transmission of audio signals, image signals, etc., and improved serviceability. In particular, high-efficiency coding of audio signals has a wide range of applications, including improving line utilization efficiency in coded transmission, reducing the amount of information stored in audio MAR/response systems, and applying to mobile communications and satellite communications. Therefore, there is a need for a highly efficient encoding/transmission device that can compress the information amount of the audio signal and provide good reproduced audio.

[Conventional technology]

Figure 4 shows an example of a conventional encoding and transmitting device.
This has already been done by Bell Laboratories in the US.
, S, ^ta I) etc. are CB L P (Coded-
Excited Linear Prediction
) method was improved by Mr. Gen Oyama of the Radio Research Institute of the Ministry of Posts and Telecommunications. The paper was published under the title ``Linear Prediction Analysis and Synthesis Method''.

In the figure, a pinch prediction analysis circuit 11 predictively analyzes an input signal S, outputs a pinch prediction coefficient, and controls a pitch prediction filter (Ppitch) 12 to extract a residual signal X from which periodic predictable components are removed. let This residual signal
pc) 14 to extract a residual signal E from which short-cycle predictable components are removed from the pitch prediction filter 12.

The residual signal E after proximity prediction extracted by removing the correlation of the input signal S by the pitch prediction filter 12 and the linear prediction filter 14 in this way is matched with the noise codebook 15 consisting of a noise sequence (pattern). 16,
Multiplexing circuit 1 selects the closest noise and uses its code number
7 to the decoding side.

On the decoding side, the demultiplexing circuit 18 receives the pitch prediction coefficient, the linear prediction coefficient, and the code number, and the readout circuit 1 receives the pitch prediction coefficient, linear prediction coefficient, and code number.
In step 9, a noise sequence is read out from the codebook 20 corresponding to the codebook 15 on the encoding side according to the sent code number, and a linear prediction filter (P
LFC) 21 and pinch prediction filter (Ppttc)
h) A reproduced signal S' is obtained by passing the signal through 22.

[Problem that the invention seeks to solve]

In such a conventional coding transmission device, matching distortion of the codebook (residual signal E and selected Since there is no means to correct this distortion, the signal reproduced using the residual signal E゛ based on this noise sequence has sufficient quality. The problem was that it could not be done.

In addition, since linear prediction analysis is performed as proximity prediction, it is necessary to transmit proximity prediction coefficients in addition to pitch prediction coefficients, which requires a large amount of auxiliary information to be transmitted for each frame, which becomes an obstacle to reducing the transmission bit rate. was.

Therefore, the purpose of the present invention to solve this problem is to
The object of the present invention is to realize an encoding/transmission device capable of correcting distortion caused in the encoding process by collating a codebook.

[Means for solving problems]

FIG. 1 is a diagram showing the principle of the coding/transmission apparatus of the present invention to achieve the above object, and 1 is a diagram showing the principle of the coding/transmission apparatus of the present invention, in which 1 is a sequential adaptive prediction system that performs sequential adaptive prediction as a reference from the residual signal X after pitch prediction of the input signal. The adaptive prediction filter 16 is a matching unit that sequentially compares the prediction residual signal E outputted from the adaptive prediction filter 1 with a codebook 15 consisting of a noise sequence, selects the closest noise E, and transmits the selection number. , 2 is a plurality of sequential adaptive predictive recovery filters ADF +~^OF with different update rates that sequentially adaptively predictively reproduce the selected closest noise E°.
, and these plurality of sequential adaptive predicted reproduction values <xl'
~X, ゛) and the residual signal X after the pinch prediction.
, and transmits the number n for use in successive adaptive predictive reproduction as proximity prediction on the decoding side.

[For production]

In FIG. 1 illustrating the present invention, a residual signal X after pitch prediction is determined by a matching unit 16 from a residual signal E after proximity prediction by successive adaptive prediction in a successive adaptive prediction filter 1 and a signal that is the closest to the signal by comparing with a codebook 15. Select E゛. Furthermore, the optimum reproduction performance determining section 2 operates a plurality of sequential adaptive predictive reproduction sections ADF, ~ADF, for this signal E' to obtain reproduced signals x1'~X,°. The S/N of these reproduced signals x1° to xII'' is determined for each frame with respect to the residual signal Decide and send.

On the decoding side, the inverse operation of adaptive prediction is performed according to the transmitted predictor number.

〔Example〕

Embodiments of the encoding and transmitting apparatus according to the present invention will be described below.

FIG. 2 shows an embodiment of the encoding and transmitting apparatus of the present invention conceptually shown in FIG. 1. The difference from FIG. Using a sequential adaptive prediction filter (ADF rt )l;
The signal E° generated in the matching unit 16 is commonly input to a plurality of parallel-connected successive adaptive predictive regeneration filters ADF, ~^Y, whose update rates are different from each other. This reproduction signal Xl
The S/N calculation circuit 3 calculates the S/N between °~X,゛ and the residual signal X.
S
/N evaluation circuit 3 determines and selects the optimum reproducibility determining section 2.

Therefore, even on the decoding side, the sequential adaptive prediction inverse filter 30 is controlled by inputting the optimal predictor number corresponding to the encoding side.
(ADF, 1) is used.

Here, the sequential adaptive prediction filter (^oF,, t) and the sequential adaptive prediction recovery filter ADF +~ADF will be explained. These filters have zero prediction coefficients ( This is already known as an adaptive filter having a winding prediction coefficient [i] and a winding prediction coefficient [i].

Zero prediction coefficient 081 (+1°l)・L, *Csj (
al+ D, *sgn(E(n))*sgn(E(n
-j)) +11 scratch prediction coefficient C,, ta+u・L,
*C,, + D, 'ksgn(E(n))'ksgn
(X(n-D) +2) where L, , L is a leak coefficient, sgn is a sign function representing the positive or negative sign, D,
, D indicates the update rate of the prediction coefficient, and the filter 80F'
+~A[lF, is the update rate, and the values of ,D, are different from each other.

Next, the operation of the embodiment of the present invention shown in FIG. 2 will be explained.

On the encoding side, long-term prediction (pitch prediction) is applied to the input signal S.
The residual signal X obtained by performing the above processing is passed through the sequential adaptive prediction filter l as a proximate prediction, and becomes the residual signal E. This residual signal E is checked against the noise codebook 15 in the matching section 16, and the code number of the closest noise is selected. The residual signal E° at this time is further inputted in parallel to successive adaptive predictive regeneration filters ADF, -ADF, and sequential adaptive predictive regeneration is performed, respectively, to obtain regenerated signals X1'' to x, . Playback signal X1
S between '~x, ° and the residual signal X after pitch pref'/ff4
/N is calculated for each frame by the S/N evaluation circuit 3, the best S/N is determined, and the number n of the reproduction filter ADF corresponding to the best S/N is sent to the multiplexing circuit 17.

In this case, the optimal adaptive predictive reconstruction filter ADF.

is selected using the intra-frame error signal power as shown in equation (3) below, and the one with the minimum value is selected.

P, -ΣIX(i)-X, '(iN '' +
3, where 1=1 to 1, and k indicates the number of samples in the frame.

On the decoding side, the demultiplexing circuit 18 extracts pitch prediction coefficients and
The code number n output by the matching section 16 and the S/N
The number n of the optimal sequential adaptive predictive recovery filter output from the evaluation circuit 3 is separated and output, and the code number is read out from the noise code book 20 by the readout circuit 19 as the noise signal E' corresponding to the code number. In the sequential adaptive predictive filter (ADF, l) 30, this noise signal E° is sequentially subjected to adaptive predictive regeneration processing according to the number n that instructs the regeneration filter ADF, which performs optimal proximity prediction as a sequential adaptive predictive regeneration filter. The signal Xn'' is outputted as a decoded signal Sn'' by the Lipinoch prediction inverse filter 22.

Therefore, the distortion (X-X') caused by the distortion (E-E') caused by matching is reduced to (X-X,').

In addition, in order to avoid data discontinuity between frames, the present applicant published a product in April 1986 entitled "High-efficiency coding transmission device."
It is disclosed in Japanese Patent Application No. 61-80063 filed on May 9th, and it is necessary to perform copy control of parameters as shown in FIG. That is, the parameter copy control unit 35 in FIG. 3 continuously inputs and monitors the parameters of the adaptive predictive filters ADF, -ADF, and selects the parameters P of the filter ADF, which is selected as the optimum one.
7 (prediction coefficients/tap data) to filter l as well as filter 8 opl~ADF at the end of the frame.

〔Effect of the invention〕

As described above, according to the encoding and transmitting apparatus of the present invention, matching distortion caused by codebook matching of the residual signal is reproduced by a plurality of successive adaptive predictive recovery filters having different update rates,
The S/N with the residual signal after pitch prediction 1jIl is calculated to select the optimal one, and the optimal proximal prediction reproduction filter is instructed to the decoding side, so that reproduced audio distortion caused by matching distortion can be reduced. In addition, since successive adaptive prediction is used as the proximate prediction, there is no need to encode and transmit prediction coefficients, thereby reducing the bit rate and enabling highly efficient encoding and transmission.

[Brief explanation of the drawing]

Fig. 1 is a principle diagram of the coding transmission device according to the present invention. Fig. 2 is a circuit diagram showing an embodiment of the coding transmission bag U according to the invention. FIG. 4 is a block diagram showing a conventional coding and transmitting device. In FIG. 1, 1 is a sequential adaptive prediction filter, and 2 is an optimal reproduction performance. The determining unit, ADF + ~ ADF, is a sequential adaptive prediction recovery filter;
15 is a codebook, 16 is a matching unit, IE is a prediction residual signal, x, ゛ to x, ゛ is a number indicating a successive adaptive prediction reproduction value, and n is a number indicating an optimal reproduction filter ADF. In the drawings, the same reference numerals indicate the same or corresponding parts. Figure 1

Claims (2)

[Claims] (1) A sequential adaptive prediction filter (1
), and compares the prediction residual signal (E) output from the sequential adaptive prediction filter (1) with the noise sequence codebook (15), selects the closest noise (E'), and sends out its selection number. a matching unit (16) that performs sequential adaptive predictive reproduction from the selected closest noise (E'); An adaptive prediction reconstruction filter that gives the optimal S/N is selected from each of the values (X_l' to X_m') and the residual signal (X) after the pitch prediction, and its number (n) is sequentially adapted on the decoding side. An encoding transmission device comprising: an optimum regenerator determining unit (2) for transmitting data for use in predictive reproduction. (2) In order to avoid data discontinuity of reproduction output when selecting the sequential adaptive prediction filter for each frame, the optimal regenerator determining unit (2) selects the sequential adaptive prediction parameters of the selected reproduction filter. The encoding/transmission apparatus according to claim 1, further comprising a parameter copying control section (35) that performs copying control on other filters.