JPH02237233A - Adaptive predictive encoding/decoding system - Google Patents

Abstract

PURPOSE:To suppress the increase of the redundancy of transmission data and to prevent the deterioration of the quality of restored data by adding an error detection code to a prediction parameter so as to attain transmission and obtaining a prediction signal through the use of the prediction parameter of a preceding block when a reception-side detects errors. CONSTITUTION:An error detection code generation means 4 installed on a transmission-side generates the error detection code to the prediction parameter. In the reception-side, an error detection means 5 detects the errors in the received prediction parameter from the prediction parameter and the error detection code, and a prediction parameter selection means 6 selects the prediction parameter of the past block stored in a prediction parameter storage means 7 and supplies it to a prediction means 8 when the error is detects. Thus, the increase of the redundancy of transmission data is suppressed and the deterioration of the quality of restored data due to the errors which occur in transmission information of the prediction parameter can be prevented.

Description

[Detailed Description of the Invention] [Summary] Regarding an adaptive predictive encoding/decoding system, the quality of restored data due to errors occurring in transmission information of prediction parameters is improved while suppressing increase in redundancy of transmission data as much as possible. The purpose of this is to effectively prevent the deterioration of the input signal, and on the transmitting side, a prediction parameter is obtained by capturing the input signal block by block and performing predictive analysis, and the prediction signal obtained based on the prediction parameter and the input signal are combined. By transmitting the difference of , and the entire prediction parameter, on the receiving side, the prediction signal is obtained using the prediction parameter, and the sum of the difference received at the same time as the prediction parameter and the prediction signal is calculated. In an adaptive predictive encoding/decoding system that restores the input signal, an error detection code is added to the prediction parameter and transmitted, and the receiving side performs error detection on the prediction parameter, and if there is an error, , the prediction signal is determined using the prediction parameters of the previous block.

[Industrial application field]

The present invention relates to an adaptive predictive encoding/decoding system.

Audio data signals, image data signals, etc. are temporally divided into blocks, predictive analysis is performed, the data signal of the new block is predicted, the difference between the data signal of the new block and the predicted data signal is determined, and the data signal of the new block is calculated. By transmitting the difference data and the prediction parameters used in the prediction, the data is compressed and transmitted, and on the receiving side, the prediction data signal is obtained from the prediction parameters, and the sum of the prediction data signal and the received difference is calculated. The adaptive predictive encoding/decoding method for restoring the data signal is well known.

In such an adaptive predictive encoding/decoding system, if an error occurs in the predictive parameters due to line conditions or the like, the quality of the restored data will be fatally degraded.

For example, in recent years, adaptive predictive encoding and decoding of voice signals has been used in digital radio, etc., but due to phasing in wireless links, pinch and prediction of voice signals transmitted as voice signal prediction parameters, etc. The problem is that many errors occur in coefficients, etc.

Therefore, there is a need for countermeasures against errors that occur in transmission information of prediction parameters in adaptive predictive encoding/decoding systems.

[Prior art and problems to be solved by the invention]

Conventionally, for example, in the above-mentioned digital wireless system, an error correction code is added to a prediction parameter and transmitted, and on the receiving side, an error correction code in the transmission data of the prediction parameter is detected by the added error correction code. There are some things to fix.

However, error correction codes greatly increase the redundancy of transmitted information, which goes against the original purpose of encoding and decoding, which is redundancy reduction.

In particular, when transmitting and receiving audio data, the amount of prediction parameter data is not very small compared to the amount of audio difference data, so adding error correction codes to such prediction parameter data increases redundancy. An increase in the number will greatly reduce the effectiveness of the original encoding.

The present invention has been made in view of the above problems, and effectively prevents the deterioration of the quality of restored data due to errors occurring in the transmission information of prediction parameters while minimizing the increase in redundancy of transmitted data. The purpose of this invention is to provide an adaptive predictive encoding/decoding system that can perform the following steps.

[Means to solve the problem]

FIG. 1 is a basic configuration diagram of the present invention. In this figure, 1
2 is a difference means, 2 is an analysis means, 3 is a first prediction means, 4 is an error detection code generation means, 5 is an error detection means, 6 is a prediction parameter selection means, 7 is a prediction parameter storage means, and 8 is a second prediction means. 9 is a prediction means and an addition means.

The analysis means 2 obtains prediction parameters by taking in the input signal block by block and performing predictive analysis.

The first prediction means 3 obtains a prediction signal based on the prediction parameter.

The difference means 1 calculates the difference between the input signal and the predicted signal.

These are provided on the transmitting side.

The second prediction means 8 uses the prediction parameters to obtain the prediction signal.

The adding means 9 calculates the sum of the difference and the predicted signal obtained by the second predicting means 8.

These are provided on the receiving side.

According to the present invention, an error detection code generating means 4 further provided on the transmitting side generates an error detection code for the prediction parameter.

According to the present invention, the receiving side further includes the error detection means 5;
Prediction parameter storage means 7 and prediction parameter selection means 6 are provided.

The error detection means 5 detects an error in the received prediction parameter from the received prediction parameter and the error detection code.

The prediction parameter storage means 7 stores prediction parameters of past blocks.

The prediction parameter selection means 6 normally selects the received prediction parameters, and when the error is detected, selects the prediction parameters of the past blocks stored in the prediction parameter storage means 7 and sends them to the second prediction means. Supply to 8.

[For production]

On the transmitting side, a prediction parameter is obtained by capturing the input signal block by block and performing predictive analysis, and transmitting the difference between the prediction signal obtained based on the prediction parameter and the input signal, and the prediction parameter, On the receiving side, the prediction signal is obtained using the prediction parameter, and the input signal is restored by calculating the sum of the prediction signal and the difference received at the same time as the prediction parameter.

According to the present invention, the prediction parameter is transmitted together with an error detection code, and the receiving side performs error detection on the prediction parameter, and when there is an error, calculates the prediction signal using the prediction parameter of the past block. , restore the received data.

In general, error detection codes require a considerably smaller amount of data than error correction codes. Therefore, the increase in redundancy is reduced.

Furthermore, the quality of restored data is improved because prediction parameters that have become abnormal values due to errors occurring in the transmission path are no longer used.

In particular, in the case of data with a strong correlation between the preceding and succeeding blocks, the deterioration in quality can be considerably reduced by using the prediction parameters of the immediately preceding block.

〔Example〕

FIG. 2 shows the configuration of a system for encoding and decoding audio signals as an embodiment of the present invention.

In FIG. 2, 11 is a subtracter, 12 is an analysis means, 13
is a predictor, 14 is an error detection encoder, 15 is a quantizer, 1
6 is an inverse quantizer, 17 is an adder, 18 is a multiplexer, 2
0 is a transmission line, 21 is a demultiplexer, 22 is an error detection decoder, 23 is a memory, 24 is a selector, 25 is a predictor,
26 is an adder, and 27 is an inverse quantizer.

To encode a voice signal, for example, as shown in FIG. 3, assume that voice is formed by vocal cord waves from a sound source (vocal cords) having a predetermined pinch passing through the vocal tract as a filter. This is done based on a model.

Therefore, the prediction parameters in this case are mainly the pitch described above and the prediction coefficients of the vocal tract filter that simulates the vocal tract.

Further, the audio input signal is divided into sections of about 20 ms, for example, and the audio signal of each section constitutes one frame of data.

The analysis means 12 in FIG. 2 uses linear predictive analysis to find the optimal prediction parameters. This prediction parameter is input to the multiplexer 18 for transmission.

The predictor 13 obtains a predicted signal based on the data stored therein and the above-mentioned prediction parameters, and the subtracter 11 obtains a difference from the input signal.

The difference is quantized in a quantizer 15 and input to a multiplexer 18 for transmission, as well as an inverse quantizer 1.
The input signal is dequantized in step 6, and then summed with the predicted signal in an adder 17, thereby restoring the input signal and supplying it to the predictor 13. This is used to generate a prediction signal later.

An error detection encoder 14 provided according to the invention generates an error detection code for said prediction parameters and inputs it to a multiplexer 18 for transmission.

The multiplexing device 18 stores the difference data, prediction parameters,
and error detection codes for the prediction parameters are multiplexed and sent onto the transmission path 20.

On the receiving side, the demultiplexer 21 separates the difference data, the prediction parameter, and the error detection code for the prediction parameter from the multiplexed signal transmitted on the transmission path 20, and separates the difference data, the prediction parameter, and the error detection code for the prediction parameter, respectively. is output to the inverse quantizer 27 , the prediction parameter is output to the selector 24 and the error detection decoder 22 , and the error detection code is output to the error detection decoder 22 .

The error detection decoder 22 detects errors in the prediction parameters from the prediction parameters and the error detection code, and controls the selector 24 based on its output.

The selector 24 selects one of the received prediction parameters and the prediction parameters stored in the memory 23 and supplies it to the predictor 25 .

The selection is normally performed so that the received prediction parameters are output, and when the error detection decoder 22 detects an error, the prediction parameters to be stored in the memory 23 are selected.

The output of the selector 24 is not only supplied to the predictor 25 but also input to the memory 23 to update its stored contents.

The predictor 25 generates a prediction signal based on the prediction parameters and past data stored by itself, and the adder 26
supply to.

The adder 26 calculates the sum of the difference data dequantized by the dequantizer 27 and the predicted signal, and this becomes the restored data, that is, the audio reproduction signal.

The audio reproduction signal is also input to a predictor 25 for generating a predicted signal.

The length of one sound of an audio signal is usually longer than several audio frames of 20 ms in length, and during this time, the above-mentioned prediction parameters are almost common.

Therefore, even if there is an error in the data of one frame, the quality of the reproduced audio signal will often hardly deteriorate by using the predictive parameters of the previous frame.

Further, even if an error is detected in a certain frame and prediction parameters from past frames are used, if the error is recovered in the next and subsequent frames, the system immediately returns to normal operation.

〔Effect of the invention〕

According to the present invention, it is possible to effectively prevent deterioration in the quality of restored data due to errors occurring in transmission information of prediction parameters while suppressing increase in redundancy of transmission data as much as possible.

[Brief explanation of drawings]

FIG. 1 is a basic configuration diagram of the present invention, FIG. 2 is a configuration diagram of an embodiment of the tree invention, and FIG. 3 is a diagram showing a model of the speech generation process. [Description of codes] 1... Differential means, 2... Analysis means, 3... First prediction means, 4... Error detection code generation means, 5... Error detection means, 6... - Prediction parameter selection means, 7...
- Prediction parameter storage means, 8... second prediction means,
9... Adding means, 11... Subtractor, 12... Analyzing means, 13... Predictor, 14... Error detection encoder,
15... Quantizer, 16... Inverse quantizer, 17...
Adder, 18... Multiplexer, 20... Transmission line, 21... Demultiplexer, 22... Error detection decoder, 23... Memory, 24... Select, 25...・・・
Predictor, 26... Adder, 27... Inverse quantizer. Diagram showing the basic configuration of the present invention]

Claims (1)

[Claims] 1. On the transmitting side, a prediction parameter is obtained by capturing the input signal block by block and performing predictive analysis, and a difference between the prediction signal obtained based on the prediction parameter and the input signal; On the receiving side, the prediction parameter is used to obtain the prediction signal, and the input signal is restored by calculating the sum of the difference received at the same time as the prediction parameter and the prediction signal. In an adaptive predictive encoding/decoding system, an error detection code is added to the prediction parameter and transmitted, and the receiving side performs error detection of the prediction parameter, and if there is an error, the prediction of the previous block is performed. An adaptive predictive encoding/decoding system characterized in that the predicted signal is obtained using parameters. 2. Analysis means (2) for obtaining a prediction parameter by taking in an input signal block by block and performing predictive analysis; a first prediction means (3) for obtaining a prediction signal based on the prediction parameter; and the input signal. a difference means (1) for calculating a difference between the predicted signal and the predicted signal; a second differential unit for calculating the predicted signal using the predicted parameters;
Adaptive predictive coding, comprising: a prediction means (8) on the receiving side; and an addition means (9) for calculating the sum of the difference and the prediction signal obtained in the second prediction means (8). In the decoding system, the transmission side is provided with an error detection code generating means (4) for generating an error detection code for the prediction parameter, the error detection code is transmitted together with the prediction parameter, and the reception side receives the error detection code. Error detection means (5) for detecting errors in the received prediction parameters from the prediction parameters and error detection codes; prediction parameter storage means (7) for storing prediction parameters of past blocks; and usually the received prediction parameters. When the error is detected, prediction parameter selection means () selects a prediction parameter of a past block stored in the prediction parameter storage means (7) and supplies it to the second prediction means (8); 6) An adaptive predictive encoding/decoding system comprising: