JPS5830780B2 - Multi-channel Shingoshiyorihoushiki - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a multichannel signal processing system.

Currently, various audio transmission methods are used, among which PCM...! (Digital modulation methods such as PCM and JM have the great advantage that there is almost no deterioration of the audio signal even if the transmitted signal is slightly affected by interference such as S/N of the transmission path.

Among these methods, JP is used for continuous signal transmission.
CM is used.

FIG. 1 is a signal waveform diagram for explaining the JPCM method.

That is, 1 is an input signal, and 2 to 6 are sample points at which sample values are created by sampling the input signal 1 at a constant period.

A general PCM processing method quantizes and encodes each sample value without any relationship between the two before and after, but in APCM, when sampling, it is compared with the sample value one sample before, and the difference is calculated. It performs signal transmission by quantizing and encoding.

For example, at sample point 4, the sample value is not transmitted as is, but the value obtained by subtracting the sample value one sample before, that is, sample point 3, is quantized and encoded, and then the signal is transmitted.

This method uses not only the sample value of the previous sample, but also
There is also a method of generating a voltage for comparing sample values to be transmitted from sample values two or three samples ago.

The voltage used for this comparison is called a predicted value.

Figure 2 is a block diagram of the JPCM system, where 7 is an input terminal;
8 is a band-limiting filter, and 9 is a subtracter, which outputs the difference between the input signal and the predicted value.

A quantizer 10 samples and quantizes the output of the subtracter 9 at a sampling frequency.

11 is an A/D converter that converts the output of the quantizer 10 into a digital quantity.

12 is an output terminal.

13 is an adder that adds the output of the quantizer 10 and the predicted value, and outputs a value equal to the input signal one sample before.

Reference numeral 14 denotes a predictor which inputs the output of the adder 13, that is, a value equal to the input signal of one sample before, and outputs a predicted value for the next sample.

However, in multi-channel signal transmission, there is considerable similarity between the signals of each channel, so
Transmitting each channel as is is wasteful in terms of constructing predicted values when considering signals with high frequencies and rapid level changes. The drawback is that you have to increase it.

Therefore, it is an object of the present invention to provide a similarity multi-channel signal processing scheme that can reduce the number of quantization bits.

FIG. 3 is a four-channel audio signal waveform diagram for explaining an embodiment of the multi-channel signal processing method of the present invention.

That is, 15 is an audio signal of the left front channel, 16 is an audio signal of the left rear channel, 17 is an audio signal of the right rear channel, and 18 is an audio signal of the right front channel.

19 to 22 are sample points of the signal 15, 23 to 26 are sample points of the signal 16, 27 to 29 are sample points of the signal 17, and 30 to 32 are sample points of the signal 18.

The sampling periods of signals 15.16 and 17.18 are shifted by 1γ4 periods.

In this invention, when creating a predicted value, the predicted value is created by considering not only the sample values of that channel but also the sample values of other channels.

For example, the predicted value at sample point 20 of signal 15 is:
Not only the sample value of sample point 19 but also sample point 2
3,2γ.

It is created considering 30 4 sample values.

FIG. 4 is a block diagram of such a four-channel audio signal processing system.

That is, 33 to 36 are input terminals for the left front, left rear, right rear, and right front channels, respectively, 37 to 40 are band limiting filters for each channel, and 41 to 44 are subtraction between the input signal of each channel and the predicted value. 45-48 are quantizers for each channel, 49-52 are A/Ds for each channel.
Converters 53 to 56 are adders that output a value equal to the input signal one sample before.

57 to 60 are predictors for each channel, and the audio signals 15 to 1 of each channel are sampled with a delay of 1γ4 periods.
8 (FIG. 3) as input and outputs a predicted value.

61 to 64 are output terminals for the JPCM signals of each channel.

By using this circuit configuration, the signal processing method described in connection with FIG. 3 can be realized.

At the outputs of adders 53-56, the actual signal (not the difference) of each channel is stored.

Therefore, the output of the adder 53 includes the A signal 15 in FIG.
The signal at sample point 19 of
The signal at the sample point 23 of the signal A 16 in FIG. 3 appears at the output of the adder 55, and the signal A of the signal 1 in FIG.
A signal at sample point 2γ of signal 18 appears at the output of adder 56, and a signal at sample point 30 of signal 18 appears.

The outputs of adders 53 to 56 for each channel are connected to predictors 57 to 60 for each channel, and predicted values are created for each channel.

For example, the predictor 57 calculates the predicted value for the value of sample point 20 of signal 15 by the value of sample point 19 of signal 15, the value of sample point 23 of signal 16, and the value of sample point 2 of signal 11.
7 and the value of sample point 30 of signal 18.

Specifically, the predicted value is calculated by performing a simple average of the above four values or a weighted average with emphasis placed on the value of the own channel.

In the case of multi-channel music, there is similarity between each channel, and as explained in Figure 3, sample points 20 and 30 in Figure 3 are separated by only 1γ4 sample periods, and the sample period Considering that frequency components higher than 1 are not included, the configuration of this embodiment makes it possible to extract a better predicted value and therefore to process the audio signal with more fidelity.

When creating a normal predicted value, the shorter the sample period, the smaller the difference between the actual signal and the predicted value, and it is possible to reproduce sound with sufficient fidelity even with a small number of bits.

In the case of multi-channel, if the signals of each channel are similar, you can create a predicted value by sequentially shifting the sample points of each channel and taking into account the values at sample points of other channels that are close to the current sample point of the own channel. , it is possible to create a predicted value that is closer to the predicted value when the sample period is shortened, and closer to the actual signal.

This means that if the signals of each channel are similar, when the sample points of each channel are sequentially shifted, the value of the sample point of another channel that is closer to the current sample point of the own channel will be one value before that of the own channel. This is because the value of the signal appearing at the current sample point of the own channel is often closer to the value of the signal appearing at the current sample point of the own channel than the value of the sample point of .

If similarity is emphasized in configuring predicted values, the following configuration may be considered in addition to the embodiments described in FIGS. 3 and 4.

That is, the combinations with the strongest similarity are the right front and right rear channels, and the left front and left rear channels.

A conceivable method for creating predicted values in this case is to create predicted values separately for the left and right channels by shifting the sample period of the front and rear surfaces by 1/2 period.

Furthermore, among the four channels of the left front, left rear, right front, and right rear, the channels on the diagonal have the weakest similarity.

Therefore, a predicted value is created from the other three channels by excluding the diagonal channel with weak similarity.

For example, the predicted value when sampling the left front channel is created from the sample value of the left front channel one sample before, the right front channel, and the left rear sample value, and the predicted value of the left rear channel is created from the sample value of the left front channel one sample before. A possible method is to create it from the sample values of the rear right channel and the front left channel.

As described above, according to the multi-channel signal processing method of the present invention, in similar multi-channel signal transmission, the number of quantization bits can be reduced compared to when JPCM processing is performed on each channel separately. If the same number of quantization bits is used, more faithful audio can be transmitted.

[Brief explanation of drawings]

Fig. 1 is a signal waveform diagram for explaining the conventional JPCM method, Fig. 2 is a block diagram of the conventional JPCM method, and Fig. 3 is a 4-channel diagram for explaining the multi-channel signal processing method of the present invention. The audio signal waveform diagram in FIG. 4 is a block diagram of an embodiment of the multi-channel signal processing method of the present invention. 15~18...4 channel audio signal, 19~
32...Sample points, 57-60...
Predictor.

Claims (1)

[Claims] 1 A multi-channel signal processing method for processing input signals of a plurality of channels having mutual similarity, the method comprising: a filter for limiting the band of the input signal; a prediction means for outputting a predicted value; and a prediction means for outputting a predicted value; subtracting means for outputting the difference between the input signal and the predicted value sent from the predicting means; quantizing means for periodically sampling and quantizing the output of the subtracting means; A/D conversion means for performing A/D conversion is provided for each channel, and the prediction means for each channel converts the predicted value for the input signal of the current sample point of the self channel into one of the current sample points of the self channel. the input signal of the own channel at the previous sample point and the input signal of at least one other channel at the sample point immediately before the current sample point of the self channel; A multi-channel signal processing method characterized in that sample points of other channels are sequentially delayed with respect to sample points of the own channel.