JPS63127630A - Voice compression processing unit - Google Patents

Abstract

PURPOSE:To reduce the processing amount remarkably as the entire processing unit by sacrificing more or less the quality of a voice with less sound volume and then in a direction toward lower priority in comparison with the application of voice compression/expansion processing with high quality in both transmission/reception. CONSTITUTION:A data inputted from a voice input terminal 1 is converted by a voice input 2, subject to band compression by a voice compression processing unit 3 and coded by a coder 4 and sent to a transmission line from an output terminal 5. Moreover, a data input from the transmission line to an input terminal 6 is decoded into a compressed data by a decoder 7, expanded by the voice compression processing unit 3, a voice signal is recovered by a voice output device 8 and outputted from a voice output terminal 9. A comparator 10 compares a parameter representing the sound volume of the transmission data from the transmission line with a parameter representing the sound volume of the reception data and selects a compressor 31 with high quality and an expander 33 with low quality when the sound volume of the transmission data is larger than the sound volume of the reception data. When the sound volume of the sent data is smaller than the sound volume of the received data, a compressor 32 with low quality and an expander 34 with high quality are selected.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a band compression processing device for audio data, and in particular,
The present invention relates to an audio compression processing device that aims to reduce the processing time of compression processing that requires large processing scale.

[Conventional technology]

Conventionally, a device that effectively utilizes a voice band in which there is no input is described in Japanese Patent Laid-Open No. 128028/1983. However, this focuses on the effective use of the transmission path, and does not mention the effective use of the capabilities of the transmission/reception processing device. In addition, this method can only be applied when there is no audio input.
No consideration is given to cases where low-volume audio is being input.

It also monitors the electrical information parameters of the received signal,
A device for controlling decoded audio is JP-A-60-1.
There is 73600. However, this is only for suppressing errors in transmission parameters, and does not take into account the processing amount of the audio processing device.

[Problem that the invention seeks to solve]

The above conventional technology does not consider reducing the processing amount of the audio compression processor, and therefore requires advanced processing to provide high-quality audio in both directions that exceeds the capabilities of the audio processor in two-way calls. It has not been possible to realize band compression processing with content.

The purpose of the present invention is to slightly sacrifice the quality of the audio in the direction where the volume is low and therefore the priority is low, compared to the case where high quality audio compression/expansion processing is performed in both the transmission and reception directions. An object of the present invention is to provide an audio compression processing device that can significantly reduce the processing amount of the processing device as a whole.

[Means for solving problems]

The above object is achieved by providing an audio compression processing device having the following configuration.

In other words, first, there are two types of compressors: a high-quality compressor that can provide relatively high-quality audio but takes a relatively long processing time, and a low-quality compressor that can only provide relatively low-quality audio but has an extremely short processing time. Prepare.

A specific configuration example will be described later. Either one of the two compressors is actually operating at a certain time, and one of them is selected based on the output of a comparator that compares the transmitted volume and the received volume on the transmission line.

In addition, as decompressors, we prepare a high-quality decompressor with relatively high quality but a long processing time, and a low-quality decompressor with low quality but a short processing time. Have them choose one.

By selecting the compressor and decompressor using the output of the same comparator, when a high quality compressor is selected, a low quality decompressor is always selected; The configuration is such that when a compressor is selected, a high-quality decompressor is always selected.

[For production]

The comparator compares a parameter representing the volume of transmitted data on the transmission path with a parameter representing the volume of received data.

If the volume of the transmitted data is greater than the volume of the received data, the comparator output selects a high quality compressor and a low quality decompressor.

As a result, high-quality compressed data is generated and transmitted at the own station, and relatively low-quality audio is reproduced on the receiving side. As for the compression processing board as a whole, the amount of processing can be reduced by the increased speed of the decompressor.

Conversely, if it is determined that the volume of the transmitted data is lower than the volume of the received data, a low quality compressor and a high quality decompressor are selected based on the comparator output. As a result, although relatively low-quality compressed data is transmitted from the local station, high-quality audio is reproduced. As for the compression processing board as a whole, the processing amount is reduced by the speed of the compressor.

As described above, compared to the case where high-quality audio compression and decompression processing is performed in both the transmission and reception directions, the audio is smaller, so compression is possible by sacrificing the quality of the audio in the direction with lower priority. The processing amount of the entire processing panel can be significantly reduced. In this case, in normal two-way calls, there is little chance that meaningful voice calls will collide in both directions, and even if they do, they will only be interrupted each other, so in most cases, quality deterioration in the direction with lower priority will not be a problem. .

〔Example〕

(Example 1) First, an example of a speech compression processor using linear prediction will be described. Linear prediction is described in detail in, for example, Kazuo Nakata, "Basics of Speech Information Processing", Ohmsha, 1981.

Figure 1 shows the overall configuration. Data input from the audio input terminal 1 is first input to the audio input device 2, where it is converted, and then band-compressed by the audio compression processor 3.

The compressed audio data is encoded by the encoder 4 in a form suitable for the transmission path, and sent from the compressed data output terminal 5 to the transmission path.

Further, the data received from the transmission line and input from the compressed data input terminal 6 is decompressed into compressed data by the decoder 7, decompressed by the audio compression processor 3, and transferred to the audio output device 8 for audio output. The audio signal is reproduced by the device 8 and output from the audio output terminal 9. 10 is a comparator that compares the volume parameter of the transmitted data and the volume parameter of the received data, and the output of this comparator 10 allows
The configuration is such that the processing content in the audio compression processor 3 changes. That is, the audio compression processor 3 includes a high-quality compressor 31 and a low-quality compressor 32 as compressors, and a low-quality decompressor 33 and a high-quality decompressor 34 as decompressors. vessel 31.32 and stretcher 33.34
According to the output of the comparator 10, when the volume of the transmitted data is larger than the volume of the received data as in the illustrated example, the contact a side becomes connected, and the high quality compressor 31 and the low quality decompressor 33
is now selected.

FIG. 2 shows a block diagram of the configuration of an audio compression processor implemented using linear prediction. Input audio data is usually divided into sections of fixed length called frames, a correlation coefficient is calculated for this section, and this correlation is used to determine linear prediction parameters. Two such linear predictors are prepared here: a high-order linear predictor 311 that operates at a high order, and a low-order linear predictor 312 that operates at a low order. Depending on the output from (not shown), either one is selected and operated. The inverse filter that predicts the speech within the same frame using the prediction parameters output from the predictor 311 or 312 and obtains the residual, which is the error from the actual speech, is also connected to the high-order inverse filter 321 and the low-order Two inverse filters 322 are prepared, and similarly, comparator 1
A zero output makes a selection. The residual calculated by the inverse filter 321 or 322 is:

High quality residual compressor 331 or low quality residual compressor 33
2, the band is compressed. As a method of f-range compression, R
There are ELP methods, APC-AB methods, etc., and here, for example, the residual compression method described in Japanese Patent Application No. 58-28726 will be used for explanation.

In this residual compression method, as shown in a specific waveform example in Figure 4, among the residuals in a frame, those with small amplitudes are thinned out, and the parts with the highest periodicity of the remaining residuals are thinned out. Bandwidth compression is performed by representing data for one representative period. Here, we use a high-quality residual compressor 331 that compresses the band as described above, and a low-quality residual compressor 331 that simply thins out small amplitudes and does not select representative parts, as shown in the example waveform in FIG. There is a container 332.

Similarly, two are prepared: a high-quality residual restorer 334 that decompresses compressed data with high quality and restores the residual, and a low-quality residual restorer 333 that restores the residual with low quality. The restored residual is used by a synthesis filter to predict speech using prediction parameters, and the residual is added to this to restore the speech.

Similar to the inverse filter, this synthesis filter has two types: a high-order synthesis filter 342 and a low-order synthesis filter 341.

Out-of-band noise is removed from this voice by an output filter 350, but without operating this output filter 350,
The output of the low-order synthesis filter 341 can also be sent as is to the audio output device 8 side.

In addition, since the residual compressors 331 and 332 use the repetition period of the audio signal, a pitch extractor 360 that calculates it
is also provided, but this period is not needed when the low quality residual compressor 332 is operating, so the pitch extractor 36
0 operation can also be stopped by the output of the comparator 10.

In the above configuration, when the comparator IO determines that the transmitted volume is higher than the received volume, each switch in FIG. 2 is connected to the a side. At this time, compressed audio data corresponding to high-quality audio is transmitted, but the received compressed data is sent to the low-quality residual expander 333 and the low-order synthesis filter 34.
1 to output audio. Here, first, the processing of the decompressor is performed by the low-quality residual decompressor 333, so 1. For example, when processing a frame of 160 8kHz samples using the residual compression method, the processing amount is reduced to about 1/10 of the total processing amount. is expected. Furthermore, in the low-order synthesis filter 341, if, for example, the 8th-order filter is suppressed to the 4th-order filter, a reduction in processing amount of about 1/10 can be expected here as well. Furthermore, by stopping the output filter 350, for example, if this filter 350 is about 2 stages of pi-Farad, 1
By using 6 or more, which can reduce the processing amount by about 1/20 to 1/10, it is possible to reduce the processing amount by 20% to 30% in total.

Next, when the comparator 10 determines that the received volume is higher than the transmitted volume, each switch in FIG. 2 is connected to the b side. This corresponds to a case where the transmitted compressed data is of low quality and the received compressed data is of high quality. First, the correlation order can be reduced by processing on the low-order linear predictor 312 side, so the 8kHz
For 320 points, if we reduce the 8th order correlation to 4th order,
It is possible to reduce the processing amount by about 1/7. Next, by reducing the order in the low-order inverse filter 322, the order is reduced from 1/20 to 1/20.
about 10, about 115 due to stopping the pitch extractor 360, and about 11 due to using the lower quality residual compressor 332.
A reduction in processing amount of about 5% is expected for each, and an overall reduction in processing amount of more than 50% is possible.

The above description has been made using the residual compression method, but the same applies to other methods such as RELP. Furthermore, a similar application is possible with a method that directly performs band compression without using linear prediction.

In this way, advanced band compression methods that exceed the processing capacity of the compression processor can reduce the overall amount of processing by sacrificing some of the sound quality in the direction of low volume, resulting in isometrically high-quality compression processing. You can get the equipment.

(Embodiment 2) Next, an example in which an echo suppressor is applied to the present invention will be described. When performing advanced compression processing, transmission delays tend to increase due to the large amount of processing involved.

When such a system is applied to a telephone set, the echo delay caused by imperfections in the hybrid transformer increases, resulting in a significant deterioration in speech quality. Therefore, an echo suppressor is used to suppress echoes by suppressing calls in low-volume directions. Regarding the echo suppressor, it is described in 1, for example, "Applications of Digital Signal Processing Technology J" edited by IEICE.In the echo suppressor, the means for detecting the direction of low sound volume can be shared with that in Embodiment 1. The block diagram is shown in the figure.The audio compression processor 3 is the same as that in the first embodiment, and 100 is an echo suppressor provided in this embodiment. A device is inserted to suppress echoes.It is also possible to replace this with a switch to cut off calls in low-volume directions.

According to this embodiment, the speech direction that is attenuated is always the direction that is compressed to a low quality. Therefore, the level of the voice restored to low quality is suppressed, and the voice is no longer unpleasant to the ears. Furthermore, if you do not completely block calls in the direction of low volume,
Since calls in this direction are never interrupted, the naturalness of the conversation is preserved.

〔Effect of the invention〕

As explained above, according to the present invention, a compressor and an expander are each provided with a high quality compressor and a compressor with low quality but short processing time, and a comparison between the transmitting volume and the receiving W volume is performed. Depending on the result of the determination, either a combination of a high-quality compressor and a low-quality decompressor or a combination of a low-quality compressor and a high-quality decompressor is selected. Compared to high-quality audio compression and decompression processing, the volume is lower and the quality of the audio in the lower priority direction is slightly lower, but the amount of processing by the compression processing equipment can be significantly reduced. , it becomes possible to achieve high-quality compression processing that cannot normally be achieved because it exceeds the processing capacity. In this case, in normal two-way calls, significant voice calls in both directions rarely collide, and even if the quality in the direction with lower priority deteriorates slightly, the call is preserved and greetings etc. are not transmitted. This will not cause any unnatural conversation.

[Brief Description of the Drawings] Fig. 1 is a diagram showing the basic configuration of the present invention, Fig. 2 is a block diagram of an embodiment in which the present invention is applied to a compression processing device using linear prediction, and Fig. 3 is a diagram showing the basic configuration of the present invention. FIG. 4 is a diagram showing an example of a waveform for high-quality compression/expansion of an audio signal, and FIG. 5 is a diagram showing an example of a waveform for low-quality/compression of an audio signal. Explanation of symbols 1...Audio input terminal 3...Audio compression processor 4
...Encoder 7...Decoder 10...
comparator

Claims (1)

[Claims] 1. Means for converting an input audio signal into a digital signal;
A compressor that removes redundancy from this digital signal and compresses the band; an encoder that encodes the band-compressed data and sends it to the transmission line; and a decoder that decodes the signal received from the transmission line. In an audio compression processing device that includes a decompressor that restores an audio signal by adding redundancy to band-compressed data from a decoder, the compressor is a high-quality compressor and the processing time is shorter than the decompressor. A compressor of low quality is provided as the decompressor, a decompressor of high quality and a decompressor of relatively low quality but with a shorter processing time are provided as the decompressor, and A switching means is provided for performing switching and switching between the high-quality expander and the low-quality expander in accordance with a comparison result between information representing the volume of the transmitted data on the transmission path and information representing the volume of the received data. An audio compression processing device characterized by: 2. The switching means selects a high quality compressor and a low quality decompressor when the information representing the volume of the transmitted data is higher than that of the received data, and when the information representing the volume of the transmitted data is higher than that of the received data. 2. The audio compression processing device according to claim 1, further comprising switching means for selecting a low quality compressor and a high quality decompressor when the output is small.