JPS63127300A - Voice silence compression system - 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] Goyi public disaster The present invention relates to a voice silence compression method.

1 digit Silence compression is one of the audio compression methods.

This is the silent interval between spoken words and between words.

Alternatively, it is a method that utilizes redundancy in sections with small amplitudes. When a silent section is detected, the output section outputs O (zero) V (volt) to compress the data during that period. However, when a small amplitude section within a word is compressed and an OV is output, discontinuous points occur at both ends of the small amplitude section, and the sounds within the word are heard as interrupted.

Purpose The present invention was made in view of the above-mentioned circumstances.
In particular, when a speech waveform synthesizer performs silence compression within a word and plays it back, by linearly approximating both ends of a small amplitude section, it is possible to prevent discontinuities from occurring, thereby producing more natural speech. It was made for the purpose of reproduction.

In order to achieve the above object, the present invention relates to an audio silence compression method for compressing and reproducing audio silence.

When performing silence compression and reproduction, a linear approximation is performed from data at both ends of the silence section and output. An explanation will be given below based on one embodiment of the present invention.

FIG. 4 is a diagram showing the envelope of the audio waveform.

The figure (a) shows the original sound, and the figure (b) shows the silence compression between words.

FIG. 5 is a diagram showing a speech waveform to explain the case of compressing a section with a small amplitude within a word. (a) shows the original sound, and between A and B is a compressed section with a small amplitude within a word. It shows. Figure (b) shows that for the original sound shown in figure (a), the output of the compression section is Ov, and both points A and B are discontinuous.

The present invention was made in order to prevent the occurrence of discontinuous points as described above, and as shown in Figure (C), a linear approximation is performed between them based on the value of point A and the value of point B. Or, as shown in figure (d), noise is superimposed on the linear approximation shown in figure (c) S to give it a natural look.

FIG. 1 is a block diagram for explaining an embodiment of the present invention (a concrete example of FIG. 5(c)), in which (a) is a block diagram of encoding, and (b) is a block diagram of encoding. This is a block diagram of decoding. In figure (a), 1 beam, P, F (low pass filter), 2 A/D converter, 3 delay memory (De
4 is a silence detection circuit, 5 is an encoder, and this encoding method is the same as the conventional silence compression method. The silence detection circuit 4 can use various methods, for example, one block consists of eight consecutive samples.

When the maximum data in a block is smaller than a predetermined value, it can be defined as a silent section. When detecting silence, the encoder 5 sends only a flag indicating the length of the silent section. In addition, in cases other than silence, the output data of the A/D conversion circuit 2 is sent out or subjected to waveform compression such as DPCM or ADPCM, and encoded data is sent out. The delay memory 3 is used to detect silence and then send data from the A/D conversion circuit to the encoder.

(b) is a block diagram of decoding, in which 10 is a decoder and 11 is a delay memory (DelayMemor).
y), 12 is a silent section/small amplitude section determination circuit, 13 is a linear approximation circuit, 14 is a selector, 15 is an "O" setting circuit,
Reference numeral 16 indicates a D/A conversion circuit, 17 indicates P and F (low pass filters), and the decoder 10 expands the compressed waveform data and separates the silence flag. The silent section/small amplitude section determining circuit 12 performs linear approximation as a small amplitude section when the length of the silent section is smaller than a predetermined value (for example, 4 blocks), and when it is greater than the predetermined value,
A signal is issued to switch the selector 14 so as to output the value of Ov as the silent interval between words. The delay memory 11 is capable of outputting data with different delays in order to obtain values at both ends of the small amplitude section for linear approximation. The linear approximation circuit 13 outputs the value of each point in the small amplitude section from the data at both ends, and selector 1
4 switches the valid output, and D/A conversion circuit 16,L.

It is output through P and F 17.

FIG. 2 is a block diagram of decoding for explaining another embodiment of the present invention (the embodiment shown in FIG. 5(d)).
18 is a full adder circuit, 19 is a sampling frequency oscillation circuit, 20 is a counter, 21 is a ROM, and the others shown in Figure 1 (b)
) are given the same reference numerals as in FIG. 1(b). Then,
In this embodiment, noise is superimposed on the linear approximation interval, and the noise is generated by an oscillator with a sampling frequency (0
, S, C) 19 is input to a counter 20, and the output of this counter 20 is entered into the address of the ROM 21 and generated. Noise data is input into the ROM 21 in advance, and the output of the ROM 21 and the linear approximation circuit 1
Noise is superimposed by inputting the digital data of the outputs of 3 to the full adder 18.

FIG. 3 is a decoding block diagram for explaining a modified embodiment of the embodiment shown in FIG.
Parts which function similarly to the embodiment shown in the figures are provided with the same reference numerals as in FIG. 2. In this embodiment, noise is generated in an analog manner using elements such as operational amplifiers and diodes, and added to data after D/A conversion in an analog manner (using an OP amplifier or the like). When superimposing noise, the analog switch 23 on the noise generation side and the switch 24 on the ONL ground side are turned off. When noise is not superimposed, 0V is added by reversing ON and OFF to cut the noise.

Effects As is clear from the above explanation, according to the present invention, more natural speech can be reproduced by linear approximation and noise superimposition with the same amount of transmission data as in conventional silence compression. Further, there are also advantages such as a small increase in the amount of hardware.

[Brief explanation of the drawing]

1 to 3 are block diagrams for explaining the present invention in detail, FIG. 4 is a diagram showing the envelope of the audio waveform, and FIG. 5 is for explaining the operating principle of the present invention. FIG. 1...L, P, F, 2...A/D converter, 3...
Delay memory, 4... Silence detection circuit, 5... Encoder. 10... Decoder, 11... Delay memory, 12
... Silent section/small amplitude section judgment circuit, 13... Linear approximation circuit, 14... Selector, 15... 110'''
Setting circuit. 16...D/A conversion circuit, 17...L, P, F, 1
8... Full adder circuit, 19... Sampling frequency oscillation circuit. 20... Counter, 21... ROM, 22... Noise generation circuit, 23.24... Analog switch. Tomi 3 figure 8; ¥4 figure broom 5 figure

Claims (4)

[Claims] (1) An audio silence compression method for compressing and reproducing audio silence, which is characterized in that when performing silence compression and playback, linear approximation is performed from data at both ends of a silence section and output. . (2) If the silent section is long, do not perform linear approximation and
Claim No. (1) characterized in that V is output.
The audio silence compression method described in . (3) The speech silence compression method according to claim (1) or (2), characterized in that noise is superimposed on the linear approximation section. (4) The speech silence compression method according to claim (3), characterized in that a noise PCM code is generated from a ROM.