JPH0266598A - Speech signal compressing and expanding device - Google Patents

Abstract

PURPOSE:To improve compressibility without deteriorating sound quality by regenerating a sample value of a waveform which is thinned out according to an interpolation polynomial of each representative waveform. CONSTITUTION:Representative waveforms A0 and A1 are extracted from the stationary part of the repetitive waveform of a speech signal, other waveforms are thinned out, and the number of samples in one cycle of the repetitive waveform of the speech signal is found. Then the speech signal is compressed, the interpolation polynomial of the waveform thinned out with the sample values of the extracted representative waveform is found, and the sample values of the thinned-out waveform are found to perform interpolation, thereby attaining expansion. Consequently, connection points of the expanded repetitive waveform continue regardless of the amplitude, pitch period, etc., of the representative waveform, so that the compressibility is improved without deteriorating the reproduced sound quality.

Description

[Detailed description of the invention] Industrial applications The present invention relates to an audio signal compression/expansion device.

BACKGROUND OF THE INVENTION FIG. 7 is a waveform diagram for explaining the operation of a conventional audio signal compression/expansion device.

As shown in Fig. 7(a), similar waveforms are usually repeated in the stationary part of the audio signal, so Fig. 7(b)
As shown in the figure, a representative waveform is cut out from this repetitive waveform and other similar waveforms are thinned out to compress the audio signal.

Then, as shown in FIG. 7(c), the thinned out waveform is expanded by repeatedly interpolating the cut out waveform.

Therefore, in the conventional example described above, similar waveforms in the stationary portion of the audio signal are compressed by thinning out, and the thinned out waveforms are repeatedly interpolated and expanded using representative waveforms.

Problems to be Solved by the Invention However, in the above-mentioned conventional audio signal compression/expansion device,
The thinned out waveform is expanded by simply repeating interpolation with the representative waveform, so at the connection point of the expanded repetitive waveform, the amplitude, pinch period (period of the repetitive waveform)
There is a problem in that discontinuities occur due to etc., and therefore abnormal noises occur in the reproduced sound.

As a result, there is a problem in that the compression ratio cannot be increased in order to prevent deterioration of the reproduced sound quality.

SUMMARY OF THE INVENTION In view of the above-mentioned conventional problems, it is an object of the present invention to provide an audio signal compression/expansion device that can increase the compression rate without deteriorating the reproduced sound quality.

Means for Solving the Problems In order to achieve the above object, the present invention extracts a representative waveform from the repetitive waveform of the stationary part of the audio signal, thins out other waveforms, and thins out the samples in one cycle of the repetitive waveform of the audio signal. The audio signal is compressed by calculating the number of samples, the interpolation polynomial of the decimated waveform is obtained from the sample value of the extracted representative waveform, and the sample value of the decimated waveform is calculated using this interpolation polynomial and the number of samples. It is designed to expand by finding and interpolating.

Effect: With the above configuration, the connection points of the expanded repetitive waveform are continuous regardless of the amplitude, pitch period (period of the repetitive waveform), etc. of the representative waveform, so the compression ratio can be increased without deteriorating the reproduced sound quality. It can be improved.

EXAMPLES Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the audio signal compression/expansion apparatus according to the present invention, FIG. 2 is a flowchart for explaining the operation of the audio signal compression/expansion apparatus of FIG. 1, and FIG. ,
A waveform diagram for explaining the operation of the voiced sound compressor shown in FIG.
FIG. 4 is a waveform diagram for explaining the operation of the voiced sound expander of FIG. 1, and FIGS. 6 and 6 are diagrams for explaining the operation of the voiced sound expander of FIG. 1, respectively.

The upper part of FIG. 1 shows an audio signal compression device, where 1 is a microphone into which audio is input, 2 is an A/D conversion circuit that converts an analog audio signal from microphone 1 into a digital audio signal, and 3 is an A/D converter. A voice/silence determiner 4 detects the voice/silence portion extracted by the voice/silence determiner 3 by determining whether the digital audio signal is a voice portion or a silent portion from the conversion circuit 2 and extracts the voice portion. A voiced/unvoiced sound determiner 5 determines between voiced and unvoiced sound parts and extracts the voiced sound part, and 5 is a voiced sound compressor that compresses the voiced sound part extracted by the voiced sound/unvoiced sound determiner 4 as described below. 6 encodes the silent part determined by the voiced/silent determiner 3, the unvoiced part determined by the voiced/unvoiced determiner 4, and the voiced part compressed by the voiced compressor 6, respectively. The encoder 7 is a memory for storing encoded data from the encoder 6.

The lower part of FIG. 1 shows an audio signal decompression device, 8 is a memory for storing encoded data, and 9 is a memory for storing encoded data, and 9 converts the encoded data read from the memory 8 into a silent part, an unvoiced part, and a voiced part. A decoder 10 decompresses the voiced part decoded by the decoder 9, and a voiced part decompressor 11 decompresses the voiced part decoded by the decoder 9 as will be described later. a D/A conversion circuit that converts the unvoiced sound part and the voiced sound part expanded by the voiced sound part expander 10 into an analog audio signal;
12 is a speaker that converts the audio signal from the D/A conversion circuit 11 into sound.

Next, referring to FIGS. 2(a) and 3, the voiced sound compressor 5
Explain the operation.

In step 21 shown in FIG. 2(a), the voiced sound compressor 5 uses a method such as an autocorrelation function to extract samples within one period of each repetitive waveform of the voiced sound part as shown in FIG. 3(a). Figure 3 (b
) are obtained, and in step 23, repetitive waveforms other than the representative waveforms AO, A, are thinned out.

Then, as the compressed data of this voiced part, the representative waveform A
The number of samples in one cycle of O, A I and thinned out waveforms, and the amplitude of each sample point of representative waveforms AO, At are stored in the memory 7.
is stored in

Next, the operation of the voiced sound expander 10 will be explained with reference to FIG. 2(b) and FIGS. 4 to 6.

In step 24 shown in FIG. 2(b), FIG. 4(a)
), the number of samples of the representative waveform A ON A I as shown in A ON A
1) AI "xl (OL xl (') r
"'r xl ("I') (Fig. 4 (
In a), the sample point x o (0) + x o ('
) *Only Xo(2) is shown), the voiced sound expander 10 defines an interpolation polynomial for each sample point X(0) to X·(g, −1).

For example, as shown in FIG. 6, this interpolation polynomial has the coordinates of the four points ao1als a2 and a3 as aO(1,xo) at(0,xt) a2(1,x2) a3(2,x3), The following conditions are met: (1) Pass through four consecutive points; (2) At each point, the first-order differential coefficients in the preceding and succeeding sections must match.

The following formula (% formula %) is obtained, and this formula is expressed as four consecutive sections xI (J
-1), xl(''), xl(''+1), X (J+2)
Find the interpolation polynomial by applying the polynomial to .

In the following step 25, representative waveforms A o, A 1
In order to match the number of samples in one cycle of the waveform to be interpolated, let the number of samples in one cycle of the second waveform between representative waveforms A O, A 1 be lk, and the representative waveforms Ao, A
Resampling is performed so that the number of samples of I becomes lk.

That is, sample points xH(0) to X of representative waveform Ao and AI
1(l, -1) is equally divided into (lk-1), and as shown in Figure 4 (
As shown in b), the respective amplitudes AO””xo’(
0) t xo'(1) l ”'r xO'(1
301) A1'-xl'(0), xl'(1), ・
..., find x1' (111).

Amplitude xl of the j-th sample point within one period of the interpolated waveform
(D is the representative waveform A, A, and the sampled j-th amplitude X, as shown in FIG. 6.'(D.

However, j=o, 1. ..., 1lk-10: Calculated using the number of interpolated waveforms k -12...n.

Therefore, according to the above embodiment, each representative waveform Ao, A
By reproducing the sample values of the thinned waveform using the interpolation polynomial of l, the thinned waveform can be smoothly interpolated, and therefore the compression ratio can be increased without deteriorating the reproduced sound quality. Can be done.

DETAILED DESCRIPTION OF THE INVENTION As described in detail, the present invention extracts a representative waveform from a stationary portion of a repetitive waveform of an audio signal, thins out other waveforms, and calculates the number of samples in one cycle of the repetitive waveform of an audio signal. The audio signal is compressed, the interpolation polynomial of the thinned out waveform is found using the extracted sample values of the representative waveform, and the sample values of the thinned out waveform are found and interpolated for expansion. The connection points of the expanded repetitive waveform are continuous regardless of the amplitude, pitch period (period of the repetitive waveform), etc. of the representative waveform, and therefore the compression ratio can be increased without deteriorating the reproduced sound quality.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the audio signal compression/expansion apparatus according to the present invention, FIG. 2 is a flowchart for explaining the operation of the audio signal compression/expansion apparatus of FIG. 1, and FIG.
6 is a waveform diagram for explaining the operation of the voiced sound compressor shown in FIG. 1, FIG. 4 is a waveform diagram for explaining the operation of the voiced sound expander shown in FIG. The figures are a diagram for explaining the operation of the voiced sound expander shown in FIG. 1, and FIG. 7 is a waveform diagram for explaining the operation of the conventional audio signal compression and expansion device. 5... Voiced sound compressor (representative waveform extraction, sample number calculation means), 6... Encoder, completion, 8... Memory, 9...
...Voiced sound expander (interpolation polynomial calculation, interpolation means). Name of agent: Patent attorney Shigetaka Awano and one other person if
! Figure 3 (cl) 14 period V] Puno Garden (b) Figure Figure (0 lid also Figure (C) 1 synchronization 4 ηL71L number supplementary Sonozawa Vaginal Maki n=riN 7 Figure (O figure) Figure C (Figure C)

Claims (1)

[Claims] means for extracting a representative waveform from the repetitive waveform of the stationary portion of the audio signal and thinning out other waveforms; means for determining the number of samples in one period of the repetitive waveform of the audio signal; and a sample value of the extracted representative waveform, An audio signal compression/expansion device comprising: means for determining an interpolation polynomial of a thinned waveform; and means for determining and interpolating sample values of the thinned waveform using the interpolation polynomial and the number of samples.