JPH0690636B2 - Speech coder - Google Patents

Description

The present invention relates to a speech coder used for speech band compression, speech accumulation, and the like.

[Conventional technology]

With the development and diversification of data networks in recent years, voice band compression technology is required to have a low bit rate for the purpose of reducing line cost and improving network efficiency.

The conventional high-efficiency coding methods for audio signals near 16 Kbit / sec include AD-PCM (adaptive differential PCM), APC (adaptive predictive coding), and APC.
-AB (adaptive prediction adaptive bit allocation coding), ATC (adaptive orthogonal transform coding), MPEC (multi-pulse drive coding), etc. have been proposed and put to practical use. (For example, Ochiai / Araseki, “Voice coding technology”, Journal of Information Processing Society of Japan, Vol.24No.8, Au
g.1983) Among these speech coding methods, in APC, APC-AB, etc., the prediction amount is increased by using a high-order pitch predictor that uses the pitch period of the speech, and high quality speech is obtained. ， Promising

As a method of extracting this pitch period, conventionally, a method has been used in which the autocorrelation function in the split frame is obtained and the maximum value of this autocorrelation function is used as the pitch period.

[Problems to be solved by the invention]

When implementing the above-described conventional pitch period extraction method using an autocorrelation function, the problem is the amount of calculation. The high-order pitch predictor uses a method of obtaining a prediction signal at a certain point from an input signal one pitch past, and therefore requires a long folding frame and a large amount of calculation. Here, the divided frame length is N, the order of the autocorrelation function is 1 to m, and the input speech signal is x (n) (n
= 1 to N), the autocorrelation function φ (n) is And the pitch period Tp ₁ is Is. Therefore, the calculation amount for extracting the pitch period Tp ₁ is the sum of products m (2N−2m) / 2 times, and the maximum value search number m
As a result, a large amount of calculation is required, which requires a large amount of calculation processing time or a large-scale calculation processing circuit. on the other hand,
When the split frame is made short in order to reduce the amount of calculation, the larger the n of φ (n) is, the smaller the number of product sums becomes, and the error becomes large, so that an accurate pitch cannot be extracted. It has the drawback of being lost.

[Means for solving problems]

In order to reduce the above-mentioned drawbacks, the speech coder of the present invention is characterized by having a pitch period extractor composed of an autocovariance function calculator, a weighting multiplier, and a maximum value searcher.

〔Example〕

 Next, the present invention will be described with reference to the drawings.

FIG. 3 is a waveform diagram for explaining the operation of the present invention.
As shown in the figure, the pitch period in section l of length N
Consider the case of extracting Tp ₂ . The position of the split frame is set to the position of k in the figure, and the autocovariance function γ (n) is And further weighted autocovariance function γw (n)
, Ask as. The Tp ₂ is Can be asked as

The pitch period extractor configured in this way will be described with reference to the block diagram of FIG. The input voice signal enters from the voice signal input terminal 1 and is guided to the autocovariance function calculator 21.
The obtained autocovariance function 23 is led to the weighting multiplier 24, and the weighted autocovariance function 25, which is the output signal of the weighting multiplier 24, is led to the maximum value searcher 26. Maximum value searcher
The output signal of 26 is the pitch cycle output terminal 27
Is output to.

An embodiment of a speech coder using this pitch period extractor will be described with reference to the block diagram of FIG. The input voice signal enters from the voice signal input terminal 1, is guided to the pitch period extractor 2 and the higher order pitch predictor 3, and the output of the pitch period extractor 2 is guided to the higher order pitch predictor 3. Higher-order pitch predictor 3
Is output to the proximity predictor 4, and the output of the proximity predictor 4 is input to the quantizer 5. Further, the output of the quantizer 5 is output to the encoded signal output terminal 6.

〔The invention's effect〕

As described above, according to the present invention, the pitch period extractor in the speech coder having the higher-order pitch predictor is composed of the autocovariance function calculator, the weighting multiplier, and the maximum value searcher. Pitch cycle extraction using a folding frame can be realized with a small amount of calculation and a relatively small error, and there is an effect that the processing delay time can be shortened and the arithmetic circuit scale can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a block diagram of an essential part of the present invention, and FIG. 3 is a waveform diagram of an audio signal for explaining the operation of the present invention. In the figure, l-1, l, l + 1 ... Partial section number, k ... Pitch-cycle folding frame section.

Claims (1)

[Claims] 1. A speech coder having a high-order pitch predictor to which an input speech signal is input, comprising a pitch period extractor for extracting a pitch period for each analysis frame and outputting it to the high-order pitch predictor. , The pitch period extractor, an autocovariance function calculator for calculating a plurality of autocovariance functions in the analysis frame from the input speech signal, and a weighted autocovariance function for weighting the obtained autocovariance function And a maximum value searcher that extracts a maximum value as the pitch period from the weighted autocovariance function.