Gain quanitzation method in analysis-by-synthesis linear predictive speech coding

Abstract

A gain quantization method, in analysis-by-synthesis linear predictive speech coding, includes these steps: determine a first gain for an optimal excitation vector from a first code book; quantize the first gain; determine an optimal second gain for an optimal excitation vector from a second code book; determine a linear prediction of the logarithm of the second gain from the quantized first gain; and quantize the difference between the logarithm of the second gain and the linear prediction.

Description

Technical field [0001] The invention relates to a gain quantization method in comprehensive analysis of linear prediction speech coding, in particular to a gain quantization method used on mobile phones. Background technique [0002] Comprehensive analysis linear prediction speech encoder usually has a long-term prediction program or adaptive codebook, and has several fixed codebooks. For example, this kind of speech encoder is described in reference (1) attached to the specification. In such a speech encoder, the total excitation vector can be described by a linear combination of codebook vectors vi, that is, each codebook vector vi is multiplied by a corresponding gain. Search the codebook sequentially. Usually the excitation from the first codebook is subtracted from the target signal (speech signal) before searching for the second codebook. Another way is orthogonal search, where all vectors in the latter codebook are orthogonal to the selected codebook v...

AI Technical Summary

Problems solved by technology

[0010] Gain ratios as described in Ref. (2) are insensitive to channel errors and are less complex. This approach requires different quantizers to be trained, which may make the characteristics of the input signal different from those in The encoder does not perform well when the features of the signal used in training

This method is also very impractical

Images