Vector quantization method for sorting and rearranging code book and vector quantizer thereof

Abstract

The invention discloses a vector quantization method for sorting and rearranging a code book and a vector quantizer thereof. The method comprises the following steps of: in an encoding process, inputting a residual vector to sort according to sorting information, determining a sub-code book search range in a sorted and rearranged code book in an encoder and determining a vector quantization value in a sub-code book; and in a decoding process, determining the reconstruction vector of the input residual vector in the sorted and rearranged code book of a decoder according to a quantization index sent to a decoder end by a channel. Before quantization, the code book is sorted previously by using a code book sorting algorithm, and the sorting information is stored in a sorter. The sorter, a sorted and rearranged code book storage unit and a quantization module in the encoder in the vector quantizer, a sorted and rearranged code book storage unit and a decoding module in the decoder are key components. In the invention, the sorter information is not necessary to be sent in a channel and does not occupy a quantization bit, the code word search range is limited in the sub-code book of the sorted and arranged code book in the encoding process, the time complexity of the vector quantizer is decreased, and the effect of transparent quantization is achieved.

Description

technical field [0001] The invention relates to a vector quantization method and a vector quantizer for voice signal processing, in particular to a vector quantization method and a vector quantizer for codebook classification rearrangement. Background technique [0002] The composition of the vector quantizer includes an encoder and a decoder, and its composition directly determines the storage requirements of the vector quantizer, the complexity of the codebook search algorithm, and the quality of the decoded voice; the codebook of the vector quantizer is stored in the encoder and decoder, The performance of the codebook and the arrangement of the codebook directly affect the quantization performance of the vector quantizer. The most commonly used vector quantizers in speech coding algorithms are split vector quantizers, multistage vector quantizers, split-mean multistage split vector quantizers S-MSVQ (Split-MultiStage Vector Quantiser) and classification vector quantizers...

AI Technical Summary

Problems solved by technology

[0010] The present invention solves the irregular arrangement of codewords in the codebook of the encoder and the codebook of the decoder in the vector quantizer under the condition that the number of quantization bits does not increase and the quantization spectrum is distorted to achieve a transparent quantization effect. The quantization vector corresponding to the residual vector, the code word search of the vector quantizer needs to adopt the method of exhaustive search, which causes the time complexity and space complexity of the vector quantizer and the quantization performance of the vector quantizer can not be well compromised, and provides a Codebook classification rearrangement vector quantization method and its vector quantizer

Images