In this paper, a novel waveform coding approach based on half-wave vector quantization is presented. The input speech signal is divided into frames, each of which is classified as silence, unvoiced and voiced. The voiced category is subdivided into 64 sub-categories according to the length of each half-wave vector. The unvoiced category is subdivided into 16 sub-categories according to the average zero-crossing rate of each frame. For each type of vectors, a set of codebook is generated by using training data. Then each kind of sub-category is encoded, transmitted, and decoded with its own particular bit allocation configuration. This variable bit coding approach can operate at medium rate and provide acceptable-to-good speech quality.
Cite as: Chen, X., Song, Y., Yu, T. (1999) A novel speech coding approach based on half-wave vector quantization *. Proc. 6th European Conference on Speech Communication and Technology (Eurospeech 1999), 1491-1494, doi: 10.21437/Eurospeech.1999-339
@inproceedings{chen99c_eurospeech,
author={Xiaoping Chen and Yantao Song and Tiecheng Yu},
title={{A novel speech coding approach based on half-wave vector quantization *}},
year=1999,
booktitle={Proc. 6th European Conference on Speech Communication and Technology (Eurospeech 1999)},
pages={1491--1494},
doi={10.21437/Eurospeech.1999-339}
}