有限状态矢量量化相关编码方案

Finite-State Vector Quantization with Correlated Encoding

本文提出了有限状态矢量量化相关编码方案。分析了该方案对线性相关高斯源的率失真性能,并将分析结果与相关编码理论界及未考虑信源间相关性的编码结果作了比较。将该方案实际应用于分析合成自适应预测语音编码系统,它有效地利用了声道参数码矢和激励参数码矢之间的冗余信息。

A finite-state vector quantization correlated encoding (FSVQ-CEC) strategy is presented in this paper. The rate-distortion performance of FSVQ-CEC for linear-correlated Gaussian observations is demonstrated, and is compared with the optimum performance theoretically attainable of correlated encoding and the performance by simple encoding which fails to capitalize on the correlation between sources. The FSVQ-CEC scheme is applied to an analysis-synthesis adaptive predictive speech encoding system so as to exploit the redundancy between vocal tract codevectors and excitational codevectors.