一种多脉冲激励线性预测声码器的仿真实现

A Vocoder of Multi-Pulse Excited Linear Prediction and Its Realization of Simulation

本文是笔者所开发的语音信号处理课程中的一个研究型实验教学案例。文中概述了多脉冲激励线性预测声器的编码方法,并用M ATLAB做了仿真实现。在本声码器实现方法中,在发送端对系统参数进行量化时,用量化LPC滤波器的PARCOR系数(部分相关系数,PAR t COR relation)代替传统的对滤波器的线性预测系数的量化;在接收端,将PARCOR系数用文中所给的算法转换为线性滤波器的线性预测系数,进行系统重构合成语音,大大降低了传输比特率,并且提高了合成语音质量。

This paper is a research experiment teaching case that the author develops in the course of speech processing. In this paper the author summarizes the coding ways for Multi-Pulse Excited Linear Prediction. In the vocoder realization way, the sender digitizes part correlation （PARCOR） coefficient rather than linear predictive coefficient of LPC filter. The receiver transforms PARCOR coefficient into linear predictive coefficient, and rebuilds system and synthesizes speech. In this way~ the transfer bit rate is reduced and the qualities of synthesized speech is improved greatly.