摘要
一.序言 语音信号压扩技术或语音编码技术主要可分为波形编码、声道编码和所谓的混合编码。在声道编码技术中,线性预测编码(LPC)技术是当今最强有力的编码技术之一,已成为估计诸如节距、共振峰及其宽度、频谱、声道截面参数等语音基本参数及语音低比特率通信和存储等数据压缩方法的主导技术,也成为语音识别技术的基础之一。对LPC技术的研究已发表了很多论文,基于LPC技术的低比特率通信用语音分析和语音合成装置也都已有样机。本文作者负责的课题组为乌鲁木齐车站研制的汉维语电话自动答询系统已运行二年。现在LPC技术也在诸如脑电图、地震波处理等领域获得了应用。 本文将介绍语音信号的LPC模型,LPC的有关理论和实践,详细讨论节距激励式LPC声码器。此外,作为对节距激励式LPC的改进,还将简单讨论多脉冲激励LPC和残差信号激励LPC技术。
Linear predictive coding (LPC) has now become the predominant coding technique in speech companding and the technique is used for estimating such fundamental speech parameters as pitch, formants, spectra, vocal tract area etc. , for low bit rate speech transmission and storage. It has also become one of the most important technique for speech identification.
A prototype established by a research group led by the author is now working well for analyzing and synthesizing Chinese and Uygur language at Urumqi Railway Station.
This paper introduces the LPC model,the theory and practice of LPC technique, discusses pitch excited LPC vocoder in detail.As the improvenments, multipulse excited LPC and residual excited LPC are briefly introduced as well.
