摘要作为最有效的语音分析技术之一，线性预测是一种基于全极点模型假定和均方预测误差最小准则下的波形逼近技术，通过对音频信号的时域和频域分析对声道参数进行估值，以少量低信息率的时变参数精确地描述语音波形及其频谱的性质，能够在尽量保持原音质的基础上合成出高质量的语音。本文重点研究了线性预测（LPC）的原理、Levinson-Durbin算法，基于MATLAB进行语音线性预测仿真，并对参数的选取做了比较分析。关键词语音信号LPCLevinson-Durbin算法MATLAB仿真AbstractAsoneofthemosteffectivespeechanalysistechnique,linearpredictionisakindofbasedonthepolemodelassumptionandminimummeansquareerrorcriterionofwaveapproximationtechnique,basedontheaudiosignaloftimedomainandfrequencydomainanalysisoftrackparametersvaluations,withafewlowinformationrateoftime-varyingparametersaccuratelydescribethenatureofthespeechwaveformandspectrum,tobeabletotrytokeeptheoriginalsoundqualityonthebasisofthesynthesisofhighqualityspeech.Thispapermainlystudiesthelinearprediction(LPC),Levinson-fromtheprincipleofthealgorithm,basedonMATLABspeechlinearpredictionsimulation,andtheselectionofparameterstodoacomparativeanalysis.KeywordsvoicesignalLPCLevinson-DurbinalgorithmMATLABsimulation目录TOC\o"1-3"\h\z\uHYPERLINK\l"_Toc264541204"第1章绪论PAGEREF_Toc264541204\h1HYPERLINK\l"_Toc264541205"1.1语音信号LPC分析技术的基本概念PAGEREF_Toc264541205\h1HYPERLINK\l"_Toc264541209"第2章线性预测编码的基本原理2HYPERLINK\l"_Toc264541210"2.1语音信号的产生2HYPERLINK\l"_Toc264541214"2.2线性预测的概念与原理2HYPERLINK\l"_Toc264541221"2.2.1线性预测分析的概念2HYPERLINK\l"_Toc264541221"2.2.2LPC和语音信号模型的关系4HYPERLINK\l"_Toc264541219"2.3解线性预测参数方程组的算法5HYPERLINK\l"_Toc264541221"2.3.1Levinson-Durbin自相关解法5HYPERLINK\l"_Toc264541222"2.3.2利用格型法求解线性预测系数6HYPERLINK\l"_Toc264541224"第3章simulink仿真的分析合成系统10HYPERLINK\l"_Toc264541226"3.1仿真内容10HYPERLINK\l"_Toc264541231"3.2仿真系统模型10HYPERLINK\l"_Toc264541231"3.3仿真工作过程11HYPERLINK\l"_Toc264541221"3.3.1语音信号采样11HYPERLINK\l"_Toc264541221"3.3.2预加重11HYPERLINK\l"_Toc264541221"3.3.3叠接窗分析13HYPERLINK\l"_Toc264541221"3.3.4汉明窗14HYPERLINK\l"_Toc264541221"3.3.5自相关算法14HYPERLINK\l"_Toc264541221"3.3.6数字滤波器16HYPERLINK\l"_Toc264541231"3.4仿真结果分析18HYPERLINK\l"_Toc264541234"参考文献1