基于线性预测编码与AMDF的高精度基音检测算法.doc

《基于线性预测编码与AMDF的高精度基音检测算法.doc》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、基于线性预测编码与AMDF的高精度基音检测算法摘要:根据语音信号产生原理，结合线性预测编码(lpc)与平均幅度差函数法(amdf)，提出了一种高精度的基音检测算法。该算法首先利用线性预测分析提取残差信号；然后采用累积平均归一化差分函数与差分信号修正，使基音周期的谷值点更加尖锐；最后利用二次函数拟合与基音周期的倍数检查筛选候选值，得到了准确的基音周期。实验结果表明，与传统方法相比,该算法的基音检测效果有了明显改善，减少了基音检测中的半频错误，在高信噪比下具有良好的准确性和鲁棒性。关键词:语音信号；基音周期；线性预测编码；平均幅度差函数；自相关函数superresolu

2、tionpitchdetectionbasedonlpcandamdfwangen.cheng1,suteng.fang1*,yuankai.guo2,wuchun.hua21.schoolofinformationengineering,northchinauniversityoftechnology,beijing100144,china;2.schoolofcomputer,beijinguniversityofpostsandtelecommunications,beijing100876,chinaabstract:accordingtothe

3、mechanismofspeechsignal,asuperresolutionpitchdetectionalgorithmwhichcombinedlinearpredictivecoding(lpc)withaveragemagnitudedifferencefunction(amdf)wasproposed.firstly,residualoflpcwasextractedbylinearpredictiveanalysis.then,cumulativemeannormalizeddifferencefunctionanddifferencesignalrev

4、isionwereusedtomakepitchvalleysharper.atlast,parabolicinterpolationandpitchmultiplecheckweretakentoselectrealpitchperiod.experimentalresultsindicatethatthepitchdetectioneffectofthealgorithmissuperiortothatoftheconventionalalgorithms.theproposedalgorithmconquershalffrequencyerrors,andhasg

5、oodaccuracyandrobustnessundertheconditionofhighsnr.accordingtothemechanismofspeechsignal,asuperresolutionpitchdetectionalgorithm,whichcombinedlinearpredictivecoding(lpc)withaveragemagnitudedifferencefunction(amdf),wasproposed.firstly,residualoflpcwasextractedbylinearpredictiveanalysis.th

6、en,cumulativemeannormalizeddifferencefunctionanddifferencesignalrevisionwereusedtomakepitchvalleysharper.atlast,parabolicinterpolationandpitchmultiplecheckweretakentoselectrealpitchperiod.theexperimentalresultsindicatethatthepitchdetectioneffectofthealgorithmissuperiortothatoftheconventi

7、onalalgorithms.theproposedalgorithmconquershalffrequencyerrors,andhasgoodaccuracyandrobustnessundertheconditionofhighsignal.to.noiseratio(snr).keywords:speechsignal;pitchperiod;linearpredictivecoding(lpc);averagemagnitudedifferencefunction(amdf);autocorrelationfunction(a