基于小波变换的语音基音周期检测

Speech Pitch Period Detection Based on Wavelet Transform

本文根据语音信号在声门闭合时刻（ＧＣＩ）的锐变特性，利用小波变换的时频局部化性质，构造了一个基音检测系统．通过使用不同的压扩因子，根据语音信号小波变换的局部最大值的位置，确定出对应的语音信号的锐变时刻（即ＧＣＩ），而相邻两个ＧＣＩ的间隔即为基音周期．本方法具有准确性高、抗嗓声性能好、检测范围宽、动态性能好等传统方法所不具备的特点．

ccording to the speech signal sharp variations caused by glottal closures, a system of pitchperiod detection is constructed based on the time-frequency domain localization of wavelet transform,and detect the local maxima POSitions of wavelet transform of speech signals are detected across severalsuccessive dyadic scales, the POsitions which indicate sharp variation point of speech signal being justglottal closure instants. So the interval of two adjacent glottal closure instants is jUst pitch period of thespeech segment. The method of pitch detection is superior to classical ones in measurement accuracy,robustness to noise, detection range and dynamic performance etc.