基于LPC预测误差DWT的语音基音检测

Pitch Detection of Speech Signals Based on LPC Predictive Error DWT

在对语音信号进行ＬＰＣ分析的基础上，提出了一种有效的语音基音周期检测算法。该算法利用小波变换中著名的Ｍａｌｌａｔ算法逐层分解ＬＰＣ预测误差信号，在最低分辨率的逼近信号中寻找峰值，然后逐层回溯各个分辨率的逼近信号，最后在ＬＰＣ预测误差中确定出峰值，从而求出相应的基音周期。

Based on the LPC analysis of speech signals an effective method for the detection of thepitch is proposed.The LPC predictive error signal is decomposed layer by layer by the well-known Mallat algorithm in wavelet transformation and the peaks are searched for in the approximation signals of the lowest resolution. The relevant pitch period can be found by thepeaks determined with the LPC predictive error after all the approximation signals have beentracked backward layer by layer,