语音信号基音检测的现状及展望

基音检测是语音处理中的一个非常重要的问题，但由于影响基音检测的因素众多．使得基音周期的准确估计非常困难。文中阐述了短时自相关函数法、短时平均幅度差函数法、倒谱法、小波变换法等几种经典的基音检测方法．分析它们各自的优点及存在的不足，并在预处理．后处理、语音信号的产生模型、语音信号的个性特征、发音时的情感及力度等基音检测的各个环节上提出了一些看法，并就一些可能出现的突破口做了一些展望。

The Current Situation and Prospects of Pitch Detection

The Pitch detection is important in speech signal procession. But it is difficult in detecting pitch accurately because many factors affect the detection. In this paper, several classical methods of pitch detection are proposed with analysis of their strong points and weak points, they are short time autocorrelation function, short time average magnitude difference function, cepstrum and wavelet transform. Also some peculiar views are presented on pro-procession, post-procession, model of speech signal, characters, emotions and physical power, and finally some possible breakthroughs are expected.