由透射波提取语音特征参数初探

研究了一维透射波提取语音参数问题．语音的物理本质是声波沿声道透射．声门是波源发生器，唇外是透射后的波，不同的声管形态产生不同的声音，从这些声波中提取不同语音的特征信息，是识别语音的一种途径．所提出的新的透射波提取语音参数公式，其理论基础是：波的传播可以由时变谱信号来刻划，表示成三元(即时间，空间和频率)复函数；介质参数(如反射系数等)均是空间和频率的复函数．利用小波变换实现时频转换后进行计算，再利用分类方法实现语音识别．

Pilot study of extracting voice feature parameters by transmitted wave

Physical quiddity of voice is sound wave transmitting along sound tube. Glottis is the source of sound wave and the outside of lip is transmitted wave. The voice is changed by different shape of voice channel. It is a way of identifying voice to pick up feature parameters from different sound wave. A kind of new expression has been put up which makes use of transmitted wave to pick up feature parameters. Its theory base is the transmission of wave can be depicted by time variant spectrum signal,which can be expressed by 3 dimensions complex function (time,space and frequency). Media parameters,such as reflectance etc.,are the complex function of space and frequency. The extracting voice parameter from 1 dimension transmitted wave is studied. The time signal is converted to frequency signal by wavelet conversion and then computing feature parameters. According to the feature parameters the different syllables can be identified by a kind of classifying method.