语音识别中的特征预检测用模拟电路提取过零率和包络

On the Pre-eigen Detection in Speech Recognition

本文讨论了一种具有实用价值的提取时域语音特征的硬件方法.使用该方法不仅能节省计算机内存,而且还能做到实时处理.借助于一种简单的硬件线路,就可以准确、可靠地提取出语音的过零率和包络.该法可以用于计算机语音识别及汉语语音分析中辅/元音的判别.

This paper gives a practical method with hardware of obtaining the time-domain features of speech, by which not only is the memory saved, but also the real time treatment is carried out in the computer.With the aid of a simple hardware circuit,it is possible to obtain the zero-crossing rate and the envelop of speech exactly and reliably. And this method can also be applied to the speech recognition and speech analysis of Han language in the computer to distinguish between consonants and vowels.