应用矢量量优的孤立词汉语语音识别

Isolated—word Chinese speech recoginition usirg vector quantization

本文提出了一种识别孤立词汉语语音的新方法,提取线性预测系数作为语音特征,利用矢量量化的聚类特性压缩数据,用多段码书作为语音样板,用最小失真法进行识别。

This paper gives a new method to recognize isolated-word chinese speech, the features of speech are extracted as linear predictive coefficients,The data are compressed by clustering character of vector quantization,the templets of speeches are represnted by multisection codebook.The words are recognized by a method called “the smallest distortion”.