基于半波差分谱的语音信号音节切分  

Segmentation of speech signal based on the half-wave differential spectrum

作　　者：张红[1] 黄泰翼[1] 李治[2] 

机构地区：[1]中国科学院自动化研究所,北京100080 [2]西南交通大学,成都610031

出　　处：《声学学报》2000年第4期323-328,共6页Acta Acustica

摘　　要：根据听觉感知系统对于动态语音特征的响应特点，提出了半波差分港的概念，在此基础上提出了一种简便的音节切分算法，可以将存在大量协同发音现象的连续语音切分成一个个音节。在连续语音上的切分实验结果说明这种切分算法是准确而可靠的．This paPer proposes the concept of halfwav differential spectrum based on the study of the response styleof the auditory system to the dynamic part of the speech spectrum. Then a simple syllab le-segment at ion algorit hm ispresellted, which can segment the colltinuous speech oontaining many co-articulation ilito single syllable. The syllable-segmelltation experiments on the continuous speech show that the algorithm is accurate and reliable.

关 键 词：半波差分谱 语音信号 音节切分 语音识别 

分 类 号：TN912.3[电子电信—通信与信息系统]