首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于DWT变换的语音端点检测
引用本文:滕旭东,袁晓,赵元英,魏永豪.基于DWT变换的语音端点检测[J].四川大学学报(自然科学版),2005,42(5):942-945.
作者姓名:滕旭东  袁晓  赵元英  魏永豪
作者单位:1. 上海工程技术大学电子电气工程学院,上海,201600;四川大学电子信息学院,成都,610064
2. 四川大学电子信息学院,成都,610064
摘    要:根据语音特性,提出一种基于离散子波变换(DWT)的语音端点检测方法.该方法利用(数字)语音信号的多尺度能量分布特性和不同分辨率下重构语音信号的相关特性来刻画(数字)语音信号,从而在噪声中检测出语音信号的端点.仿真试验结果表明:基于DWT变换的端点检测法与常用的端点检测方法相比,具有更好的抗噪性和识别稳定性.

关 键 词:多分辨率分析  互相关函数  能量分析  标准差
文章编号:0490-6756(2005)05-0942-04
收稿时间:2004-08-10
修稿时间:2004-08-10

Speech Endpoint Identification Based on DWT
TENG Xu-dong,YUAN Xiao,ZHAO Yuan-ying,WEI Yong-hao.Speech Endpoint Identification Based on DWT[J].Journal of Sichuan University (Natural Science Edition),2005,42(5):942-945.
Authors:TENG Xu-dong  YUAN Xiao  ZHAO Yuan-ying  WEI Yong-hao
Abstract:A detection method of speech start-end points is presented. In terms of speech properties, the method is based on multi-scales energy distribution and the cross-correlation function of the reconstruction digital speech signal in the different scales. Therefore, authors can keep track with the energy change along speech duration to check start-end points. Comparison with classical endpoint detection are given to show how the way performs under different signal-noise ratios. The experimental results show the method is more stability and immunity to noise.
Keywords:multiresolution analysis  cross-correlation function  energy distribution  standard deviation
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号