| 基于二阶统计量的多径混合语音分离算法 |
| |
| 引用本文： | 金辉,陈晓曙. 基于二阶统计量的多径混合语音分离算法[J]. 现代电子技术, 2005, 28(24): 43-45,48 |
| |
| 作者姓名： | 金辉 陈晓曙 |
| |
| 作者单位： | 东南大学,江苏,南京,210096 |
| |
| 摘 要： | 介绍了一种实用的语音分离算法CoBliSS,他是一种基于二阶统计量的多通道盲反卷积算法.但是源信号自相关函数的序列长度必须足够大,而大多数语音信号不满足这一要求.但是针对非平稳的语音信号,非严格的理论分析显示二阶统计量可以用于语音分离,将CoBliSS算法与回声消除相结合,给出了ECoBliSS算法.结果表明当ECoBliSS算法用于远程会议系统时,分离出了较高质量的语音信号,而且速度也比较快.
|
| 关 键 词： | 二阶统计量 盲反卷积 语音分离 CoBliss算法 |
| 文章编号： | 1004-373X（2005）24-043-03 |
| 收稿时间： | 2005-08-29 |
| 修稿时间： | 2005-08-29 |
Multipath Mixture Voice Separate Algorithm Based on Second Order Statistics |
| |
| JIN Hui,CHEN Xiaoshu. Multipath Mixture Voice Separate Algorithm Based on Second Order Statistics[J]. Modern Electronic Technique, 2005, 28(24): 43-45,48 |
| |
| Authors: | JIN Hui CHEN Xiaoshu |
| |
| Affiliation: | Southeast University, Nanjing, 210096,China |
| |
| Abstract: | This paper first presents an applicable voice separation algorithmCoBliSS.It is a multichannel blind deconvolution algrithm based on Second Order Statistics(SOS).However,the length of the source correlated function sequence must be large enough,and most of the sound signal can not meet this need.But as far as nonstationary sound signal is concerned,nonrigorous theoretical analysis has shown that SS can be used to separate sound.We combine CoBliSS and echo elimination and get the ECoBliss algorithm.Results have shown that when EBlBlisis algorithm is used in longdistance conference systems,it can separate highquality sound signals,and also in a fast speed. |
| |
| Keywords: | second order statistics blind deconvolution voice separation CoBliss algonithm |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
