结合ICA预处理的麦克风阵列语音增强系统  

Microphone Array Speech Enhancement Systems with ICA Preprocessing

作　　者：杜军[1,2] 刘琚[1] 李万龙[1] 

机构地区：[1]山东大学信息科学与工程学院,济南250100 [2]山东师范大学传播学院,济南250014

出　　处：《数据采集与处理》2009年第5期588-593,共6页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(60572105)资助项目;新世纪优秀人才支持计划(NCET-05-0582)资助项目;教育部留学回国人员科研启动基金([2005]55)资助项目

摘　　要：在强背景噪声和强反射环境中,麦克风阵元接收的信号质量很差,从而影响麦克风阵列语音增强系统的性能。ICA能够仅从观测信号中提取出潜在的独立成分,基于此特性,本文将ICA引入麦克风阵列语音增强系统,利用ICA对麦克风阵元接收信号进行分析,从中提取出较纯净的目标语音作为系统的输入信号。ICA预处理可以有效抑制背景噪声和回声,提高输入信号的质量。真实环境中的实验表明,ICA预处理能够显著改善麦克风阵列语音增强系统的性能。In surroundings of strong background noise and reverberations, the quality of signals captured by microphones is very poor, thus impacting on the performance of microphone array speech enhancement systems. Independent component analysis（ICA） can extract independent components only from observations. Based on this characteristic, ICA is introduced into micro- phone array speech enhancement systems to obtain cleaner target speech signals as inputs to microphone systems. The ICA preprocessing is effective in suppressing both the ambient noise and reverberations, and dramatically improves the quality of input signals. Experimental re- sults in real environments show that ICA preprocessing can improve the system performance.

关 键 词：麦克风阵列语音增强 独立分量分析(ICA) ICA预处理 背景噪声和回声 

分 类 号：TN912.3[电子电信—通信与信息系统]