基于LMS的广义互相关时延估计  

Generalized Cross Correlation Time Delay Estimation Based on LMS

作　　者：何蒙[1] 祖丽楠[1] 孙昊[1] 杨鹏[1] 

机构地区：[1]河北工业大学电气学院,天津300130

出　　处：《电声技术》2010年第9期46-48,共3页Audio Engineering

基　　金：河北省自然科学基金(F2010000137)

摘　　要：针对复杂环境下听觉系统中对声音信号先验知识获取较难的问题,提出了一种基于最小均方自适应滤波的广义互相关时延估计方法。对传声器接收到的声源信号先进行自适应滤波,提高信号的相似度,再对滤波器权值进行平滑相干加权得到广义互相关函数来估计时延。该方法具有较强的鲁棒性,又不依赖于统计先验知识。试验证明其适用于复杂的噪声环境,提高了精度。In complex environment, it is difficult to get the prior knowledge of the sound signal in auditory system. Therefor, a method of time delay estimation is proposed on generalized correlation based on LMS（adaptive filtering）. The sound source signal received by the microphones should be adaptive filtered first, so that the signal similarity can be improved. Then the filter weight is multiplied by a smooth coherent weight function to get the generalized cross correlation function to estimate the time delay. This method of time delay estimation has strong robustness and is not dependent on prior knowledge of statistics. The experimental results show its application in complex noise environments, and it improves the precision.

关 键 词：LMS自适应 广义互相关 时延估计 

分 类 号：TN911[电子电信—通信与信息系统]