基于麦克风阵列的实时声源定位  

REAL-TIME SOUND SOURCE LOCALIZATION BASED ON MICROPHONE ARRAY

作　　者：刘真旭[1] 

机构地区：[1]上海交通大学计算机科学与工程系,上海200240

出　　处：《计算机应用与软件》2010年第6期43-45,共3页Computer Applications and Software

基　　金：国家高技术研究发展计划(2007AA01Z157)

摘　　要：由于噪音和多路的存在,声源定位总是很难达到一个很高的精度。针对这个问题提出了一个在智能教室环境下使用麦克风阵列对声源位置进行实时估计的算法,该算法分为三个阶段——去除噪音、延时估计和空间定位,即首先通过一个滤波器来对语音信号中噪音进行预处理,然后通过一个改进的GCC-PHAT算法来对多路语音信号的时延进行估计,最后通过构建几何关系来估计声源的位置。实验结果表明,在有一定噪音的自然教室环境下,算法的平均定位精度能够控制在5 cm以内,具有很强的应用性。With the existence of noise and multipath,it is difficult to obtain highly accurate sound source localization.This paper proposed a real-time estimation algorithm for sound source localization,which includes three parts-noise extraction,time delay estimation and geometry localization,based on microphone array to deal with this problem in a smart classroom.Firstly,it made use of a filter to pre-process the noise in speech signal;and then,a modified GCC-PHAT algorithm was used to do time delay estimation between each pair of speech signals;and at last,it estimated the sound source location according to the constructed geometry relationship.The method was also evaluated with multiple experiments.In noisy environment in an actual smart classroom,the average accuracy of localization could be controlled within 5cm,it is of very practical in usage.

关 键 词：声源定位 麦克风阵列 延时估计 去噪 

分 类 号：TN912.3[电子电信—通信与信息系统]