基于双耳互相关函数的声源定位算法  

Sound source localization based on binaural cross-correlation functions

作　　者：万新旺[1] 吴镇扬[1] 

机构地区：[1]东南大学信息科学与工程学院,南京210096

出　　处：《东南大学学报（自然科学版）》2011年第5期903-906,共4页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(60971098)

摘　　要：为了提高受生物启发的定位算法的定位成功率,提出了一种基于双耳互相关函数的声源定位算法.该算法包括2个阶段:离线阶段和在线阶段.在离线阶段,测量与头相关脉冲响应(HRIR),并计算所有HRIR的双耳互相关函数(BCCF).当处于在线阶段时,首先,计算接收信号的BCCF;然后,计算接收信号的BCCF与所有HRIR的BCCF之间的Pearson相关系数;最后,将最大的Pearson相关系数对应的方位作为估计的声源方位.实验结果表明,在有2个声源同时存在的情况下,相对于互通道(cross-channel)定位算法,该算法把主峰定位成功率和次峰定位成功率分别提高了约1.64%和26.48%.To improve the successful localization rate of the biologically inspired sound source localization method,a sound source localization algorithm based on binaural cross-correlation functions is proposed.The proposed method contains two phases： offline phase and online phase.During the offline phase,the head-related impulse responses（HRIR） are measured,and the binaural cross-correlation functions（BCCF） of the HRIRs are calculated.During the online phase,first,the BCCF of the received signals is computed;then Pearson correlation coefficient between the BCCF of the received signals and that of the HRIRs are calculated;finally,the azimuth corresponding to the largest Pearson correlation coefficient is regarded as the azimuth of the sound source.The experimental results show that when two sound sources are simultaneously present,compared with the cross-channel sound source localization method,the proposed algorithm increases the successful localization rate of the main peak and that of the second peak by about 1.64% and 26.48%,respectively.

关 键 词：声源定位 双耳互相关函数 与头相关脉冲响应 与头相关传递函数 

分 类 号：TN912.3[电子电信—通信与信息系统]