快速SRP-PHAT多声源定位算法  

Fast SRP-PHAT multi-source localization

作　　者：张二亮[1] 曹雪林 朱松 付康[1] ZHANG Erliang;CAO Xuelin;ZHU Song;FU Kang(School of Mechanical and Power Engineering,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学机械与动力工程学院,郑州450001

出　　处：《重庆理工大学学报（自然科学）》2023年第9期167-172,共6页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(61873244)。

摘　　要：相位变换加权指向响应功率(SRP-PHAT)算法在低信噪比和强混响环境下具有较好的鲁棒性,但是空间遍历带来的海量计算给其声源实时定位带来了挑战。提出了一种适用于多声源的随机区域收缩SRP-PHAT算法,通过最小描述长度(MDL)准则确定声源数量,利用K-means聚类算法进行空间区域划分,且引入瑞利限驱动的K-means聚类纠错机制,通过随机区域收缩算法进行空间区域快速收缩,利用波束主瓣宽度准则判别和界定有效声源,实现多声源定位。结果表明:针对3个声源和信噪比0 dB的情形,所提算法能够在保证定位准确率高于95%的前提下,提高30倍的计算效率,实现快速多声源定位。The Steered Response Power-Phase Transform(SRP-PHAT)algorithm has better robustness in the environment with low signal-to-noise ratio and strong reverberation,but the massive computation brought by spatial scaning brings challenges to the real-time location of sound sources.Therefore,a stochastic region contraction SRP-PHAT algorithm for multiple sound sources is proposed.The algorithm firstly determines the number of sound sources by the minimum description length(MDL)criterion,then uses the K-means clustering algorithm to partition the spatial region,and introduces the Rayleigh limitation to correct(possible)K-means clustering errors.Then,the stochastic region contraction algorithm is used to rapidly shrink the space region,and uses main lobe width to distinguish effective sound sources.It is shown by the results that,for the case of 3 sound sources and 0 dB signal-to-noise ratio,the proposed algorithm can improve the computing efficiency by 30 times under the premise that the localization accuracy is higher than 95%,and achieve fast multi-sound source localization.

关 键 词：声指向响应功率 随机区域收缩 判别准则 声源定位 

分 类 号：TN929.3[电子电信—通信与信息系统]