粒子滤波 PF( Particle Filtering) 是在

贝叶斯滤波框架下基于 Monte Carlo 采样的统计滤波方法，可解

决非线性、非高斯的滤波问题，声源跟踪就属于这类问题。近年

来，基于粒子滤波的声源跟踪算法已成为研究的热点。

-he real acoustic data recorded in a typi-

cal meeting room using a small-scale microphone array is used for tracking experiments． Results show that the proposed algorithm can realise

effectively the orientation tracking of the randomly walking speaker，in which，the root mean square error of tracking in direction of the hori-

zontal angle and the elevation are all less than 5°．