一种基于粒子滤波的多声源跟踪算法  

Algorithm for multiple target acoustic source tracking based on particle filtering

作　　者：蔡卫平[1] 徐健[1] 周琳[2] 

机构地区：[1]九江职业技术学院电气工程学院,江西九江332007 [2]东南大学信息科学与工程学院,南京210096

出　　处：《计算机应用研究》2014年第8期2326-2330,共5页Application Research of Computers

基　　金：国家自然科学基金青年科学基金资助项目(61201345)

摘　　要：研究了声源轨迹交叉情况下的两个声源跟踪问题。针对传统跟踪算法容易丢失较弱声源的缺点,提出一种基于粒子滤波的改进多声源跟踪算法。该算法在跟踪过程中估计声源之间的距离,当此距离小于0.3 m时,根据声源粒子的速度方向来调整粒子权重,引导粒子向各自声源的运动方向移动,从而避免较弱声源的粒子被较强声源的粒子"吸引",保持了粒子的多样性。仿真结果表明,在不同的交叉角度下,相比于传统算法,改进算法能大幅度减小对较弱声源的跟踪丢失率,并能有效提高两个声源的跟踪精度。This paper studied the problem of two acoustic source tracking in which source trajectories were cross. The conventional tracking algorithm easily lost the weaker source in this case. Aiming at the shortcoming,it proposed an improved algorithm for multiple target tracking based on particle filtering. In this method,it estimated the distance between two sources in the course of tracking. When the distance was less than 0. 3 m,means two sources become close,it regulated the particle weight according to direction of the particle velocity. This could lead the particles to shift along the direction of motion of their respective sources,thus the particles of the weaker source could avoid the attraction of that of dominant source,and the particle diversity could be retain. Simulation results show that the proposed algorithm can greatly reduce the tracking loss percentage for the weaker source and improve the tracking accuracy for both of sources compared with the conventional algorithm under the conditions of various cross angle.

关 键 词：多声源跟踪 声源轨迹交叉 粒子滤波 跟踪丢失率 跟踪精度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN912[自动化与计算机技术—计算机科学与技术]