Multiple source DOA tracking based on Multi-Bernoulli Particle Filter under acoustic vector sensor array

Abstract

According to the problem of multiple source direction of arrival (DOA) tracking of acoustic vector sensor (A VS) array, a two-dimensional DOA tracking algorithm based on Multi-target Multi-Bernoulli (MeMBer) filtering is proposed. A Multi-Bernoulli random finite set (RFS) framework is employed to characterize the randomness of the number of active targets, as well as measurement process, so as to establish the observation equation. Since the closed-form solution of the multi-target posterior probability density is difficult to implement, the particle filter algorithm is used to obtain an approximate solution. The MUSIC spectrum function has been regarded as the likelihood function of particle filtering and exponentially weighted to enhance the weight of particles at high likelihood area. The merits of this proposal are that the MeMBer filtering is able to operate more directly on A VS array signals, effectively solving the problem of multiple time-varying target recognition. Simulation experiments show that this algorithm can accurately track the source state and estimate the number of sources.