Hybrid BS-IS Particle Filter based acoustic source tracking algorithm

Tracking moving acoustic sources in a reverberant environment has been a challenging research goal for many years. The Bayesian Filtering approach, in particular the Bootstrap (BS) and the Importance Sampling (IS) based Particle Filter (PF) tracking algorithms, has been proposed recently and promising results have been obtained. While these two algorithms out-perform other algorithms, they can only implement the optimal importance function in respective sub-optimal form and suffer related drawbacks in tracking performances. A novel hybrid BS-IS Particle Filter based acoustic source tracking algorithm is proposed in this paper. The height of the target source is used to control a sampling switch importance function that allows the tracking algorithm to switch between using the BS based and the IS based algorithm. Numerical result shows that the proposed algorithm is able to out-perform the BS and the IS based tracking algorithms in reverberant environments.