Abstract
This paper utilizes an adaptive species-based particle swarm optimization (ASPSO) algorithm in conjunction with the multi-signal classification (MUSIC) algorithm to track multi-source in a reverberant environment. In order to alleviate the impact of reverberation on source localization, this work proposes a dereverberation algorithm based on the autoregressive (AR) model. In finding the local maxima in the MUSIC pseudo-spectrum, ASPSO requires much less processing time than the uniform search method. The update rule of ASPSO consists of a single inertia weight and two acceleration coefficients. A progressively decreasing inertia weight is employed to update particle velocity. Adaptive coefficients based on different evolutionary states are dynamically adjusted to accelerate ASPSO in tracking drastic source movements. In multi-source tracking, ASPSO exploits crowding and species procedures. ASPSO equipped with AR-based dereverberation (ASPSO-AR) is investigated via numerical simulations. In adverse acoustic conditions, the results demonstrate that the ASPSO-AR algorithm is robust in tracking multiple moving sources.