DOA Tracking Algorithm Based on AVS Pseudo-Smoothing for Coherent Acoustic Targets

A direction-of-arrival (DOA) tracking algorithm based on acoustic vector sensor (AVS) pseudosmoothing, referred to as the FOC-Mδ-GLMBF algorithm, is proposed to track coherent acoustic targets. This algorithm adapts the marginalized δ-generalized labeled multi-Bernoulli (Mδ-GLMB) fast filtering algorithm with the fourth-order cumulants pseudosmoothing. It introduces higher-order cumulants capable of suppressing Gaussian noise, and constructs the cumulant matrices and the likelihood function that can be used for AVS pseudosmoothing. The processing enhances the signal-to-noise ratio (SNR) by suppressing measurement noise, and can accomplish decoherence when there are coherent targets. Based on the labeled random finite set (RFS), it additionally introduces the index label to distinguish different motion models as hidden states, and achieves better tracking performance through the weighted mixture of multiple models. By using the AVS hybrid signal as the measurement, the algorithm avoids measurement-to-track association maps in the filtering process, to effectively support the tracking problem when targets are close to each other or have intersecting trajectories. In addition, as a joint prediction-and-update strategy, the algorithm performs the hypothesis truncation by the K-shortest path method only once, thereby further compensating for the burden of cumulant calculation. Simulations and field experiments verify the superiority of the proposed tracking algorithm for coherent targets under low SNR.