An Improved PHD Filtering for DOA Tracking With Sparse Array via Unscented Transform Strategy

Traditional direction of arrival (DOA) tracking methods are sensitive to the number of sources, and their performance degrades when the number of sources changes with time. To solve the problem of multi-sources time-varying DOA tracking, an improved probability hypothesis density (PHD) method for DOA tracking is proposed in this brief. We improve the particle diversity of the resampling method and enhance the tracking performance by preprocessing the predicted particles in the prediction step of the traditional PHD particle filtering method. Moreover, the proposed method is extended to sparse array scenario to further improve the DOA tracking accuracy. Performance of the algorithm can be demonstrated by simulations with better performance than the spatial smoothing PAST (SS-PAST) and existing PHD methods.