A Modified δ-Generalized Labeled Multi-Bernoulli Filtering for Multi-Source DOA Tracking With Coprime Array

Abstract

For the target tracking problem where the number of targets fluctuates with time and the measurement is a point measurement, the random finite set (RFS) class filtering is an available solution. However, in direction of arrival (DOA) tracking, the array observation is a super-positional value, and the tracking performance can be severely impaired if the RFS-based filter approach is applied. As a result, a novel measurement association mapping (NMAP) approach has been presented to cope with the mapping problem between the array observations and sources. Nevertheless, the tracking performance is poor when the number of particles is small. In this paper, a modified delta-Generalized Labeled Multi-Bernoulli (δ-GLMB) DOA tracking particle filter is proposed in combination with the NMAP strategy, which can achieve the same tracking performance with a smaller number of particles by modifying the particles in the δ-GLMB prediction step. Furthermore, the approach is extended to a coprime array and can achieve better DOA tracking performance than a uniform linear array. Simulation experiments validate the effectiveness of the proposed algorithm.