Student-t mixture labeled multi-Bernoulli filter for multi-target tracking with heavy-tailed noise

A new labeled multi-Bernoulli (LMB) filter is proposed for multi-target tracking with joint heavy-tailed noises of the state and measurement. In contrast to the Gaussian assumption, the proposed method models both the process and measurement noises as multivariate Student-t distributions to handle the heavy-tailed noises. A closed form recursion of the LMB filter to propagate the parameters of Student-t mixture components is derived based on the multi-target Student-t models. Some approximations are applied to make the filter available in practice. The gating technique is also developed for the proposed method. Furthermore, a strategy of managing the number of Student-t components is introduced here to ensure the efficiency of the proposed method. Simulations with joint heavy-tailed noises of the state and measurement are performed to assess the proposed filter, and results demonstrate effectiveness of the new LMB filter.