Vardiational Bayesian Hybrid Multi-Bernoulli and CPHD Filters for Superpositional Sensors

This paper addresses the problem of multi-target tracking with superpositional sensors, while covariance matrices measurement noise are not known. The proposed method is based on hybrid multi-Bernoulli cardinalized probability hypothesis density (HMB-CPHD) filter, which has been developed for sensors-based known noises. Specifically, we firstly propose Gaussian mixture (GM) implementation HMB-CPHD and then noises augmented into target state vector, resulting in inverse Wishart (GIWM) representation state. Then variational Bayesian (VB) exploited to approximate posterior distribution so that it maintains same form as prior distribution. A remarkable feature can jointly perform estimation. performance algorithm demonstrated via simulations.