Professor Ba-Ngu Vo (Curtin University)
Department of Statistics, LG.03
Abstract: In a finite-set-valued Hidden Markov Model (HMM), the hidden state is a finite set. Such an HMM describes a multi-object system, a system in which the number of objects and their states are unknown and vary randomly with time. Multi-object systems arise in many research disciplines including surveillance, computer vision, robotics, biomedical research and machine learning. Indeed, most systems in nature can be regarded as multi-object systems. The last decade has witnessed exciting developments with the introduction of stochastic geometry to finite-set-valued HMMs. This talk presents recent developments in smoothing and filtering for finite-set-valued HMMs with applications in multiple object tracking, especially large-scale problems.