Multitarget detection/tracking based on hidden Markov models

In several remote sensing applications, multitarget detection/tracking (D/T) of the backscattered wavefields is a very demanding task. Wavefield signals, sampled by an array of sensors, can be described by an hidden Markov model (HMM). As a consequence, the time of delay (TOD) profiles for each of the wavefield (or target) can be estimated by any of the known methods for state-sequence estimation such as the Viterbi (VA) and the backward/forward (BFA) algorithms. Some assumptions, that arise in the wavefield separation problem, allow one to include some additional constraints that preserve the target/tracker association. When an improved resolution is required, the choice of the multitarget Viterbi algorithm (MVA) is mandatory even if its complexity increases exponentially.