Energy-Based Source Tracking and Motion Pattern Recognition Using Acoustic Sensor Networks

Acoustic sensor networks can be used for localization of an acoustic-energy emitting source. While maximum-likelihood (ML) methods are widely used for estimating the pattern of motion, more advanced machine learning schemes should be employed for improving the accuracy of localization. In this paper, we develop a learning Bayesian tracking algorithm that is capable of reconstructing the target transition model using passive wireless acoustic sensors. The adaptive scheme is intended to track targets that exhibit a complex motion pattern that cannot be adequately modeled prior to the implementation of the filtering algorithm. The derivation of the algorithm begins by modeling the likelihood assuming perfect knowledge of the sensors locations. Since this assumption is inadequate when the number of sensors is large, it is further relaxed by resorting to a probabilistic representation of the underlined locations. A Markov random field (MRF) model facilitates the implementation of an optimization method for estimating the unknown sensor locations. The convergence of this method is proven using the Kullback-Leibler divergence measure. Modeling the source path as a stochastic process yields a Bayesian localization filter. The filtering algorithm is rendered adaptive by incorporating a novel motion pattern recognition procedure based on the Baum-Welch (BW) algorithm, implemented when the target dynamics is inadequately modeled or completely unknown. Simulations show that the the tracking accuracy of the new adaptive algorithm outperforms the conventional ML scheme.