| A wireless sensor network consists of numerous intelligent sensor nodes that are capable of monitoring large regions and performing collaborative signal processing algorithms in a distributive manner. A crucial collaborative signal processing task is source localization and tracking of multiple moving targets. Algorithms developed over a wireless sensor network must exhibit low-power, and low communication overhead property so as to prolong the utilization of the service.; In this dissertation, we propose novel source localization and tracking algorithms for wireless ad hoc sensor network applications. Using source signal intensity as the key feature, these algorithms are designed to be distributively executed at different sensor cliques, and will collaboratively perform localization and tracking tasks with very moderate communication bandwidth overhead.; The thesis is composed of two parts. In the first part, a novel acoustic Energy Based source Localization (EBL) algorithm using maximum likelihood estimation method is explored. Several fast and efficient solutions, including an expectation-maximization (EM) based formulation, an adaptive iterative formulation, a projection formulation, and an approximate linear least square formulation have been developed. In addition, we also investigated, through theoretical analysis and empirical simulation, the performance and robustness of the proposed source localization algorithm.; The second part of this thesis presents two novel Distributed Gaussian Mixer Particle Filters to localize and track multiple moving targets in a wireless sensor network. Sensors are dynamically clustered into a set of uncorrelated sensor cliques based on the moving target trajectories. The proposed DPF algorithms are proved to be convergence almost surely to the centralized sequential Bayesian estimation. Moreover, a data-adaptive application layer communication protocol is proposed to facilitate sensor self-organization and collaboration. Extensive simulations are conducted to verify the performance improvement as well as the communication reduction for the proposed methods. |
