Distributed, robust acoustic source localization in a wireless sensor network

A distributed, robust source location estimation method using acoustic signatures in a wireless sensor network (WSN) is presented. A contaminated Gaussian (CG) noise model is proposed to characterize the impulsive, non-Gaussian nature of acoustic background noise observed in some real-world WSNs. A bi-square M-estimate approach then is applied to provide robust estimation of acoustic source locations in the presence of outlier. Moreover, a Consensus based Distributed Robust Acoustic Source Localization (C-DRASL) algorithm is proposed. With C-DRASL, individual sensor nodes will solve for the bi-square M-estimate of the source location locally using a lightweight Iterative Nonlinear Reweighted Least Square (INRLS) algorithm. These local estimates then will be exchanged among nearest neighboring nodes via one-hop wireless channels. Finally, at each node, a robust consensus algorithm will aggregate the local estimates of neighboring nodes iteratively and converge to a unified global estimate on the source location. The effectiveness and robustness of C-DRASL are clearly demonstrated through extensive simulation results.