Energy-efficient robust acoustic source localization with local outlier rejection in Wireless Sensor Networks

Outliers can be devastating for conventional MLE and LS based acoustic source localization methods. To solve this problem with less energy consumption is the focus of this paper. A contaminated Gaussian model is proposed for acoustic energy sensor readings and shown to be consistent with real-world experiment data. Then a novel energy-efficient robust acoustic source localization protocol (ERASLP) is introduced where each node examines its own observation by a local outlier rejection rule (LORR). Detected outliers will not be sent to the fusion center for the final localization result calculation. The average communication energy cost and applicability of ERASLP is analyzed and compared to typical centralized localization method and distributed optimization method. Simulation results show LORR can surpass more than a half of the outliers even with very sparsely deployed sensors and ERASLP has better tradeoff between robustness and energy consumption in most cases.