Tracking moving acoustic sources with a network of sensors

This paper presents issues and algorithms for the problem of source tracking with a network of aeroacoustic sensors. We study fusion of data from sensors that are widely separated, and we give particular attention to the important issues of limited communication bandwidth between sensor nodes, effects of source motion, coherence loss between signals measured at different sensors, signal bandwidth, and noise. We compare the tracking performance of various schemes, including joint (coherent) processing of all sensor data, as well as data-reduction schemes that employ distributed computation and reduced communication bandwidth with a fusion center. The key result of our analysis is a quantification of the potential gain in source tracking accuracy that is achievable with greater communication bandwidth and joint processing of sensor data. We show that the potential gain in accuracy depends critically on the scenario, as determined by the source motion parameters, signal coherence between sensors, bandwidth of the source signals, and noise level. For scenarios that admit increased accuracy with joint processing, we present a bandwidth-efficient algorithm that involves beamforming at small-aperture sensor arrays combined with time-delay estimation between widely-spaced sensor arrays.