Coherent acoustic array processing and localization on wireless sensor networks

Compressive wireless arrays for bearing estimation

“…Our approach is fundamentally different in many ways from the earlier works for wireless arrays and compressive wireless sensing [6,7].…”

“…In [6], authors use audio codes and compression techniques to send the full individual sensor acoustic data to a cluster head, which in turn determines source bearings using the maximum likelihood (ML) and leastsquares (LS) methods. In [7], power-distortion-latency trade-offs are given for a compressive sensing scheme for sensor networks, which employs the compressive sensing framework as a universal encoding scheme to send and recover signals from multiple distributed sensors.…”

“…When compared to [6,7], our compressive wireless array approach provides a wireless sensing strategy to directly determine a sparse bearing vector in the angle domain by exploiting the redundancies in the sensor signals for the bearing estimation problem. In our approach, (i) the inter-sensor messages may not recover the original acoustic data sent by a sensor and we do not require any auxiliary signal reconstruction at the processing node in contrast to [6,7], (ii) we may not be able to determine the source signals even after determining their bearings, and (iii) the intersensor messages require significantly smaller communication bandwidth than [6] and smaller bandwidth than the scheme in [7].…”

“…In our approach, (i) the inter-sensor messages may not recover the original acoustic data sent by a sensor and we do not require any auxiliary signal reconstruction at the processing node in contrast to [6,7], (ii) we may not be able to determine the source signals even after determining their bearings, and (iii) the intersensor messages require significantly smaller communication bandwidth than [6] and smaller bandwidth than the scheme in [7]. We also do not use the ML or LS methods in obtaining our bearing estimates.…”

Target tracking in urban environments using audio-video signal processing in heterogeneous wireless sensor networks

“…Beamforming methods have successfully been applied to detect single or even multiple acoustic sources in noisy and reverberant environments [5], [16]. Adaptive background-modeling methods for motion detection based on video include the work in [10], which modeled each pixel in a camera scene by an adaptive parametric mixture model of three Gaussian distributions and the adaptive nonparametric Gaussian mixture model to address background modeling challenges presented in [23].…”