Collaborative Transient Event Detection, Localization and Classification in Distributed Sensor Networks

A critical need for distributed sensor networks employed in various military operations, e.g. in MOUT, is an innovative transient event localization and classification system that fully exploits the limited communication, processing and power resources. A system level solution is sought that can provide an accurate assessment of threat events on the battlefield. The algorithms should be able to process a wide range of transient events in real-time and without confusion for different sensor modalities. The goal of this Phase I research is to develop innovative system level solution that (a) can detect and agree on dynamically occurring transient events using simple sensor-level detection schemes, (b) perform collaborative and confusion-free transient event localization using multiple time difference of arrival (TDOA) method, (c) estimate and restore transient signals from noisy and faded signatures, (d) extract salient time-frequency features using wavelet-based analysis, (e) perform transient classification using subband fusion and mixtures of decision experts, (f) and develop an overall battlefield transient event assessment based upon the temporal-spatial history of occurrence and types of transient events detected, localized and classified. We propose to test our algorithms and demonstrate their effectiveness on multiple transient signature data sets corresponding to ground vehicles, artillery fires, mortar fires, small arms fires, etc. that will be acquired from the US Army ARDEC.