A New Automated Undersea Wireless Sensor Detection and Fusion System

Acoustic sensors are the primary sensor of choice to detect threat submarines operating below periscope depth. The increasingly quieter nuclear threat and the diesel-electric-on-battery threat limits traditional narrow-band processing, yielding shorter detection ranges and requires more array gain via more sensors and adaptive signal processing to counter the quieting trends. One of the key functionalities of embedded sensor networks is to detect events of interest efficiently. Traditional event services allow for the definition of events including correlated events, registering for events, and notification of events upon its occurrence. In embedded sensor networks, events are not binary, but are based on sensor fusion from several noisy sensors in a complicated environment. In Phase I, we propose to develop wireless network services focusing on functions such as real-time, energy efficiency, reliability, and security. In particular, we propose to develop a location-aware event service which takes both temporal and spatial issues into consideration for accuracy. For the event detection, we propose a new cluster trending analysis method which is sensitive to small abnormal signals. To classify undersea quieter targets, we propose to develop a set of hierarchical Gaussian Mixture classifiers and a multi-classifier fusion mechanism based on the statistical evidence theory.