Large N ASW False Alarm Reduction

The Large N ASW problem can be summarized as pervasive short-range sensors with limited communication capabilities. The driving questions are how are false alarms controlled-not necessarily from an individual sensor node, but from the overallfield-performance of individual sensor nodes, the amount of information and frequency of communication. From its inception, distributed detection has produced surprising results such as the fact that detector thresholds for local decisions are coupledtogether. Most systems optimize the detection performance of a single node, and then replicate the node-unaware of the distributed detection result that for optimal system performance, the decision thresholds should be coupled. We propose to applydistributed detection theory to the problem of Large N ASW for the purpose of reducing false alarms. Many new problems must be addressed, including a target transiting through the sensor field that is detectable by only a few-if any-sensors at a time, thedynamic formation of ad hoc networks as the target is sequentially detected through the sensor field, and the impact of the communication capability on local and field level performance. We will apply and develop results in distributed detection theorythat directly addresses these Large N ASW problems. Pervasive sensor networks for monitoring and surveillance are becoming more popular as concerns against threats and terrorism increase. The typical concern for automated detection systems is the controlof false alarms, causing the system to be deemed unreliable if the false alarm rate is too high. Our anticipated results directly address this problem of false alarm reduction. We also believe that our results will be directly applicable to medium sizenetworks as well as the original problem of large size networks.