Optimal quantization of likelihood for low complexity sequential testing

Distributed sensor systems composed of spatially distributed micro sensor nodes have been proposed for large scale monitoring applications. In these systems, nodes aggregate their sensor data to provide real time information about the underlying state. To extend the lifetime each node of the system has to limit the complexity of the sequential fusion algorithm. In this paper we derive optimal likelihood quantization rules for maximizing sequential detection performance. The resulting sequential detection algorithm is in the form of a finite state machine ideal for implementation in low complexity/low power devices.