Parallel and Serial Two-Sensor Distributed Detection Considering Binary Symmetric Channel

Abstract

In wireless sensor networks (WSNs), communication channels between sensors and fusion center (FC) are inevitably band-limited and error-prone. In this paper, we study the effect of binary symmetric channel (BSC) on a two-sensor distributed detection network with both parallel and serial topologies, in terms of the overall Bayes error of detection. For hard-decision (H-D) in the serial topology, we derive a set of coupled relations that the optimal sensors' decisions should satisfy. For soft-decision (S-D) with both topologies, we obtain sub-optimal quantization schemes at local sensors, including the sensor thresholds and the codewords (CWs) assignment, using the genetic algorithm (GA). The obtained quantizers present an inherent trade-off between data representation and error control. For a 3-bit per sensor case, these quantizers work better than the one explicitly using a repetition code.