Designing Local Quantizers for Distributed Detection of a Source with Unknown Location

A wireless sensor network is used for distributed detection of a source with unknown location. To conserve communication resources, quantization is required at the sensors prior to data transmission. At each sensor, it is assumed that the received signal power decays with distance following a certain attenuation model. Sensors quantize their signal intensity measurements locally to obtain binary decisions, which are transmitted to a fusion center (FC), and then fused by a generalized likelihood ratio test (GLRT), to obtain a global decision. To achieve better detection performance, three new local quantizer design approaches, based on the principles of maximum average Kullback-Leibler divergence (KLD), maximum average Fisher information (FI), and equiprobable quantizer outputs respectively, are proposed. The KLD- and FI-based approaches are also extended to design multi-bit quantizers. Simulation results are provided and the quantizers designed based on the average KLD and average FI lead to superior performance.