Multilevel quantizer design for decentralized detection in wireless sensor networks

We consider the problem of distributed detection involving multiple sensor nodes to jointly detect the presence of an unknown signal. To circumvent power/bandwidth constraints, a multilevel quantizer is employed in each sensor to quantize the original observation. The quantized data are transmitted through distortion channels to a fusion center where a generalized likelihood ratio test (GLRT) detector is employed to perform a global decision. We propose a quantizer design approach by maximizing the Fisher information with respect to the quantization thresholds. Numerical results demonstrate that with 2- or 3-bit quantization, the GLRT detector can provide detection performance very close to that of the unquantized GLRT detector which uses the original observations without quantization.