DATA QUANTIZATION IN STOCHASTIC-SIGNAL DETECTION SYSTEMS.

Data quantization schemes are considered for the discretization of systems designed to detect purely stochastic signals in noise. Two previously-established channel models, one of which is based on additive-noise assumption and the other on a scale-change representation for signal presence, are considered. By using earlier results for general signal-detection quantizer performance analysis, expressions are presented making it possible to compare quantizers on the basis of detection efficiency. Extensive numerical results are presented for the case where the noise has a generalized Gaussian density. Locally-optimum quantizer parameters are computed, and the relative detection performance of several quantization schemes (including minimum-distortion quantization) is considered. This analysis shows that detection performance can be enhanced considerably by proper choice of quantizer parameters.