A fuzzy approach to signal integration

Abstract

A new fuzzy concept for postdetection signal integration is presented. The fuzzy integrator is developed as a simple extension of the classic binary integrator by replacing the crisp binary threshold, which quantizes the observed data, with a fuzzy threshold. The performance of the fuzzy integrator is illustrated for detection of a simple nonfluctuating signal in Gaussian noise and is shown to exceed that of the binary integrator, approaching that of the optimal detector with Neyman-Pearson decision rule. Furthermore, the fuzzy integrator has the characteristic that the false alarm rate can be tuned using a single threshold, more easily than that of the dual-threshold binary integrator.