Improving Nonparametric Detectors via Stochastic Resonance

This paper investigates the problem of stochastic resonance in some non-parametric detection schemes such as the sign detector, Wilcoxon detector and dead-zone limiter detector. Detection performance comparisons are made between the original detectors and noise modified detectors. Potential improvement of detection performance via stochastic resonance (SR) for each detection scheme is determined. The optimal SR noise for the sign detector is derived. Asymptotic detection performance is evaluated for the three detectors. For the finite sample size problem, we demonstrate improvability via some detection examples where the detection performance of these detectors can be improved when suitable noise is added.