Optimum detection of Rayleigh signals in nonstationary noise

The authors propose an adaptive thresholding procedure for Rayleigh envelope distributed signals and noise, where noise power residues instead of noise power estimates are processed. It is shown that it is the optimum constant false alarm rate (CFAR) detector when the noise samples are statistically independent and identically distributed in the sense that its detection performance approaches that of the ideal (fixed threshold) detector as the number of noise samples becomes very large. The authors also discuss the merits of the proposed technique in rejecting interferences in the noise samples