A Nonlinear Filter for Non-Gaussian Interference

A nonlinear filter is investigated and its effectiveness in improving signal detectability in the presence of certain types of non-Gaussian noise is determined. The filter consists of a zeromemory nonlinear device followed by a low-pass filter, the nonlinear device being designed on the basis of only the first-order statistics of the interfering noise and of the sum of signal and noise. The class of noise used in the study is that obtained by passing Gaussian noise through a zero-memory nonlinear element. Because of this the non-Gaussian process can still be characterized by relatively few parameters. The results of the study indicate that, when the noise probability density function is sufficiently different from Gaussian, a considerable improvement in detection reliability can be obtained. When the noise is Gaussian the filter reduces to a linear one.