Robust estimation of Gaussian-noise variance in the presence of pulsed interference

A robust estimator of the mean power of a Gaussian noise signal in the presence of pulsed interference is proposed for remote sensing. The technique is based on the following rationale: under certain conditions, the histogram of a signal contaminated by pulsed interference is close to a normal distribution in a certain range of snapshot values, i.e., the histogram is an estimate almost insensitive to contaminating pulsed interference. Correspondingly, we propose a modified minimum χ² method, which is called the method of robust approximation, for estimating the mean power of the “useful” Gaussian component. It is shown that the proposed estimate yields a higher accuracy compared with, e.g., median estimates and is close to the Cramér-Rao lower bound. The results of experimental testing of the technique using hydroacoustic data are presented.