Statistical normalization of non-Rayleigh reverberation

A method for dealing with non-Rayleigh reverberation in active sonar systems has been proposed and examined through simulation and application to real data. The method first removes auxiliary data cells corrupted by a spatially compact target or interference through the use of a nonparametric pruning algorithm. This algorithm most importantly does not remove reverberation outliers when no spatially compact target or interference is present as do standard order statistic methods. The clean auxiliary data set is then used to estimate the reverberation distribution through a Rayleigh mixture model with maximum likelihood estimates obtained from the expectation-maximization (EM) algorithm. Finally, the data are transformed from an unknown non-Rayleigh distribution to being approximately Rayleigh distributed through the application of a tail-deemphasizing nonlinearity. The individual parts of the statistical normalization are examined through simulation and the combined algorithm is applied to real active sonar data