Normalization procedure for nonstationary reverberation with Weibull or lognormal distribution

It is pointed out that common normalization techniques, estimating a single distributional parameter of background, maintain uniform sonar performance in Gaussian reverberation, but are not effective against nonGaussian reverberation. Motivated by the reported fitting of lognormal and Weibull distribution to experimental reverberation data, the authors propose a biparametric normalization which uses a logarithmic transformation to convert the input distribution of either form into location-scale type, and then exploits best linear unbiased estimation of location-scale parameters. A performance analysis is carried out on both for locally homogeneous and for transient situations. >