Two typical gaussianizing filters and their application in active signal detection

Abstract

For test definite signal in non-Gaussian interference background, the matched filter or the correlation test is no more optimal. The gaussianizing filter can weaken the bigger and strengthen the smaller samples in order to enhance the gaussianity of observed data, and to improve performance of subsequent matched filter or correlation test. Firstly, a widely applied non-Gaussian model, viz. Gaussian mixture, is described. Secondly, two typical gaussianizing filters, named U-filter and G-filter respectively, are studied. Thirdly, the matched filter with gaussianizing filter preposed is proposed and its detection performance is discussed. Finally, a numerical instance is illustrated.