Performance of robust detector for very heavy‐tailed noise distribution

Abstract

AbstractIn previous studies of the robust detector, it is assumed that the noise density function of the nominal function is a “strongly unimodal” density function, with a lighter‐tail than the exponential function. On the other hand, the actually observed noise density function often has a heavier‐tailed density function (more strictly, probability density function) than the exponential function; and the traditional method cannot be applied.From such a viewpoint, this paper considers the heavier‐tailed density function than the exponential function and introduces the concept of “nonstrongly unimodal” property. The concept is defined; and using the corresponding nominal density function, the robust detector is designed based on the actually observed underwater acoustic data. The performance is examined.In terms of the performance, the robust detector designed by the authors is compared with the linear and sign detectors. It is shown as a result that the robust detector designed by the authors is much better and useful.