On the performance of cumulant-based detection of nonGaussian signals

Detection of a non-Gaussian signal in additive Gaussian noise is assumed to be done in two steps. A selected set of sample cumulants and/or moments of the data are computed; next a detection statistic based on these sample moments/cumulants is evaluated. The paper outlines the asymptotic performance analysis of this detector for a fairly general class of signal and noise models. A simple example is worked out in detail.<
>