On the inferiority of higher-order detection in narrowband processing

In certain scenarios higher-order statistic (HOS)-based detectors for non-Gaussian signals have been shown to have performance superior to widely-used second-order (SOS) techniques, owing to the HOS Gaussian noise rejection property. The authors study the effects of commonly-used narrow-band processing on these detectors. As the processing bandwidth is reduced, they characterize the tradeoff between the HOS and SOS detectors and demonstrate that the performance of the HOS detector deteriorates faster than that of the SOS detector. Distribution wideband detection is considered by a bank of L narrowband detectors whose decisions are fused together. It is shown that, given a fixed-bandwidth input signal, as L increases, the performance of the HOS distributed detector deteriorates faster than that of the SOS detector.<>