The bispectrum and bicoherence for quadratically nonlinear systems subject to non-Gaussian inputs
IEEE Transactions on Signal Processing
| Hi-index | 35.69 |
In certain scenarios, higher order statistic (HOS)-based detectors exploiting the Gaussian noise rejection property of HOS have been shown to have performances superior to widely used second-order (SOS) techniques. However, a crucial limitation on these detectors that has not been addressed is a constraint on the processing bandwidth. We study the effects of commonly used narrowband processing on these detectors. As the processing bandwidth is decreased, we characterize the tradeoff between HOS and SOS detectors and demonstrate that the performance of HOS detectors degrades faster than that of SOS ones. We then consider distributed wideband detection by a bank of local narrowband detectors whose decisions are fused together. Given a fixed-bandwidth input signal, we show that as the number of local detectors increases, corresponding to a decreasing amount of bandwidth for each local detector, the performance of such a scheme using HOS local detectors degrades more quickly than a scheme using SOS ones