Application of neural networks to radar signal detection in K-distributed clutter

Radar signal detection is a complex task that is generally based on conventional statistical methods. In real applications, these methods require a lot of computing to estimate the clutter parameters and that they are optimal only for one type of clutter distribution. Recently, artificial neural networks (ANNs) have been used as a means of signal detection. Following on from this work, the authors consider the problem of radar signal detection using ANNs in a K-distributed environment. Two training algorithms are tested, namely, the back-propagation algorithm, and genetic algorithms for a multi-layer perceptron (MLP) architecture and also for the radial basis function architecture. The simulation results show that the MLP architecture outperforms the classical cell-averaging constant false alarm rate and order statistics constant false alarm rate detectors.