Performance Analysis of MLP-based Radar Detectors for Swerling 1 Targets

摘要

This paper deals with the application of multilayer perceptrons (MLPs) to radar detection. The dependence of the neural detector performance on the network size and on the signal-to-noise ratio selected for training (TSNR) is considered. MLPs with different numbers of neurons in their hidden layers have been trained using different values of TSNR. Results show that, when the number of hidden neurons is increased, not only is the neural detector performance close to the Neumann-Pearson optimum detector performance, but the dependence of the MLP performance on TSNR is reduced. Due to its practical interest, a very low probability of false alarm values has been considered. To estimate the probability of false alarm, importance-sampling techniques have been used in order to reduce computational cost while maintaining a low relative error.