Target discrimination using neural networks with time domain or spectrum magnitude response

Abstract

Several different memory-based neural networks are used to discriminate radar targets based on their early-time, aspect-dependent response. The beginning of the response is difficult to locate in practice, so we use only the magnitude of the time response's DFT Spectrum as input to the neural network, thus eliminating time-shift uncertainty. Especially promising is the Recurrent Correlation Accumulation Adaptive Memory-Generalized Inverse (RCAAM-GI) cascade neural network. From the simulation results, the network demonstrates a decision strategy which is flexible, parallel adaptive, computation space efficient, and highly noise tolerant. Performances of the networks presented in this paper are compared with those of existing networks.