Predicting patterns in time using a modified back propagation neural network

SD-Scicon have undertaken research work for the Royal Aerospace Establishment, Farnborough, on the suitability of applying artificial neural networks to various problems in the field of ESM. The paper reports on an attempt at using a neural network for radar pulse prediction and deinterleaving. The problem was to produce a neural network that recognises and predicts radar pulse trains. If there are many trains, it was required to exact just one train from the background. The problem was split into two smaller problems. Firstly, the network had to be able to predict the next pulse from a single train. Once this had been achieved, the problem of using the pulse predictor for deinterleaving was tackled