Wavelet Autoencoder for Radar HRRP Target Recognition with Recurrent Neural Network

A Wavelet Autoencoder model with Recurrent Neural Network (WaveletAE with RNN) is developed for radar automatic target recognition (RATR), with an encoder-decoder layer, in which the weights of decoder are fixed as a set of overcomplete bases derived from mother wavelet. Imposing an sparsity constraint on the hidden units of encoder-decoder layer, interesting structure in the data is discovered, and superior recognition performance is achieved on the measured High Resolution Range Profiles (HRRP) data, showing the effectiveness of the proposed model. Specific results are represented in our experiments.