Radar Emitter Identification Based on Stacked Long and Short Term Memory

With the increasing complexity of electromagnetic environment and the rising of operating patterns of new radars, emitter identification is becoming more and more difficult. This paper presents a radar emitter identification method based on stacked long and short term memory (SLSTM). Radar pulse train can be directly used as input without extracting other features, which greatly simplifies the data preprocessing and realizes the "end-to-end" recognition of radar emitter signal. The timing characteristics of the pulses are automatically extracted by SLSTM, and the optimal network parameters are trained to complete radar signal identification. Compared experiments with conventional methods are conducted, and the results show that the proposed model outperforms other existing techniques. Moreover, simulation experiments in different noise and loss pulse environment show that the method is effective and robust in solving problems of radar emitter recognition.