Automatic Recognition for LPI Radar Waveform Using Sequence-based Deep Learning Model

Automatic recognition of low probability of intercept (LPI) radar waveforms is an essential technology in contemporary electromagnetic warfare systems. In recent years, sequence-based deep learning (DL) models have been proposed for the recognition of LPI radar waveforms, which can effectively recognize radar waveforms in rapidly changing environments. In this paper, we propose a sequence-based DL model for recognizing 12 different kinds of LPI radar waveforms. The proposed model is based on a residual network (ResNet), which can extract rich features through its deep structure. For inputs of the model, the discrete Fourier transform (DFT) and autocorrelation function (ACF) of the radar waveform are considered. Simulation results show that the proposed model outperforms existing sequence-based DL models. It is also shown that utilizing the DFT as the model's input provides an advantage over utilizing ACF in terms of recognition performance.