BiLSTM Based Phase Modulation Detection of Radar Emitters

In modern Electronic Warfare (EW) systems, recognition of the intrapulse modulation schemes of intercepted Low Probability of Intercept (LPI) radar signals in real-time is a crucial survival task. Lately, Convolutional Neural Network (CNN) had proven effective for emitter recognition but the problem lies with the recognition of phase-coded waveforms at low Signal to Noise Ratio (SNR). We propose a Bidirectional Long Short Term Memory (BiLSTM) network-based recognition technique for the analysis and recognition of phase-coded waveforms. In order to avoid heavy processing, the time-domain radar signals are directly fed as an input to the BiLSTM network for recognition without any preprocessing and feature extraction. The BiLSTM layer can extract the contextual information of signals well and is followed by a fully connected layer. Finally, a softmax classifier is employed to accomplish the recognition task. Six distinct types of phase-coded waveforms are simulated corrupted with Additive White Gaussian Noise (AWGN) with the SNR ranging from -10 to 10 dB. It has been demonstrated that the proposed methodology considerably enhances recognition accuracies over existing techniques.