An Integrated Time-Frequency Detector and Radar Pulse Modulation CNN Classifier

Radar pulse detection and intra-pulse modulation classification are important capabilities in electronic warfare (EW) applications, including electronic intelligence (ELINT), electronic support measures (ESM), and radar warning receivers (RWRs). This work investigates radar pulse detection and modulation classification using a spectrogram-based time-frequency detector (TFD) integrated with a convolutional neural network (CNN) classifier. The TFD was tested over a pulse width (PW) range from 100 nsec to 100 μsec with signal-to-noise ratio (SNR) varying from −20 dB to + 30 dB, and achieved good detection at −10 dB SNR for 1–100 μsec, The CNN was trained over seven intra-pulse modulation cases with PW from 1 μsec to 100 μsec and SNR down to −6 dB. The resulting TFD-CNN is able to accurately detect and classify modulated pulses at the lower classification test limit of −6 dB SNR. The TFD-CNN was trained and tested on simulated signals, then validated using a laboratory signal transmitter and receiver system with offline processing.