Automatic Modulation Classification of Common Communication and Pulse Compression Radar Waveforms using Cyclic Features

This research develops a feature-based MAP classification system and applies it to classify several common pulse compression radar and communication modulations. All signal parameters are treated as unknown to the classifier system except SNR and the signal carrier frequency. The features are derived from estimated duty cycle, cyclic spectral correlation, and cyclic cumulants. The modulations considered in this research are BPSK, QPSK, 16-QAM, 64-QAM, 8-PSK, and 16-PSK communication modulations, as well as Barker coded, Barker coded, Barker coded, Frank coded, Px49 coded, and LFM pulse compression modulations. Simulations show that average correct signal modulation type classification %C 90% is achieved for SNR 9dB, average signal modulation family classification %C 90% is achieved for SNR 1dB, and an average communication versus pulse compression radar modulation classification %C 90% is achieved for SNR -4dB. Also, it is shown that the classification cation performance using selected input features is sensitive to signal bandwidth but not to carrier frequency. Mismatched bandwidth between training and testing signals caused degraded classification cation of %C 10% - 14% over the simulated SNR range