Underwater non-cooperative communication signal recognition with deep learning

Abstract

The channel of underwater acoustics is time-varying and space-varying, which leads to the severe interference in the underwater acoustics communication signals. That makes a big challenge to distinguish the underwater acoustics communication signals modulation types. The traditional methods of identifying the non-cooperative signal modulation rely on statistics. They regard the statistical parameters in time-domain or frequency-domain of the signals as the features. In order to avoid extracting the features artificially we utilize the deep learning method to distinguish the raw time-domain signals into different modulation types such as Binary Phase Shift Keying (BPSK), Quadrature Phase Shift Keying (QPSK), 8 Phase Shift Keying (8PSK), Direct Spread Spectrum Sequence (DSSS) and Orthogonal Frequency Division Multiplexing (OFDM). The result of recognition for the experimental data shows the validity of our method. Finally we achieve 100% accuracy rate to BPSK, QPSK, DSSS modulation types and 90% accuracy rate to 8PSK, OFDM modulation types. Comparing with the statistics method, the accuracy of our method is higher.