Demodulation of Unitary Space-Time Modulated Signals Based on Deep Learning

Underwater acoustic channel is a frequency selective fading channel with severe multipath effect and significant Doppler shift, which greatly limits the performance of underwater acoustic communication system. Unitary space-time modulation (USTM) technology can achieve considerable diversity gain for underwater acoustic communication, and no channel state information is required to avoid spectrum resource waste. Signal demodulation is a key step to achieve reliable communication. In underwater acoustic communication, using traditional methods to demodulate USTM signals has high complexity and high bit error rate. This paper proposes a demodulation method CNN-GRU based on convolutional neural network and recurrent neural network for USTM. By fitting the mapping between the received signal and the likelihood information, demodulate the received signals of different constellation points. Simulation experiments show that the proposed demodulation method based on deep learning is superior to the traditional demodulation method, with a signal-to-noise ratio gain of about 0.6 dB and better adaptability to the channel environment.