Deep learning-based M-ary spread spectrum communication system in shallow water acoustic channel

Abstract

A deep learning-based M-ary spread spectrum (SS) underwater acoustic (UWA) communication system is proposed. The long short-term memory (LSTM) model is used as the receiver of the system, enabling the direct demodulation of received signals without de-carrier and de-spreading operations. LSTM architecture-based neural network models are fed with the time-domain waveform of each symbol during the training process to learn the source information carried by each symbol. Simulation results show that the deep learning-based M-ary SS UWA communication system outperforms the conventional system under the low signal-to-noise ratio and complex shallow water acoustic channel. For optimal performance, different LSTM models can be selected according to the simulation conditions. Furthermore, a water tank experiment simulating the complex shallow water channel was conducted and the results show that experimental equipment, different training sample acquisition methods, and LSTM models can affect the performance of the system.