Signal Detection for Full-duplex Cognitive Underwater Acoustic Communications with SIC Using Model-Driven Deep Learning Network

This paper aims to handle the model-driven deep learning network based signal detection for full-duplex cognitive underwater acoustic communications (FDCUACs) with self-interference cancellation (SIC). The FDCUACs play an important role in underwater wireless communications, which employs the index modulation-orthogonal frequency division multiplexing-spread spectrum (IM-OFDM-SS) that carries subcarrier index bits and symbol bits simultaneously to further enhance data rate. It is shown that the proposed signal detector for the FDCUACs with the SIC can be modelled as a model-driven deep learning network, i.e., an easy index bit recovering processor utilizing the uniqueness behavior of spread codes plus a deep learning network with eight essential layers employed to directly regain the data belonging to the pre-selected carrier index in the transmitter. Compared with the original receiver and signal detection for the IM-OFDM-SS communications, the proposed scheme omits the traditional steps such as channel estimation, equalization and demodulation, and demonstrates the remarkable performance.