Unique word orthogonal signal division multiplexing with complex unitary neural network for underwater acoustic communication

Although orthogonal signal division multiplexing (OSDM) offers improved performance for underwater acoustic communication (UWAC), it still faces two major challenges related to the high peak-to-average power ratio (PAPR) and increased sensitivity to inter-vector interference (IVI). This paper proposes a novel OSDM system, termed precoded unique word OSDM based on unitary neural network (UW-OSDM-UNN), to address these issues effectively. The proposed scheme embeds the guard interval within the fast Fourier transform duration to mitigate inter-symbol interference and employs a UNN-based precoder at the transmitter to reduce PAPR and significantly overcome the IVI sensitivity. The UNN-based transmitter is completely independent of the UWAC channel, eliminating the need for receiver-side training or additional testing-stage training. Furthermore, zero vectors and frequency-shifted Chu sequences are incorporated to enable robust Doppler shift estimation and multipath compensation, respectively. The Chu sequences are inserted in the frequency domain to generate deterministic sequences within the guard interval without introducing additional inter-symbol interference. The system is validated through both simulations and real-world sea trials over a 300-meter underwater connection. Results show that the proposed scheme achieves up to a 4 dB PAPR reduction, a 5 dB improvement in bit error rate (BER), and superior robustness against challenging UWAC channel conditions compared to state-of-the-art OSDM-based systems.