Underwater optical wireless communication interference minimization using MMSE and deep neural networks based receivers

Abstract Underwater optical wireless communication (UOWC) has emerged as a promising technology for high-speed and low-latency underwater data transmission, offering significant advantages over traditional acoustic and radio-frequency systems. However, UOWC performance is severely degraded by ambient light noise, multipath scattering, turbulence-induced fading, and multiuser interference, which collectively limit communication reliability and range. To address these challenges, this paper proposes a deep neural network (DNN)-based interference minimization framework for multiuser UOWC systems. The proposed model learns complex nonlinear relationships between received optical signals and underwater channel distortions, enabling adaptive interference suppression and accurate signal reconstruction. Simulation results demonstrate substantial performance improvements over conventional linear receivers. In a multiuser scenario with N = 6 users, the proposed DNN receiver achieves a BER of 1.2 × 10 −4 at 25 dB SNR, compared to 3.8 × 10 −3 for the MMSE receiver and 1.5 × 10 −2 without interference mitigation. Furthermore, the DNN-based approach provides an average spectral efficiency gain of approximately 1.5 bits/s/Hz over MMSE across the SNR range of 0–30 dB, along with a 22 % reduction in bandwidth occupancy as confirmed by PSD analysis. These results validate the robustness, adaptability, and superior interference mitigation capability of the proposed framework, making it a strong candidate for next-generation underwater networks and Internet of underwater things (IoUT) applications.