Deep Learning and LPF-Based Decision Technique in Free-Space Optical Communication Links

This study proposes a threshold decision technology to compensate for the turbulence effect in free-space optical (FSO) communication links, which integrates deep learning (DL) with a low-pass filter (LPF) to enhance system performance. Firstly, we introduce DL model of fully connected neural network (FCNN) for the sake of adaptive threshold decision (ATD) capability improvement. Then, in the cascaded LPF and FCNN approach, in order to improve the accuracy of channel state information (CSI) signal acquired from LPF, FCNN model is deployed behind LPF with a fixed cut-off frequency set for different turbulence channel degrees and data rates. For the adaptive cut-off frequency scheme of FCNN-based LPF technology, we utilize the FCNN model to determine the cut-off frequency value of LPF according to the estimated turbulence channel characteristics, enabling flexible variation of cut-off frequency values across diverse turbulence channel degrees and data rates. Finally, we conducted simulations to evaluate this technology. Simulation results demonstrate that FCNN-based adaptive cut-off frequency LPF technology outperforms LPF-based ATD with a fixed cut-off frequency, the FCNN-based ATD, and our proposed cascaded LPF and FCNN approach. Furthermore, its performance is approximate to theoretical ATD with comprehensive CSI knowledge. Therefore, the proposed method is a promising solution to compensate turbulence effect in FSO links. Received: 27 November 2024 | Revised: 18 March 2025 | Accepted: 11 June 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available from the corresponding author upon reasonable request. Author Contribution Statement Yan Gao: Writing – original draft. Yan-Qing Hong: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Visualization, Supervision, Project administration, Funding acquisition. Chao-Yue Zhai: Writing – original draft, Writing – review & editing. Xiao-Xue Ren: Writing – review & editing.