AI-Assisted Digital Signal Recovery for Nonlinear Optical Wireless Links: A Software-Defined Approach

This study addresses signal integrity challenges in high-speed optical wireless communication (OWC) links caused by noise and nonlinear signal distortions. A degraded OWC channel is emulated in Python using an OOK-modulated binary sequence of 20,000 bits subjected to additive white Gaussian noise (SNR = 1, 3, 5, 7, and 10 dB) and a controlled deterministic sinusoidal distortion component. The distortion is introduced as an additive nonlinear disturbance to provide a reproducible test condition rather than as a complete physical model of optical-channel nonlinear propagation.

Two signal recovery strategies are compared: a conventional threshold detector and a lightweight multilayer perceptron (MLP) operating on a temporal context window of [N] samples. The MLP consists of [number] hidden layers with [architecture, e.g., 16–8 neurons] and is trained using [N] samples from the simulated received signal. Both approaches are implemented within a software-defined digital signal processing pipeline. Under severe noise (SNR = 1 dB), the threshold detector and MLP produce comparable BER values of 0.2268 and 0.2301, respectively. As channel quality improves, the MLP provides lower BER, achieving 0.1691 versus 0.1807 at 3 dB and 0.0330 versus 0.0549 at 10 dB, corresponding to an approximately 40% reduction in BER at 10 dB SNR.

The results demonstrate that a compact MLP can learn nonlinear distortion patterns present in the simulated channel and improve signal recovery under moderate and high SNR conditions. The proposed neural detector requires additional computational operations compared with the threshold-based method; however, its small architecture enables a software-defined implementation with low computational overhead. The proposed AI-assisted approach provides a flexible basis for adaptive digital signal recovery in variable OWC link conditions. Future work will investigate recurrent and sequence-aware architectures, physically grounded optical-channel models, and hardware-oriented latency evaluation for integration into high-capacity terrestrial, data-center, and inter-satellite optical communication systems.