NONLINEAR CHANNEL EQUALIZATION AND ADAPTIVE LEARNING METHODS FOR LOW-POWER VLC SYSTEMS

To address the challenges of limited modulation bandwidth and nonlinear channel impairments in low-power visible light communication (VLC) systems, this paper proposes a hardware-software co-designed adaptive learning equalization scheme. At the physical layer, a hardware pre-equalization circuit is designed to compensate for Light Emitting Diode (LED) high-frequency attenuation, thereby expanding the system's physical bandwidth. At the algorithmic layer, an adaptive gradient descent learning algorithm arctan-softsign variable-step least mean square (A-SVSLMS) based on the softsign activation function is proposed. This algorithm leverages the nonlinear mapping mechanism between the step size factor and the error gradient to achieve dynamic optimization of weight updates, effectively resolving the challenge faced by traditional algorithms in balancing convergence speed and steady-state accuracy. Experimental results demonstrate that under a 0.06 W light source, the system's -3 dB bandwidth increases from 1.6 MHz to 13.5 MHz. Compared to traditional LMS algorithms, the proposed algorithm exhibits faster learning rates and enhanced robustness, successfully achieving 2 Mbps error-free transmission at a 0.55 m distance. This validates the application potential of lightweight intelligent algorithms in resource-constrained devices.