Deep-learning-based Adaptive Predistorter for Nonlinear LED Compensation in Visible Light Communication Systems

Abstract

In typical visible light communication (VLC) systems, light emitting diodes (LEDs) are adopted as light transmitters by modulating the optical power of LEDs with input current. However, LEDs has a well-known nonlinear distortion which cannot be ignored and would inevitably degrade the performance of VLC systems. Since the deep-learning (DL) method has a reputation for nonlinear approximation, a DL-based predistorter is proposed to mitigate the LED nonlinearity. Simulation results show that the proposed method has superior performance than existing linear predistortion methods, such as the look-up-table (LUT), normalized least mean square (NLMS) algorithms and the nonlinear predistortion method, such as the Chebyshev polynomial-based algorithm.