Extreme Learning Approach for Timing Synchronization of ACO-OFDM under Nonlinear Distortions

Asymmetrically-clipped optical orthogonal frequency division multiplexing (ACO-OFDM) is the optimal power-efficient modulation technique for the intensity-modulated (IM) and the direct-detection (DD) based visible light communication (VLC) systems. In ACO-OFDM, the non-linear distortions of light-emitting diodes (LEDs) and diffused channels will affect the estimation of delay offset from the received data frame. These distortions corrupt the timing metric (TM) by generating several peaks around the position of the actual delay, thus affecting the delay offset estimation. In this paper, we propose a timing synchronization method for ACO-OFDM using an extreme learning machine (ELM) based delay offset estimator by considering the identification of delay offset position as a feature extraction problem. Then, the simulations are carried out considering the real-time effects of high-power LED distortions and diffused channel characteristics. The simulation results show a reduction in the probability of error (PoE) of delay offset estimation and bit error rate (BER) for the received signal of different frame lengths, diffused channels, and randomly varied LED non-linear conditions compared to the conventional methods.