DCO-OFDM Signals With Derated Power for Visible Light Communications Using an Optimized Adaptive Network-Based Fuzzy Inference System

Abstract

Direct current-biased optical orthogonal frequency division multiplexing (DCO-OFDM) signals used in visible light communications suffer from high peak-to-average-power ratio (PAPR) or cubic metric (CM). It strongly degrades the performance due to the great back-off necessary to avoid the clipping effect in the light-emitting diode. Thus, PAPR and CM reduction techniques become crucial to improve the system performance. In this paper, an adaptive network-based fuzzy inference system (ANFIS) is used to obtain efficient DCO-OFDM signals with a low power envelope profile. First, signals specially designed for DCO-OFDMwith very low CM, as the ones obtained from the raw cubic metric (RCM)-active constellation extension method, are used to train the fuzzy systems in time and frequency domains. Second, after the off-line training, the ANFIS can generate a real-valued signal in a one-shot way with 8.9 dB of RCM reduction from the original real-valued signal, which involves a gain in the input power back off larger than 2.8 dB, an illumination-to-communication conversion efficiency gain of more than 35% and considerable improvements in bit error rate.