Patterns Quantization With Noise Using Gaussian Features and Bayesian Learning in VLC Systems

Abstract

Nonlinear impairments seriously affect the transmission performance of high-speed visible light communication (VLC) systems, which have become the bottleneck of VLC systems in practical applications. To compensate for the nonlinear impairments of VLC systems, we propose a Pattern quantization method (PQM) based on Gaussian feature inputs and Bayesian learning in the time-domain. First, we investigate the signal distribution of VLC systems in the presence of additive white Gaussian noise (AWGN) and nonlinear impairments. Second, probability statistics are used to quantify the noise. Finally, a nonlinear compensator for PAM-8 VLC systems is realized using the Bayesian machine learning principle. In the experiment, a PAM-8 VLC system with a maximum transmission data rate of 1 Gb/s and bit error rate (BER) below the threshold of hard decision forward error correction (HD-FEC) is successfully demonstrated. The results indicate that the PQM compensator can significantly reduce nonlinear impairment and improve BER performance while requiring little computational effort. The Q-Factor is improved by 2.1 dB, the transmission distance is increased by 1/3, and the BER is reduced by 90.05% when compared with conventional nonlinear Volterra equalizers. As far as we know, our study is the first to investigate the input features and quantify the white noise in VLC systems, thereby providing a novel approach to machine learning input data processing.