Nonlinear Compensation of Multi-CAP VLC System Employing Clustering Algorithm Based Perception Decision

Nonlinearities induced by the electrical amplifiers and the optoelectronic devices can be detrimental effects in visible light communication (VLC) systems. In this paper, clustering algorithm based perception decision (CAPD) is proposed to mitigate the nonlinear distortion in a VLC system. Aided by CAPD nonlinear compensation, we experimentally demonstrate a multiband CAP modulated VLC system consisting of a red light-emitting diode as a transmitter and a p-i-n photodiode based differential receiver. The system performances including the
Q
factor, bit error rate (BER), and computational complexity are thoroughly investigated when using a pure linear blind equalization scheme (modified cascaded multimodulus algorithm, M-CMMA) and when using hybrid linear and nonlinear equalizers (M-CMMA + Volterra series based nonlinear equalizer). The experiment results show that compared to pure linear equalizer case, the measured BER can be enhanced up to 1e−6, correspondingly the
Q
factor of each subband can be improved for around 1.6–2.5 dB by employing CAPD. The CAPD method can outperform the Volterra series based nonlinear equalizer with a lower BER value (at least 10% reduction) and relatively lower complexity. To the best of our knowledge, this is the first time that the clustering algorithm in machine learning is successfully applied to VLC systems.