Unsupervised Multistage-Clustering-Based Hammerstein Postdistortion for VLC

Recently, there has been a huge interest in research toward visible light communication (VLC) targeted toward fifth generation (5G) and beyond standards. One of the performance limiting factors in visible light communications is the nonlinear characteristics of the light-emitting diode (LED), which is used as a transmitter in VLC. Among seminal works for mitigating performance-limiting nonlinearity, the modified cascaded multimodulus algorithm (MCMMA) based Volterra filtering approaches have been proposed recently for blind LED postdistortion. This work proposes an unsupervised normalized Hammerstein improved multistage clustering (HIMSC) based postdistorter that is formulated as a more suitable cost function, as compared with the Volterra-MCMMA-based approaches. Theoretical bounds on step size are found to ensure convergence of the proposed postdistorter. Simulations over IEEE 802.15 personal area network (PAN) channels indicate that the proposed normalized HIMSC delivers much better performance as compared with the Volterra-MCMMA in terms of mean square error (MSE) and bit error rate (BER) performance.