Abstract: To address the problem that the nonlinear response of light-emitting diodes (LEDs) leads to serious degradation of visible light communication (VLC) performance, a K-nearest neighbor (KNN) algorithm improved based on rough set theory is proposed, and further, a cascaded equalizer is designed by combining it with least mean square (LMS). First, the training set data space is divided into different regions according to the distribution characteristics of constellation points at the receiver side, and different classification strategies are used for different regions to reduce the computational complexity of the traditional KNN algorithm. Then, the LMS and improved KNN cascade equalizer are proposed, and the first stage LMS algorithm can reduce the dispersion of sample points, which provides conditions to improve the classification accuracy and reduce the computational complexity of the second stage improved KNN. Finally, Monte Carlo BER simulation is used, and the results show that the complexity of the improved KNN algorithm is about 1/9 of the traditional KNN algorithm without sacrificing the classification accuracy; meanwhile, the proposed LMS with improved KNN cascade equalizer can significantly improve the BER performance.

Key words: visible light communication, LED nonlinearity, rough set theory, K-nearest neighbor algorithm, LMS algorithm