Mitigation of nonlinearities in analog radio over fiber links using machine learning approach

Abstract

Machine learning (ML) techniques are looked upon as an innovative and realistic direction to cope up with nonlinearity issues in fiber optics communication. In this paper, a 64-quadrature amplitude modulation (QAM) based radio over fiber (RoF) system is demonstrated for 10 km of standard single mode fiber length utilizing support vector machine (SVM) method to indicate an effective nonlinearity mitigation in front-hauls. The comparison of SVM is drawn with conventional ML classifiers to optimize symbol decision boundary that will reduce the RoF link impairments. The results are reported in terms of BER, Eye-linearity and Quality factor.