Radial Basis Function Neural Network Nonlinear Equalizer for 16-QAM Coherent Optical OFDM

摘要

We propose a radial basis function neural network (RBFNN)-based nonlinear equalizer (NLE) for coherent optical orthogonal frequency division multiplexing (CO-OFDM) systems. The hidden layer neuron weights of the RBFNN-NLE are calculated using the K-means clustering algorithm and the output layer weights are updated using the least mean square algorithm. With only 3% overhead in training, the proposed RBFNN-NLE was found to provide up to 4-dB performance improvement in terms of Q-factor for 70-Gb/s 16-QAM CO-OFDM transmission over 1000 km (10 × 100 km) fiber. Numerical results show that the operating data rate is 80 Gb/s at Q = 6.25 dB with the proposed RBFNN-NLE, compared with previously reported value of 70 Gb/s with artificial neural network-based NLE.