Nonlinear Equalization Based on CV-CNN for 16-QAM Coherent Optical OFDM Signal at 350GHz

In the 350 GHz 16-QAM coherent optical orthogonal frequency division multiplexing (OFDM) communication system, we propose an equalizer based on complex-valued convolutional neural networks (CV-CNN) to achieve nonlinear equalization of the signal. In terms of improving system performance, not only can complex signals be directly processed without using two networks respectively, but can also achieve a 9.2 dB Q-value improvement over 40 km standard single mode fiber (SSMF) compared to traditional equalization methods that do not use neural networks (NN).