A Deep Neural Network Equalizer for FSO Transmission System

A deep neural network (DNN) equalizer for 16-QAM signal in free space-optics (FSO) transmission system is proposed and experimentally demonstrated in this paper. When the fiber and FSO links in fiber-wireless system are cascaded, the nonlinearity will become serious. Therefore, a DNN equalizer is designed to reduce the nonlinear effect in the FSO system based on the channel nonlinearity. The result shows that DNN equalizer approach can minimize the nonlinear and reduces error vector magnitude (EVM).