Free-space optical multiplexed orbital angular momentum beam identification system using Fourier optical convolutional layer based on 4f system

Here we introduce a free-space optical communication (FSOC) system that is capable of adjustment to alignment drift, varying atmospheric turbulent conditions of multiplexed spatial structured laser beams such as orbital angular momentum (OAM) beams. The detection system is based on heterogeneous convolutional neural network with first Fourier convolution neural network layer implemented in optics as a 4f system driven by kilohertz-fast reprogrammable high-resolution digital micromirror devices (DMDs). We utilize this optical-filtering-based convolutional neural network to realize the training and identification of two co-propagating OAM beams among 12 different multiplexed modes under simulated turbulent condition using modified von K´arm´an atmospheric model. The current implementation shows test accuracy of 95.04% (under weak turbulence) and 87.52% (under strong turbulence).