High-efficiency and high-precision identification of transmitting orbital angular momentum modes in atmospheric turbulence based on an improved convolutional neural network

In this paper, we have proposed an improved convolutional neural network model based on the ShuffleNet V2 network for recognizing the orbital angular momentum (OAM) modes for the OAM based free space optical communication systems in the environments of atmospheric turbulence (AT). The network is trained by inputting the intensity images of the Laguerre Gaussian beams, which can effectively finish the training process due to its special designs, and can recognize the OAM modes with high accuracy. Compared with previous works for the single and multiplexing OAM modes, the proposed network model has high-precision and high-efficiency characteristics. Especially for the multiplexing OAM modes, our proposed system can achieve the recognition accuracy of 99.5% under strong AT and long-distance transmission. In addition, in order to prove that our system has good generalization ability and strong robustness, we used the trained model to test several groups of data obtained under untrained AT intensities, and the results showed that our model could still maintain high accuracy under the untrained AT intensities, which is very important to the realization of high-capacity optical communication technologies based on OAM in the future (paper)