Research on Orbital Angular Momentum Multiplexing Communication System Based on Neural Network Inversion of Phase

An adaptive optical wavefront recovery method based on a residual attention network is proposed for the degradation of an Orbital Angular Momentum multiplexing communication system performance caused by atmospheric turbulence in free-space communication. To prevent degeneration phenomenon neural networks, used as backbone network, and multi-scale hybrid constructed. Distributed feature extraction convolutional kernels at different scales to enhance network’s ability represent light intensity image features. The mechanism improve recognition rate broken spot loss function designed combining realistic evaluation indexes so obtain Zernike coefficients that match actual aberration. Simulation experiments are carried out conditions, results show can reconstruct turbulent phase quickly accurately. peaks valleys recovered aberrations were between 0.1 0.3 rad, root means square was 0.02 0.12 rad. obtained better than those conventional SNRs.