ResNet Equalizer for OAM Mode Divsion Multiplexing Obtical Fiber Communication System

In recent years, orbital angular momentum mdoe division multiplexing(OAM MDM) has become an important technology to enhance the transmission capacity of optical fiber communication systems. However, the nonlinear interference within the system seriously affects its performance. In this paper, a nonlinear equalization method based on residual neural networks(ResNet) is proposed, which uses deeper neural networks to compensate for nonlinear interference. In addition, this paper proposes a mixed-mode training method to improve the compensation ability of ResNet equalizer when transmitting at higher mode in fiber. Finally, based on computational cache, we design a low time complexity equalizer scheme. Through experiments, ResNet equalizer has better bit error rate(BER) performance reaching to -4.74dB and -3.37dB at mode 3 and 4, compared to CNN and Volterra equalizer. Besides, combined with the methods proposed in this paper, ResNet equalizer has stronger compensation ability at higher transmitting mode and lower time complexity.