Performance Investigation of OAM-SK Optical Communication System Based on PCNN-NMS

This paper introduces a PCNN-NMS demodulation method for an OAM-SK optical communication system operating under atmospheric turbulence, by integrating the data-driven probabilities of a Convolutional Neural Network (CNN) model with the Normalized Min-Sum (NMS) LDPC decoding algorithm. Then we simulated the average bit error rate (ABER). Initially, the transmission of Laguerre-Gaussian (LG) beams through atmospheric turbulence is simulated using the power spectrum inversion method, and a dataset is created for training the CNN network. This training achieves a direct classification accuracy of 97.3% with the CNN demodulator. Subsequently, the classification probabilities from the CNN are incorporated into the NMS algorithm, using the output of the CNN to refine the credibility of channel estimation. The final simulation results show that the PCNN-NMS significantly improves the system's BER performance, offering valuable insights for enhancing the BER performance in free-space optical communication system.