Performance investigation of an OAM-SK optical communication system based on a PCNN-LDPC under atmospheric turbulence

This paper proposes a probabilistic convolutional neural network-low density parity check (PCNN-LDPC) demodulation scheme for orbital angular momentum shift keying free-space optical (OAM-SK-FSO) communication systems, with a focus on its bit error rate (BER) performance. Initially, the original information sequences were encoded by a low-density parity check (LDPC) and transformed into superposition state Laguerre Gaussian (LG) beams using a 16-Ary mapping scheme. The transmission of LG beams through atmospheric turbulence was then simulated using the power spectral inversion method, and the dataset was constructed and trained using a convolutional neural network (CNN). During demodulation, the CNN adaptive demodulator was integrated with the LDPC decoding algorithm. Classification probabilities from the CNN adaptive demodulator were used in the LDPC decoding process to enhance the channel estimation credibility. The computational results demonstrate that the BER of the PCNN-LDPC significantly outperforms both the CNN adaptive demodulator and CNN-LDPC demodulation strategy. Additionally, the performances of three LDPC decoding algorithms-min-sum (MS), normalized min-sum (NMS), and offset min-sum (OMS)-are compared, revealing that the probabilistic convolutional neural network-normalized min-sum (PCNN-NMS) achieves the best BER performance, highest convergence efficiency, and greatest decoding stability. These findings provide theoretical references and technical support for studying the BER performance of OAM-SK-FSO communication systems.