Physics-aware restoration and calibrated soft demodulation for turbulence-degraded OAM mode recognition

Authors:
Publisher website
Google Scholar
Abstract
Turbulence-degraded orbital angular momentum (OAM) mode recognition is critical for free-space optical (FSO) communication but is limited by atmospheric distortion. We propose a cascaded receiver integrating a vision transformer-based Uformer-OAM intensity-restoration network, an SE-ResNet-18 neural demodulator, and an entropy-modulated scaling (ETS) strategy for low-density parity-check (LDPC) soft decoding. In split-step beam propagation method (SS-BPM) simulations, Uformer-OAM achieves 31.34 dB peak signal-to-noise ratio (PSNR) and 96.9% recognition accuracy under strong turbulence ( C n 2 =10 −13 m −2/3 , z =1000m), outperforming UNet by 7.19 dB and 18.3 points, and the full cascade attains bit error rate (BER) −6 at z =1300m where baselines exhibit error floors.
Funding Information
  • the Dreams Foundation of Jianghuai Advance Technology Center (2023-ZM01K012)
  • National Natural Science Foundation of China (U21A20457)
  • National Natural Science Foundation of China (62071353)

This publication has 21 references indexed in Scilit: