Vision Mamba demodulation with efficient channel attention for LDPC-coded orbital angular momentum shift keying over turbulent free-space optical links
Optics Express · 24.08.2026 · 34 (17) · 31546
DOI: 10.1364/oe.601445 ↗Kurz gefasst
Fachartikel in der Zeitschrift Optics Express, online erschienen am 12.08.2026. Verfasst von Junfeng Zhai und 4 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
We propose a Vision Mamba demodulator with efficient channel attention (Vim-ECA) for low-density parity-check (LDPC)-coded orbital angular momentum shift keying (OAM-SK) over turbulent free-space optical (FSO) links. Despite the structured intensity signatures of multi-mode OAM-SK symbols, atmospheric turbulence severely distorts their received intensity patterns, increasing inter-mode confusion and degrading demodulation performance. Vim-ECA combines a Vision Mamba (Vim) backbone with an efficient channel attention (ECA) gate to capture turbulence-induced global distortions and enhance turbulence-robust spatial features, thereby improving symbol separability. We evaluate the end-to-end performance of an LDPC-coded 16-ary OAM-SK link over turbulent channels. Extensive simulations compare Vim-ECA with Vim, convolutional neural network (CNN)/CNN-ECA, and Swin Transformer (Swin)/Swin-ECA baselines across a wide range of turbulence strengths, using post-decoding bit error rate (BER), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM) for evaluation. Under strong turbulence at C n 2 =1.94×10 −13 m −2/3 , Vim-ECA achieves a BER of 1.992 × 10 −2 , with 21.86 dB PSNR and an SSIM of 0.588, outperforming its non-ECA Vim counterpart as well as the ECA-enhanced CNN and Swin baselines. Across the evaluated turbulence range, Vim is more robust than CNN and Swin under moderate-to-strong turbulence, while ECA consistently improves all three backbones; together, these factors make Vim-ECA the best-performing model in end-to-end evaluation.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Zhai, J., Li, Z., Shang, T., Guo, B., & Chang, Z. (2026). Vision Mamba demodulation with efficient channel attention for LDPC-coded orbital angular momentum shift keying over turbulent free-space optical links. Optics Express, 34(17), 31546. https://doi.org/10.1364/oe.601445
Vancouver
Zhai J, Li Z, Shang T, Guo B, Chang Z. Vision Mamba demodulation with efficient channel attention for LDPC-coded orbital angular momentum shift keying over turbulent free-space optical links. Optics Express. 2026;34(17):31546. doi:10.1364/oe.601445
BibTeX
@article{Zhai2026,
author = {Zhai, Junfeng and Li, Zhaokun and Shang, Tao and Guo, Binquan and Chang, Zheng},
title = {Vision Mamba demodulation with efficient channel attention for LDPC-coded orbital angular momentum shift keying over turbulent free-space optical links},
journal = {Optics Express},
year = {2026},
volume = {34},
number = {17},
pages = {31546},
doi = {10.1364/oe.601445}
}
EUVIMED