Forward-Error-Correction Dynamic Network for High-Accuracy Transmission in Disturbed Multimode Fiber
Journal of Lightwave Technology · 01.06.2026 · 44 (11) · 4430-4436
DOI: 10.1109/jlt.2026.3680462 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Lightwave Technology, erschienen am 01.06.2026. Verfasst von Yang Hu und 4 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Chengdu Municipal Bureau of Science and Technology Innovation and R & D Projects.
Abstract (Originalsprache)
Lasers propagating through scattering media like multimode fibers (MMFs) form distorted speckle patterns via multipath scattering, hindering signal recovery. Though data-driven deep learning has potential, neural networks cannot achieve 100% accuracy in noisy dynamic scattering media. This is due to irreversible physical information loss, incomplete training data for time-varying channels, and models lacking physical priors. To solve this, we propose a Forward-Error-Correction Dynamic Network (FECDN), fusing deep residual networks (ResNet) with Reed-Solomon (RS) coding. Guided by forward error correction in information theory, this framework leverages RS coding, a typical information-theoretic tool, to introduce redundancy at the transmitter and achieve real-time byte-level error correction at the receiver, fundamentally protecting the self-supervised training of the neural network from corrupted pseudo-labels, while ResNet captures transmission pattern changes with fewer samples to boost efficiency. Under a 100 m MMF with manual perturbations and without temperature control, the channel exhibits severe instability. In this scenario, the long-term structural similarity index (SSIM) dropped below 0.1 and the short-term SSIM dropped below 0.9. Despite these challenging conditions, FECDN achieved real-time transmission of H.265-compressed 200 × 200 video with zero frame loss and 100% symbol accuracy. This framework addresses instability and accumulation in dynamic scattering channels, providing a robust solution for real-time optical communication and laying the groundwork for intelligent, adaptive optical systems in complex or long-distance scenarios.
Förderung
- National Natural Science Foundation of China
- Chengdu Municipal Bureau of Science and Technology Innovation and R & D Projects
Zitieren
APA
Hu, Y., Fan, M., Liu, K., Jiang, N., & Wang, S. (2026). Forward-Error-Correction Dynamic Network for High-Accuracy Transmission in Disturbed Multimode Fiber. Journal of Lightwave Technology, 44(11), 4430-4436. https://doi.org/10.1109/jlt.2026.3680462
Vancouver
Hu Y, Fan M, Liu K, Jiang N, Wang S. Forward-Error-Correction Dynamic Network for High-Accuracy Transmission in Disturbed Multimode Fiber. Journal of Lightwave Technology. 2026;44(11):4430-4436. doi:10.1109/jlt.2026.3680462
BibTeX
@article{Hu2026,
author = {Hu, Yang and Fan, Minyu and Liu, Kun and Jiang, Nan and Wang, Sha},
title = {Forward-Error-Correction Dynamic Network for High-Accuracy Transmission in Disturbed Multimode Fiber},
journal = {Journal of Lightwave Technology},
year = {2026},
volume = {44},
number = {11},
pages = {4430--4436},
doi = {10.1109/jlt.2026.3680462}
}
EUVIMED