DARNet: Deep Attention Receiver Network for 5G
IEEE Communications Letters · 2026 · 30 · 572-576
DOI: 10.1109/lcomm.2025.3645100 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Communications Letters, 2026. Verfasst von Zhefu Wu und 3 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 13 Literaturangaben. Gefördert von: Zhejiang Natural Science Foundation, National Natural Science Foundation of China.
Abstract (Originalsprache)
Conventional OFDM receivers suffer performance degradation under dynamic 5G channels with high mobility and large delay spreads. Although deep learning-based receivers show promise, most existing designs emphasize frequency-domain modeling, limiting robustness in time-varying scenarios. To address this, we propose DARNet, an end-to-end receiver that directly processes time-domain signals. DARNet integrates complex-valued convolutional layers with a native sparse attention mechanism to extract and fuse time–frequency features for accurate bit recovery. Using datasets generated from 3GPP CDL channel models via the Sionna platform, evaluations show that DARNet surpasses traditional methods, achieving notable BER gains under complex channel conditions.
Förderung
- Zhejiang Natural Science Foundation
- National Natural Science Foundation of China
Zitieren
APA
Wu, Z., Zhang, T., Wan, Y., & Paul, A. (2026). DARNet: Deep Attention Receiver Network for 5G. IEEE Communications Letters, 30, 572-576. https://doi.org/10.1109/lcomm.2025.3645100
Vancouver
Wu Z, Zhang T, Wan Y, Paul A. DARNet: Deep Attention Receiver Network for 5G. IEEE Communications Letters. 2026;30:572-576. doi:10.1109/lcomm.2025.3645100
BibTeX
@article{Wu2026,
author = {Wu, Zhefu and Zhang, Tao and Wan, Yuxuan and Paul, Agyemang},
title = {DARNet: Deep Attention Receiver Network for 5G},
journal = {IEEE Communications Letters},
year = {2026},
volume = {30},
pages = {572--576},
doi = {10.1109/lcomm.2025.3645100}
}
EUVIMED