EUVIMEDEUVIMED entdecken
Crossref

Physics-Constrained Self-Supervised Learning for Few-Shot Channel-Robust Specific Emitter Identification

Zhengwu Liu, Jiexin Yin, Qi Yu, Jian Luo, Kaishuo Shen

IEEE Communications Letters · 2026 · 30 · 3655-3659

DOI: 10.1109/lcomm.2026.3744903 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Communications Letters, 2026. Verfasst von Zhengwu Liu und 4 weiteren Autor:innen. Die Arbeit nennt 13 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Youth Talent Recruitment Project in the Military Science and Technology Field, Outstanding Youth Science Foundation of Henan Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Specific emitter identification (SEI) relies on subtle radio frequency fingerprints (RFFs), which are often masked by stronger channel effects in practical environments. Although self-supervised learning (SSL) can leverage unlabeled IQ signals, conventional augmentation-based SSL may learn unsuitable invariances for SEI. Specifically, weak augmentations fail to remove channel shortcuts, whereas strong or unrealistic augmentations may damage RFFs. To address this issue, we propose Physics-Constrained Positive-pair Self-Supervised Learning (PCP-SSL) for few-shot channel-robust SEI. PCP-SSL constructs hard-yet-faithful positive pairs by adversarially searching a physically plausible channel space. Furthermore, a lightweight diagonal-bispectrum-based fidelity regularizer is introduced to preserve nonlinear RFFs during this search. These physics-constrained positives guide the encoder to suppress channel bias while retaining RFFs. Architecturally, PCP-SSL employs a phase-preserving encoder and adaptive latent consistency learning to improve RFF representation. Experiments on ORACLE and WiSig datasets demonstrate that PCP-SSL consistently improves few-shot cross-domain SEI, especially under cross-distance, cross-day, and low-SNR conditions.

Förderung

  • National Natural Science Foundation of China
  • Youth Talent Recruitment Project in the Military Science and Technology Field
  • Outstanding Youth Science Foundation of Henan Province

Zitieren

APA

Liu, Z., Yin, J., Yu, Q., Luo, J., & Shen, K. (2026). Physics-Constrained Self-Supervised Learning for Few-Shot Channel-Robust Specific Emitter Identification. IEEE Communications Letters, 30, 3655-3659. https://doi.org/10.1109/lcomm.2026.3744903

Vancouver

Liu Z, Yin J, Yu Q, Luo J, Shen K. Physics-Constrained Self-Supervised Learning for Few-Shot Channel-Robust Specific Emitter Identification. IEEE Communications Letters. 2026;30:3655-3659. doi:10.1109/lcomm.2026.3744903

BibTeX
@article{Liu2026,
  author = {Liu, Zhengwu and Yin, Jiexin and Yu, Qi and Luo, Jian and Shen, Kaishuo},
  title = {Physics-Constrained Self-Supervised Learning for Few-Shot Channel-Robust Specific Emitter Identification},
  journal = {IEEE Communications Letters},
  year = {2026},
  volume = {30},
  pages = {3655--3659},
  doi = {10.1109/lcomm.2026.3744903}
}
Kostenlos starten →