EUVIMEDEUVIMED entdecken
Crossref

PSD 2 SAN: An Unsupervised Specific Emitter Cross-Domain Identification Framework Under Class Inequality Conditions

Wenqiang Shi, Yingke Lei, Fei Teng, Hu Jin, Jin Wang, Bisheng Pan, Xu Yan, Ke Pei

IEEE Transactions on Cognitive Communications and Networking · 2026 · 12 · 10265-10280

DOI: 10.1109/tccn.2026.3706517 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Cognitive Communications and Networking, 2026. Verfasst von Wenqiang Shi und 7 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Gefördert von: National Natural Science Foundation of College of Electronic Engineering, National University of Defense and Technology.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Specific emitter identification (SEI) under cross-domain conditions becomes significantly more challenging when the target domain contains only a subset of source domain classes, leading to severe negative transfer in conventional domain adaptation frameworks. To address this realistic yet underexplored problem, this paper proposes a Power Spectral Density based Deep Subspace Alignment Network (PSD2SAN), which achieves cross-domain identification under class inequality conditions through partial domain adaptation (PDA). The key innovation lies in a power spectral density (PSD)-driven similarity mechanism that estimates class relevance between domains and assigns adaptive transfer weights, thereby suppressing source-private class interference. A multi-stage joint optimization strategy with multi-constraint subspace regularization is further designed to enhance intra-class compactness and cross-domain alignment while maintaining inter-class separability. Additionally, an adaptive class weight transfer mechanism dynamically refines the source contribution during training, stabilizing convergence and reducing negative transfer. PSD2SAN achieves the highest recognition rates of 99.52%, 98.18%, and 97.31% across different datasets, outperforming all state-of-the-art methods.

Förderung

  • National Natural Science Foundation of College of Electronic Engineering, National University of Defense and Technology

Zitieren

APA

Shi, W., Lei, Y., Teng, F., Jin, H., Wang, J., Pan, B., Yan, X., & Pei, K. (2026). PSD 2 SAN: An Unsupervised Specific Emitter Cross-Domain Identification Framework Under Class Inequality Conditions. IEEE Transactions on Cognitive Communications and Networking, 12, 10265-10280. https://doi.org/10.1109/tccn.2026.3706517

Vancouver

Shi W, Lei Y, Teng F, Jin H, Wang J, Pan B, et al. PSD 2 SAN: An Unsupervised Specific Emitter Cross-Domain Identification Framework Under Class Inequality Conditions. IEEE Transactions on Cognitive Communications and Networking. 2026;12:10265-10280. doi:10.1109/tccn.2026.3706517

BibTeX
@article{Shi2026,
  author = {Shi, Wenqiang and Lei, Yingke and Teng, Fei and Jin, Hu and Wang, Jin and Pan, Bisheng and Yan, Xu and Pei, Ke},
  title = {PSD 2 SAN: An Unsupervised Specific Emitter Cross-Domain Identification Framework Under Class Inequality Conditions},
  journal = {IEEE Transactions on Cognitive Communications and Networking},
  year = {2026},
  volume = {12},
  pages = {10265--10280},
  doi = {10.1109/tccn.2026.3706517}
}
Kostenlos starten →