EUVIMEDEUVIMED entdecken
Crossref

High Efficient and Robust Broadband Beamforming With Improved LSTM and Transformer

Bi Wen, Jiali Li, Shurui Zhang, Cong Xue, Hairui Zhu, Jie Luo, Siqin Jiang, Yanbo Liu, Weixing Sheng

IEEE Transactions on Cognitive Communications and Networking · 2026 · 12 · 8272-8287

DOI: 10.1109/tccn.2026.3693273 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Cognitive Communications and Networking, 2026. Verfasst von Bi Wen und 8 weiteren Autor:innen. Die Arbeit nennt 48 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities, National Key Laboratory of Transient Physics.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Broadband beamforming is one of the most important techniques in modern communication systems. With the widespread application of deep learning across diverse domains, adaptive broadband digital beamforming algorithms integrated with neural networks have emerged. However, insufficient prior information, such as inaccurate directions of the signal of interest (SOI) and interference signals, degrades the performance of these approaches. Furthermore, their performance also deteriorates in low snapshot scenarios. To tackle these challenges, we propose two algorithms for distinct scenarios. One is the fast and robust broadband beamforming network based on long short-term memory (LSTM), abbreviated as FRBL. The other is the fast and robust broadband beamforming network based on Transformer, abbreviated as FRBT. These two algorithms respectively utilize the LSTM model and Transformer model as the foundational architectures, and introduce a decimal separable system to improve the network fitting performance, which could adaptively align the desired signal within a certain range and generate accurate nulls in the interference directions without prior knowledge of the exact angle in limited snapshot situations. Simulation results demonstrate that compared with existing neural network-based algorithms, FRBL and FRBT reduce the required number of signal snapshots to 16%, achieving higher system efficiency. Meanwhile, the two proposed algorithms exhibit a significant reduction in the demand for prior information, leading to superior robustness in low snapshot scenarios.

Förderung

  • National Natural Science Foundation of China
  • Fundamental Research Funds for the Central Universities
  • National Key Laboratory of Transient Physics

Zitieren

APA

Wen, B., Li, J., Zhang, S., Xue, C., Zhu, H., Luo, J., Jiang, S., Liu, Y., & Sheng, W. (2026). High Efficient and Robust Broadband Beamforming With Improved LSTM and Transformer. IEEE Transactions on Cognitive Communications and Networking, 12, 8272-8287. https://doi.org/10.1109/tccn.2026.3693273

Vancouver

Wen B, Li J, Zhang S, Xue C, Zhu H, Luo J, et al. High Efficient and Robust Broadband Beamforming With Improved LSTM and Transformer. IEEE Transactions on Cognitive Communications and Networking. 2026;12:8272-8287. doi:10.1109/tccn.2026.3693273

BibTeX
@article{Wen2026,
  author = {Wen, Bi and Li, Jiali and Zhang, Shurui and Xue, Cong and Zhu, Hairui and Luo, Jie and Jiang, Siqin and Liu, Yanbo and Sheng, Weixing},
  title = {High Efficient and Robust Broadband Beamforming With Improved LSTM and Transformer},
  journal = {IEEE Transactions on Cognitive Communications and Networking},
  year = {2026},
  volume = {12},
  pages = {8272--8287},
  doi = {10.1109/tccn.2026.3693273}
}
Kostenlos starten →