DBF-Net: A Dynamic Mask-Aware Dual-Branch Fusion Network for Radar Emitter Identification With Variable-Length Sequences
IEEE Transactions on Aerospace and Electronic Systems · 2026 · 62 · 13055-13062
DOI: 10.1109/taes.2026.3705028 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2026. Verfasst von Shuo Zhuang und 3 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
In complex electromagnetic environments, radar emitter identification is hindered by the inherent variability of pulse descriptor word sequence lengths and severe noise interference. Conventional deep learning approaches typically rely on static zero-padding to standardize input lengths, a practice that inevitably distorts intrinsic feature distributions and introduces spurious noise, thereby degrading recognition accuracy. To address this challenge, we propose a dynamic mask-aware dual-branch fusion network (DBF-Net) designed for the distortion-free modeling of variable-length radar sequences. Unlike standard hybrid architectures, DBF-Net integrates a unified dynamic masking mechanism that permeates both a bidirectional long short-term memory branch with temporal attention and an adaptive residual convolutional branch. By eliminating padding artifacts across recurrent state updates, attention weighting, and global pooling, this strategy guarantees that feature representations are derived solely from valid signal segments. Experiments on a challenging simulated dataset comprising ten radar emitter classes demonstrate that DBF-Net achieves superior recognition performance. Furthermore, the model exhibits exceptional robustness, maintaining approximately 90% accuracy even under a severe 45% pulse loss, validating its capability to mitigate the noise amplification and information loss issues inherent in traditional padding methods.
Förderung
- National Natural Science Foundation of China
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Zhuang, S., Sun, Y., Ai, J., & Qi, M. (2026). DBF-Net: A Dynamic Mask-Aware Dual-Branch Fusion Network for Radar Emitter Identification With Variable-Length Sequences. IEEE Transactions on Aerospace and Electronic Systems, 62, 13055-13062. https://doi.org/10.1109/taes.2026.3705028
Vancouver
Zhuang S, Sun Y, Ai J, Qi M. DBF-Net: A Dynamic Mask-Aware Dual-Branch Fusion Network for Radar Emitter Identification With Variable-Length Sequences. IEEE Transactions on Aerospace and Electronic Systems. 2026;62:13055-13062. doi:10.1109/taes.2026.3705028
BibTeX
@article{Zhuang2026,
author = {Zhuang, Shuo and Sun, Yicheng and Ai, Jiaqiu and Qi, Meibin},
title = {DBF-Net: A Dynamic Mask-Aware Dual-Branch Fusion Network for Radar Emitter Identification With Variable-Length Sequences},
journal = {IEEE Transactions on Aerospace and Electronic Systems},
year = {2026},
volume = {62},
pages = {13055--13062},
doi = {10.1109/taes.2026.3705028}
}
EUVIMED