.png){kind=icon}
.png){kind=icon}
Return
Specific Emitter Identification Based on Semi-Supervised Active Learning
.png){kind=icon}
PRE
AI
DOI：10.1109/TIM.2025.3635832.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Specific emitter identification (SEI) plays a critical role in the security and management of communication systems, particularly within the context of instrumentation and the Internet of Things (IoT). Current deep learning-based SEI approaches rely heavily on large-scale labeled datasets; however, the high cost and expertise required for annotation limit their practical application. To address these challenges, we propose SSAL-SEI, a radiation source identification method that integrates active learning with semi-supervised learning. During the active learning phase, a balanced sampling strategy that considers both uncertainty and diversity is employed to dynamically select the most informative unlabeled samples while maintaining class distribution equilibrium for annotation. In the subsequent semi-supervised phase, unlabeled data are further leveraged to enhance feature representation capabilities. To evaluate the robustness of the proposed method, experiments were conducted on the Automatic Dependent Surveillance Broadcast (ADS-B) dataset with 10, 20, and 30 classes, as well as on a Wi-Fi dataset comprising 16 classes, in conjunction with multiple SOTA models. The results demonstrate that the proposed strategy not only significantly improves the performance of semi-supervised learning but also provides effective guidance for optimizing SEI techniques under limited labeling conditions.
Keywords:
Active learning
dynamic sample selection
limited samples
semi-supervised learning
specific emitter identification (SEI)
Journal
IF:
5.9
Papers:
1.9W
Citations:
5.8W