An Adaptive Deep Ensemble Learning for Specific Emitter Identification.

Area of Science:

• Electrical Engineering
• Computer Science
• Signal Processing

Background:

• Specific emitter identification (SEI) relies on hardware-intrinsic radio frequency fingerprints (RFFs) for classifying radio transmitters.
• Existing SEI methods struggle with noise robustness, limited training data, and class imbalance.

Purpose of the Study:

• To introduce a novel framework, Adaptive Deep Ensemble Learning (ADEL), to overcome SEI limitations.
• To enhance the robustness and generalization capabilities of SEI systems.

Main Methods:

• ADEL integrates heterogeneous neural networks (CNN, MLP, Transformer) for hierarchical feature extraction.
• The framework employs adaptive weighted predictions based on reconstruction errors and hybrid losses.
• It refines feature space structure and preserves feature integrity for improved generalization.

Main Results:

• ADEL significantly outperforms existing competing methods in SEI tasks.
• The proposed method demonstrates effectiveness under conditions of limited training samples and imbalanced classes.
• Experimental validation confirms the superiority of the ADEL framework.

Conclusions:

• ADEL establishes a new paradigm for SEI through robust feature extraction and adaptive decision integrity.
• The framework shows potential for deployment in applications like space target identification and situational awareness.
• ADEL offers a robust solution for SEI in challenging real-world scenarios.