Specific Emitter Identification Based on Multi-Feature Fusion

With the continuous development of the intelligent information industry, non-cooperative Specific Emitter Identification (SEI) plays a significant role in fields such as information security, electronic warfare, and spectrum management. Currently, the feature engineering of SEI mainly includes two categories, traditional statistical methods and deep learning methods. Traditional methods, implemented by extracting statistical features, usually cannot fully extract the complex information of signals and rely on experts' domain knowledge. Therefore, deep learning methods can automatically extract the subtle features of the signals but are sensitive to noise, resulting in constrained recognition performance and stability. To address these issues, we propose a feature fusion method that combines the performance of deep learning methods with the domain prior knowledge of traditional methods. Experimental results on simulation data show that the proposed method achieves higher accuracy and better robustness than these two basic methods, demonstrating its effectiveness.