Cross-Platform Multitag Specific Emitter Identification Method

Multi-platform specific emitter identification shows performance degradation owing to the significant distributional discrepancies in heterogeneous features. To address this challenge, we propose a graph convolutional deep residual network for classification. First, a platform–target association matrix is constructed based on the statistical features of radar data. Tag features are then implicitly modeled via tag embedding, and deep signal features are extracted using a deep residual convolutional module. Subsequently, the data and tag features are fused through attention-weighted feature splicing. To fully exploit tag dependencies, the graph convolutional network is enhanced to accept the spliced features as input, enabling the generation of feature maps that yield learnable classifiers. The extracted features and generated classifiers are integrated to produce the final classification results. Experiments on real-world datasets reveal that the proposed method achieves the dynamic fusion and efficient classification of cross-platform multi-tag features. Compared to traditional independent multi-classification approaches, it improves accuracy and computational efficiency by 17% and 9.3%, respectively. Moreover, the integration of implicit tag modeling and feature splicing contributes an additional 8% gain in accuracy, fully meeting the practical application requirements of real-world scenarios.