Specific Emitter Identification Based on Deep Subdomain Adaptation Network

To address the performance degradation of specific emitter identification (SEI) models caused by significant distribution discrepancies between training and test data, as well as insufficient labeled samples under low signal-to-noise ratio (SNR) conditions, a specific emitter identification method based on a deep subdomain adaptation network is proposed. A deep residual network is adopted for feature extraction, and transfer learning is introduced to optimize the local maximum mean discrepancy and the classification loss function. Experimental results show that the proposed method can effectively boost recognition performance across different SNRs, maintain high recognition accuracy with small sample sizes, and improve the generalization ability of the model.