Few-Shot Specific Emitter Identification via Deep Metric Ensemble Learning

Specific emitter identification (SEI) is a highly potential technology for physical-layer authentication that one of the most critical supplements upper-layer authentication. SEI based on radio frequency (RF) features from circuit difference, rather than cryptography. These are inherent characteristics hardware circuits, which difficult to counterfeit. Recently, various deep learning (DL)-based conventional methods have been proposed, and achieved advanced performances. However, these proposed close-set scenarios with massive RF signal samples training, they generally poor performance under condition limited training samples. Thus, we focus few-shot (FS-SEI) aircraft via automatic dependent surveillance-broadcast (ADS-B) signals, novel FS-SEI method metric ensemble (DMEL). Specifically, consists feature embedding classification. The former complex-valued convolutional neural network (CVCNN) extracting discriminative compact intracategory distance separable intercategory distance, while latter realized by an classifier. Simulation results show if number per category more 5, average accuracy our higher 98%. Moreover, visualization demonstrates advantages in both discriminability generalization. code dataset can be downloaded https://github.com/BeechburgPieStar/FS-SEI .