Specific Emitter Identification Based on the Graph

With the increasing complexity of the electromagnetic environment, multi-modal features are often required for specific emitter identification. Multi-modal features can provide more comprehensive information about the signal from different sources and dimensions, but also easily bring redundancy and more noise. The feature fusion process in traditional specific emitter identification methods is cumbersome and difficult to analyse the relationship among features and the contribution of features. Therefore, in this paper, feature fusion and significance analysis of features are performed through graph and deep learning. The core of this method consists in autonomously learning the relationship among features and their contribution degree to the identification through the network, avoiding the limitation of manual feature selection, and guiding the network to acquire complementary features during the feature relationship learning process. The simulation experiment results verify the effectiveness and feasibility of the proposed method.