Abstract:

In the case of few label samples of the communication emitter signal, it is difficult to extract individual features of similar emitter signals and the identification accuracy is low. To this regard, a semi-supervised feature extraction method of communication emitter under the small sample condition is proposed. A small number of labeled emitter signal samples and a large number of unlabeled emitter signal samples are subjected by variational mode decomposition to extract high-dimensional steady-state information entropy. The exponential semi-supervised discriminant analysis is used to map the information entropy to form individual features. In addition, XGBoost is used to identify the communication emitter to verify the identification effect. Experiments show that the proposed method reduces the computation time by 76.17% compared with the unsupervised feature extraction method, and the identification rate reaches 85.33%, which proves that it has better performance in different individual identification of similar communication emitters.

Key words: individual communication emitter identification, feature extraction, variational mode decomposition, exponential semi-supervised discriminant analysis