Specific Emitter Identification Based on CNN via Variational Mode Decomposition and Bimodal Feature Fusion

Specific emitter identification (SEI) refers to the technology of identifying specific emitter individuals by the RF fingerprint carried by the emitter transmitting signals. RF fingerprint is a hardware difference feature caused by the nonlinearity and tolerance of the components inside the emitter, and can be used as the unique identification of the emitter. In this paper, we propose an emitter identification scheme based on convolutional neural network (CNN) via variational mode decomposition (VMD) and bimodal feature fusion. Firstly, four stray components of LoRa signals are extracted by VMD, which contain unintentional modulation features of the original signals. Then the time-frequency features (VMD-S) and temporal features (VMD-T) are extracted from the four stray components. Consequently, a CNN-based two-channel feature extraction network is built with an attention module and a bimodal feature fusion strategy which realizes the fusion of VMD-S and VMD-T. The experimental results show that the proposed SEI scheme achieves higher identification accuracy in comparison with the scheme without feature fusion at the low signal-to-noise ratio (SNR) region.