Radio Frequency Fingerprint Identification Based on High-Low Frequency Orthogonal Constraints

Radio frequency fingerprinting is an effective technical method for device authentication based on the radiation characteristics of devices. However, the characteristic energy of radio frequency fingerprints is weak and prone to noise interference, which leads to degraded device identification performance under low signal-to-noise ratio (SNR) conditions. To solve this problem, this paper proposes a radio frequency fingerprinting method based on high-low frequency orthogonal constraints, aimed at improving the identification performance of ZigBee devices under low SNR. This method enhances the spectral representation capability of signals by introducing a residual network with high-low frequency grouped convolution, thereby improving the noise robustness of the model. Meanwhile, a learnable high-low frequency filter is designed, and orthogonal constraint loss is used to constrain high-low frequency components. This strengthens the decoupling and discriminability of features, avoids interference between high and low frequency features, and improves the model’s sensitivity to device differences. Experimental results show that the proposed method achieves an identification accuracy of 84.07% on a dataset of ZigBee device signals actually collected in real environments, effectively improving the identification performance.