Narrowband Radio Device Identification Based on RF Fingerprint Features

This paper proposes a rapid and efficient method for extracting narrowband device radio frequency fingerprints (RFF), which remains robust in the presence of time-varying channels and baseband voice signals. Furthermore, the transient signal and steady-state features are combined with the signal-to-noise ratio (SNR) for integrated training, significantly enhancing recognition accuracy. Our experimental results demonstrate that at an SNR of 46 dB, the RFF identification system achieves a classification accuracy exceeding 93%. Notably, the characteristics of the steady-state signal effectively suppress noise, enabling the RFF identification system to attain nearly 90% classification accuracy even at low SNR and time-varying conditions.