A New Radio Frequency Fingerprint Identification Method Based on CEEMDAN

Radio Frequency Fingerprint Identification (RFFI) is an emerging technology for device identification that relies on the emitter-specific inherent impairments. This paper presents a new RFFI method based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN). CEEMDAN decomposes the signal into a collection of intrinsic mode functions (IMFs) accurately. Signals carrying with emitter fingerprints are decomposed by CEEMDAN and fingerprint features are extracted after Hilbert Transformation, namely energy entropy, first-order and second-order moment. Support Vector Machine (SVM) is used as classifier. The training and test sets are randomly disrupted and divided according to the ratio of 8: 2. Simulation results show that the proposed method has excellent performance. The utilization of CEEMDAN before extracting fingerprint features from the Hilbert spectrum greatly enhances their distinguishability. when 3 emitters are to be recognized, the identification accuracy attains 95% for SNR=9 dB. When the number of emitters up to 5, it reaches 85% under 12 dB in the additive white Gaussian noise (AWGN) channel.