RF Fingerprint Extraction Method Based on CEEMDAN and Multidomain Joint Entropy

Specific emitter identification (SEI) can distinguish communication radio emitters with the fingerprint features carried by received signal, and this technology has been widely used in military civilian fields. However, real electromagnetic environment, number of is large signal-to-noise ratio (SNR) low, which leads to poor nonlinear analysis SEI a single domain. Therefore, combining exploration multiple domains spatial information resources, paper proposed frequency (RF) extraction method based on complementary ensemble empirical mode decomposition adaptive noise (CEEMDAN) multidomain joint entropy. The an attempt further domains. Firstly, considering nonstationarity article adopts CEEMDAN decompose signal intrinsic functions (IMF). Then, decomposed represented spaces multidimensional phase space reconstruction technique. Nonlinear original performed spaces: differential approximate entropy space, singular spectral power space. Finally, support vector machine (SVM) adopted classification stage. To demonstrate robustness method, verified universal software peripheral (USRP) dataset Northeastern University public dataset. In terms accuracy, performs 98.5% accuracy 5-class USRP It also 94.7% 16-class experimental results show that stable performance more than 85% recognition rate SNR above 5dB.