Because of RF power amplifier’s nonlinearity and the discrete distribution of electronic components value, it is possible to form individual RF fingerprints. How to identify individual RF fingerprints is the key problem because received RF signals are weak to be affected by channel noises. In order to identify steady RF signal fingerprints, this paper proposed a tensors-based method for obtaining identification patterns from high order cummulants. Firstly, high order cummulants tensors of received RF signal were formed as distinguished patterns of RF Fingerprints. Secondly, these cummulants tensors were passed to classifiers as training data and testing data. The RF Fingerprint classifiers based on SVM and RBF networks are compared with their performance. Finally, the identification results of two classifiers were obtained. The results showed that the method is effective for RF Fingerprints Identification.