Normalization method for online learning on radio access technology identification in cognitive radio
Kenta Asakura, Haruhisa Ichikawa, Yuusuke Kawakita · 2017
The identification of Radio Access Technology (RAT) of Primary User (PU) by Secondly User (SU) is important to avoid interference for spectrum sharing techniques using Cognitive Radio (CR). RATs have become more diversified with the introduction of various kinds of services using wireless communication. Therefore, it is desirable that a RAT identification system, which can easily cope with the diversification of RATs, is applied to cognitive radio. The purpose of this study is to identify multiple RATs in the same frequency band by using online learning. In order to improve the identification accuracy of RATs with similar features, we proposed a normalization method for radio features extracted from the signal's spectrogram. We evaluated the RAT classifier created using the proposed method by calculating the receiver operating characteristics (ROC) curve and the identification accuracy. The results of the ROC curve revealed that the proposed method is effective for some supervised learning. Moreover, the results of the identification accuracy revealed that it could improve the identification accuracy for multiple RATs on the same frequency by nearly 40%.