SVM and Dimensionality Reduction in Cognitive Radio with Experimental Validation

Abstract

There is a trend of applying machine learning algorithms to cognitive radio. One fundamental open problem is to determine how and where these algorithms are useful in a cognitive radio network. In radar and sensing signal processing, the control of degrees of freedom (DOF)---or dimensionality---is the first step, called pre-processing. In this paper, the combination of dimensionality reduction with SVM is proposed apart from only applying SVM for classification in cognitive radio. Measured Wi-Fi signals with high signal to noise ratio (SNR) are employed to the experiments. The DOF of Wi-Fi signals is extracted by dimensionality reduction techniques. Experimental results show that with dimensionality reduction, the performance of classification is much better with fewer features than that of without dimensionality reduction. The error rates of classification with only one feature of the proposed algorithm can match the error rates of 13 features of the original data. The proposed method will be further tested in our cognitive radio network testbed.