Secondary User QoE Enhancement Through Learning Based Predictive Spectrum Access in Cognitive Radio Networks

Quality of experience (QoE) of a secondary spectrum user is mainly governed by its spectrum utilization, the energy consumption in spectrum sensing and the impact of channel switching in a cognitive radio network. It can be enhanced by prediction of spectrum availability of different channels in the form of their idle times through historical information of primary users’ activity. Based on a reliable prediction scheme, the secondary user chooses the channel with the longest idle time for transmission of its data. In contrast to the existing method of statistical prediction, the use and applicability of supervised learning based prediction in various traffic scenarios have been studied in this paper. Prediction accuracy is investigated for three machine learning techniques, artificial neural network based Multilayer Perceptron (MLP), Support Vector Machines (SVM) with Linear Kernel and SVM with Gaussian Kernel, among which, the best one is chosen for prediction based opportunistic spectrum access. The results highlight the analysis of the learning techniques with respect to the traffic intensity. Moreover, a significant improvement in spectrum utilization of the secondary user with reduction in sensing energy and channel switching has been found in case of predictive dynamic channel allocation as compared to random channel selection.