Learning-based predictive dynamic spectrum access framework: a practical perspective for enhanced QoE of secondary users

In a cognitive radio environment, the optimality in channel selection by a secondary user (SU) is directly dependent on its spectrum-sensing efficiency, and quality of experience (QoE) in terms of the channel-switching frequency (CSF) and the interference caused to the primary users (PUs). Modelling the spectrum through statistical methods becomes, sometimes, difficult due to the lack of a-priori information of the PU activity. This work proposes a framework for learning-based prediction of the future idle times of the PUs thereby opportunistically allocating the channel with enhanced QoE of SUs. The idea is to minimise the spectrum-sensing energy requirement by sensing only if the channel is predicted to be idle, thereby reducing the CSF and mitigating the SU–PU interference. Initially, the authors have tested the accuracy of the prediction approach in various traffic scenarios for a single PU channel case. Later, it is extended to the multiple channel case for a particular data traffic. Furthermore, a practical scenario has been considered where the efficacy of the proposed framework is validated for PU data traffic in GSM and ISM bands. The results highlight the practicability of prediction-based opportunistic dynamic spectrum access with improvement in the SU QoE over random channel selection.