Traffic-pattern aware opportunistic wireless energy harvesting in cognitive radio networks

Each traffic application follows a unique packet transmission pattern, which can be used to identify traffic applications. Current literature on cognitive radio networks (CRNs) assume that primary user (PU) channel idle and busy time probabilities are predefined and known. However, in practice, those probabilities are application-specific. In this paper, from application-dependent traffic features, we propose a Bayesian nonparametric method to detect and classify primary transmitter's (PT's) applications to estimate the secondary user (SU) spectral access and energy harvesting opportunities related to each application. To this end, the Dirichlet process mixture model (DPMM) is employed and a mean-field variational method is proposed. We demonstrate the effectiveness of the proposed method by both simulations and experiment data obtained from the WiMax networks.