Variational Bayesian learning technique for spectrum sensing in cognitive radio networks

Abstract

The successful implementation of dynamic spectrum access in cognitive radio networks requires that the secondary user has an autonomous knowledge of the true status of the licensed user activities. This paper investigates and proposes a robust blind spectrum sensing technique that is based on the variational Bayesian learning for Gaussian mixture model framework for use in multi-antenna cognitive radio networks. The results obtained from the proposed scheme, averaged over 1000 Monte-Carlo simulations show that a probability of detection greater than 90% is achievable at the signal - to - noise ratio (SJVR) of -18 dB when the false alarm probability is kept at less than 10%. An interesting feature of the proposed scheme is its ability to determine the number of active licensed users.