Kurtosis based spectrum sensing in cognitive radio

In this paper, we consider spectrum sensing in cognitive radio networks based on higher order statistics. The kurtosis, a fourth order statistic, which is a measure of deviation from Gaussianity, is used as a detection statistic. An optimum threshold is set up based on the Neyman–Pearson criterion and an analytical expression for upper bound on probability of miss is derived for a single pair of primary and secondary users. Further, we also propose a collaborative spectrum sensing scheme for more than one secondary user and it is shown by simulations that the proposed kurtosis based method outperforms the energy based spectrum hole detection method significantly.