Blind spectrum sensing based on the ratio of mean square to variance

Abstract

Anderson-Darling (AD) sensing, characteristic function (CF) sensing and order statistic (OS) sensing are three common spectrum sensing (SS) methods based on goodness of fit (GOF) testing. However, AD and OS sensing needs the prior information of noise variance; CF and OS sensing have high computation complexity. To circumvent those difficulties, in this paper, the ratio of the mean square to variance (RM2V) of the samples, after deriving its probability density function (PDF), is employed as a test statistic to detect the availability of the vacant spectrum in the cognitive radio (CR) system. Then a blind SS method based on RM2V is proposed, which is dubbed as RM2V sensing, and its exact theoretical threshold is obtained via the derived PDF of RM2V. The performance of RM2V sensing is evaluated by theoretical analysis and Monte Carlo simulations. Comparing with the conventional energy detection (ED), AD, CF and OS sensing, RM2V sensing, with no need of noise variance, has advantages from the aspect of computation complexity and detection performance.