Statistical Spectrum Sensing in Cognitive Radio

Abstract

Statistical spectrum sensing is a promising method which can reliably detect the primary users while requiring little prior information in cognitive radio networks. In this paper, we present an overview of sensing methods based on Goodness-of-Fit tests. We discuss the performance of Energy Detector (ED) sensing, Anderson Darling (AD) sensing, Cram'er VonMises(CVM) sensing and Order Statistic (OS) sensing and we compare the results using Monte-Carlo simulations. It is shown that OS sensing outperforms ED sensing, CVM sensing and AD sensing. Next it is shown through simulations that the OS test statistic does not provide maximum probability of detection for a desired probability of false alarm and results are provided showing the regions of high probability of detection for desired probability of false alarm.