Distributed Spectrum Sensing for Cognitive Radio Networks Based on the Sphericity Test

Abstract

We consider spectrum sensing in a cognitive radio network with arbitrary numbers of primary and secondary users. Based on the sphericity test, we first analyze the centralized spectrum sensing where all the data available at the secondary users are combined for the signal detection of primary users. We derive accurate approximations for the false alarm and detection probabilities that are also compared against the approximations already available in the literature. Next, we analyze the distributed spectrum sensing where only partial data from each secondary user are used in the signal detection of primary users. Two novel techniques, namely, the multisample sphericity test and metaanalysis, are proposed and analyzed. Instead of sending all the raw data received at the secondary user terminals, in the multisample sphericity test and metaanalysis, only one or two real numbers are required to be sent to a central processor to make a decision about the presence of primary users. Accurate analytical expressions on the false alarm and detection probabilities are derived, and numerical examples are provided to verify their accuracy. Receiver operating characteristic curves are also presented to compare the performance of the proposed methods.