Distributed Adaptive Largest Eigenvalue Detection with SNR Weighted Observations

Abstract

Cognitive radio (CR) systems have to be able to detect the presence of a primary user (PU) signal by sensing the spectrum area of interest. Due to radiowave propagation effects like fading and shadowing, spectrum sensing is often complicated, because the PU signal can be attenuated in a particular area. In this paper, we explore a distributed spectrum sensing approach that exploits the largest eigenvalue of correlation matrices (CMs) that are adaptively estimated, based on the combine and adapt least (CTA) type of diffusion method with no fusion center (FC). More specifically, CR nodes exchange also observations with a subset of neighbouring nodes and combine the neighbouring observations based on the locally estimated signal to noise ratio (SNR) values. We analyse the resulting detection performance and verify the theoretical findings through simulations.