Towards Robust Spectrum Sensing in Cognitive Radio Networks

Abstract

This thesis focuses on multi-antenna assisted energy based spectrum sensing. The studies leading to this thesis have been motivated by some practical issues with energy based detection. These include the noise uncertainty problem at the secondary receiver, the presence of multiple active primary users in cognitive cellular networks, the existence of unknown noise correlations and detection in the low signal-to-noise ratio regime.

In this thesis, the aim is to incorporating these practical concerns into the design of spectrum sensing algorithms. To this end, we propose the use of various detectors that are suitable for different scenarios. We consider detectors derived from decision-theoretical criteriors as well as heuristic detectors. We analyze the performance of the proposed detectors by deriving their false alarm probability, detection probability and receiver operating characteristic. The main contribution of this thesis consists of the derived closed-form performance metrics. These results are obtained by utilizing tools from multivariate analysis, moment based approximations, Mellin transforms, and random matrix theory.

Numerical results show that the proposed detectors have indeed resolved the concerns raised by the above practical issues. Some detectors could meet the needs of one of the practical challenges, while others are shown to be robust when several practical issues are taken into account. The use of detectors constructed with decision-theoretical considerations over the heuristically proposed ones is justified as well.