Optimal Multi-Cycle Cyclostationarity-based Spectrum Sensing for Cognitive Radio Networks

Reliable detection of primary users (PUs) in the presence of interference and noise is a crucial problem in cognitive radio networks. To address above issue, cyclostationary feature detectors that can robustly detect weak primary signals have been proposed in the literature. Among different candidates, in this paper we focus on the method which is based on asymptotic properties of cyclic autocorrelation estimates. The objective is to establish some optimal strategies for multi-cycle cyclostationary detection method, within which the linear combination of multiple independent test statistics corresponding to different cycle frequencies is computed. The optimality criteria considered here is the deflection coefficient and modified deflection coefficient maximization. In each case, we derive analytical approximations for the distribution of proposed test statistic under null hypothesis. Also, we study the agreement between empirically estimated distribution and proposed analytical approximation. In addition, we analytically characterize the impact of channel fading on the cyclostationarity of received signals and verify our analysis via simulation. Simulation results confirm the asymptotic detection performance of proposed optimal methods, as compared with suboptimal detectors.