Improving Spectrum Sensing Performance by using Eigenvectors

Abstract—In this paper we present a method to improve the performance of eigenvalue-based detection, facilitated with eigenvectors of the sample covariance matrix. We focus on the multi-sensor detection of a single source case. If the channel is constant over adjacent sensing slots, it can be blindly estimated by using the eigenvector associated to the largest eigenvalue on condition of the source’s presence. We introduce a new test where the eigenvector value, computed over some previous auxiliary slots, is properly used by the detection algorithm. The ROC curves show that the new test is able to outperform popular algorithms like the Roy Largest Root Test and the Energy Detection for both PSK and Gaussian sources, and to approach the optimal Neyman-Pearson performance with a very small number of auxiliary slots. Keywords-largest eigenvector; spectrum sensing; cognitive radio; signal detection I