LRS- $$G^2$$ G 2 Based Non-parametric Spectrum Sensing for Cognitive Radio

Abstract

In this paper, a novel non-parametric spectrum sensing scheme in cognitive radio (CR) is proposed based on robust Goodness of Fit (GoF) test. The proposed scheme uses likelihood ratio statistics (LRS-\(G^2\)), from which goodness of fit test is derived. The test is applied assuming different types of primary user (PU) signals such as static or constant, single frequency sine wave and Gaussian signals, whereas different types of channels such as additive white Gaussian noise (AWGN), block fading and time-varying channels. Considering a real time scenario, uncertainty in noise variance is also assumed. The performance of the proposed scheme is shown using receiver operating characteristics (ROC) and it is compared with energy detection (ED) and prevailing GoF based sensing techniques such as Anderson-Darling (AD) sensing, Order Statistic based sensing and Kolmogrov-Smirnov (KS) sensing. It is shown that the proposed scheme outperforms all these prevailing schemes.