Higher Order Eigenvalue Moment Ratio Based Spectrum Sensing in Cognitive Radio

Abstract

The Radio spectrum is a most precious natural resource in this era of development in wireless technologies. The spectrum utilization can be enhanced to a great extent by the use of cognitive radio (CR) technique.Spectrum sensing is a technique to find out the empty spectrum band. In conventional spectrum sensing technique such as energy detection and matched filter detection, full or partial information of the primary user signal characteristics, noise variance information etc. are required at the cognitive radio receiver. But such information may not be available in real cognitive radio network. So various blind spectrum sensing technique that require no primary signal knowledge or noise power is presented.Out of various blind detection techniques the methods which are discussed are eigenvalue arithmetic to geometric mean(AGM), Scaled largest Eigenvalue (SLE) and Eigenvalue Moment Ratio (EMR). In order to achieve efficient spectrum utilization, spectrum sensing should have short sensing period. Short sensing period reduces interference and is achieved using small data samples. Out of the AGM and SLE, EMR has better performance in small sample environment. Further, a new blind spectrum sensing is discussed which work better in data limited environment. The proposed method distribution is determined by Random Matrix Theory (RMT) when signals does not exist and this method is able to predict the theoretical limit. As proposed method detection is proposed from RMT concept and it uses eigenvalues of the signal for detection,so it can be considered superior to other blind detection techniques such as AGM, SLE and EMR in small sample scenarios. Asymptotic distribution of proposed technique statistic is obtained in presence of primary signal. The theoritical probability of detection is also derived. In computation of decision statistic of proposed method is obtained using matrix trace operation and internal Frobenius product operation instead of eigenvalue decomposition(EVD) technique, which uses less computational time. The simulation results shows that the proposed technique outperforms other techniques such as AGM,SLE and EMR in small sample scenarios.