Performance analysis of eigenvalue-based sensing algorithm with Monte-Carlo threshold

Abstract

Eigenvalue-based spectrum sensing algorithms, such as the maximum-minimum eigenvalue (MME) algorithm and the Marčhenko-Pastur (MP) law based algorithm, are based on the asymptotic behavior of large random matrices and have very high sensing performance with an appropriate threshold. The advantage of such algorithms is that they can work very well without the estimation of noise variance, and this feature is very attractive for practical applications because of the hardness of obtaining an exact noise variance. In practical applications, threshold-setting is the key problem of such algorithms and it is important to find a simple and efficient way to make it work well with any specific dimensions (i.e. the sizes of samples and transceivers). In this paper, a Monte-Carlo threshold is provided, which shows how eigenvalue-based spectrum sensing algorithm can work well with the new threshold for any specific dimensions. Performance analysis over the E-UTRA channel model in 3GPP LTE demonstrate that, compared with the original MME detection and the MP-law-based detection, as well as the classical energy detection, the improved scheme with Monte-Carlo threshold offers superior detection performance.