Spectrum sensing using principal component analysis

In the recent past considerable research has been performed on blind signal detection techniques that exploit the covariance matrix of the signals received at a cognitive radio (CR). These techniques overcome the noise uncertainty problem of the energy detection (ED) method and can even perform better than ED for correlated signals. Contrary to the previous work where the main evaluation technique has been theoretical analysis and simulations, in this paper we use Software defined radios (SDRs) with correlated signal reception capability to evaluate the sensing performance of the existing covariance based detection (CBD) techniques. The existing techniques considered in this work are; Covariance absolute value (CAV), Maximum-minimum eigenvalue (MME), Energy with minimum eigenvalue (EME) and Maximum eigenvalue detection (MED). Most importantly this paper presents a novel technique for blind signal detection that uses Principal Component (PC) Analysis. The PC based signal detection algorithm and the CBD algorithms are tested in a real scenario with SDRs and their sensing performance is compared. The PC algorithm outperforms the MED and EME algorithms under all conditions and it performs better than the MME and CAV algorithms under certain conditions.