Abstract:
Uncertainties in spectrum sensing are the fundamental problems in cognitive radio. The prevailing effect of impulse, thermal and environmental noise on existing sensing techniques such as energy detection, matched filter and cyclostationary feature detections has necessitated the need for
comparative studies of Eigenvalues based sensing techniques in a noisy system. Two among the Eigenvalue based spectrum sensing techniques, the Maximum Minimum Eigenvalue (MME) based and the Energy Minimum Eigenvalue (EME) based, have outstanding performances in the presence of noise signal as against the popular Energy Detection (ED) technique. Using SNR values between 0dB and -20dB and multiple SU receivers, the result showed that the probability
of detection for MME technique (P dMME ) is 92% which is higher than the probability of detection for EME technique (P dEME ) of 86%, while ED technique performs very poorly with the probability of detection (P dED ) of 1.8%. However, for fewer numbers of SU’s receivers, MME slightly performed better. Also, MME attains convergence faster than EME technique, MME attained convergence at SNR of -6dB but EME attained convergence at SNR of -2dB. In this work, the
three techniques, the MME, EME and ED methods were applied to broadcasting frequency bands in a Rayleigh channel at low Signal to Noise Ratio (SNR), and their comparative performance was analysed. The algorithms for the three detection techniques were implemented in MATLAB, with simulated Primary User’s (PU) signal in the 30 - 833 MHz broadcasting frequency bands. The
primary users’ signal samples received together with noise by Secondary Users (SUs) was used to get the covariance matrix of the received signal. The difference between the statistical properties of the covariance matrix of the received signal and the noise signal gave a reliable detection test without requiring the knowledge of the primary user’s parameters or the received noise variance. The ratio of the maximum and minimum Eigenvalues was derived from the covariance matrix todetermine the detection test statistics for MME, while the ratio of the received signal’s energy with the minimum Eigenvalue was used for EME detection test statistics. The detection tests were compared to the determined thresholds. The thresholds for the Eigenvalue based techniques (EVB) were determined using random matrix theories (RMT), that is, the probability of false alarm (P fa ,) , number of samples (N S ) and smoothing factor (L) while the ED threshold was determined based on the received noise power.
