Intelligent selection of threshold in covariance-based spectrum sensing for cognitive radio networks

Abstract

The radio spectrum sensing has been an important issue of research in cognitive radio networks over the last decade and the appropriate selection of threshold plays a crucial role in the process of spectrum sensing. The conventional channel sensing methods generally employ a fixed threshold, which is either based on the principle of constant false alarm rate (CFAR) or the principle of constant detection rate (CDR). The sensing performance of these schemes degrades under low signal to noise ratio (SNR) and noise uncertainty. The problem of noise uncertainty occurring in energy detection (ED) based spectrum sensing method can be overcome by using covariance-based spectrum sensing scheme. However, the performance of covariance based spectrum sensing degrades at low SNR. This paper proposes a covariance-based channel sensing method, where the adaptive threshold is selected in an intelligent manner to minimize the probability of error with sufficient protection to primary user (PU). First, an adaptive threshold is derived by considering both probability of detection and probability of false alarm such that the total decision error probability is minimized. This adaptive threshold is then considered along with two other thresholds based on CFAR and CDR schemes, for the final selection of threshold such that the protection to PU is maximized. The proposed approach also provides the minimum number of samples required for reliable spectrum sensing. As shown by the simulation results, the proposed scheme exhibits better detection performance compared to ED based schemes as well as the existing covariance-based detection method in terms of probability of detection and probability of decision error.