Cooperative Spectral Covariance Sensing under Correlated Shadowing

This paper investigates the theoretical limits of white space sensing in a cognitive radio (CR) network limited by channel correlation. In a log-normal shadowing channel, the received signal power is correlated based on the distance between the sensors and this makes sensing the presence of a signal difficult, even with several cooperative sensors. In the proposed system, each sensor uses the spectral covariance sensing (SCS) algorithm to detect the primary signal and then sends its decision statistic to the base station (BS). The BS, using the Neyman Pearson log-likelihood ratio test, makes the final decision. We analyze the probability of a false alarm (P_FA) and compare it with that of the cooperative energy detector. We show that an asymptotic lower bound on the P_FA is an order of magnitude lower than that of the energy detector. We also demonstrate improvements in the cooperation gain in terms of the effective number of independent sensors, and the required number of sensors for a given detection metric. The results of this paper show that cooperative SCS detection has far better white space sensing properties than cooperative energy detection in correlated channels.