On the effectiveness of the Gaussian approximation in cognitive radio systems with fading channels

In cognitive radio (CR) systems, the Gaussian approximation (GA) has been widely used in determining the optimal threshold for spectrum sensing owing to its analytical simplicity because the GA has been known to be accurate when the number of samples is large. However, this paper shows that if the received signals are weak, the determined sensing threshold by the GA causes significant performance degradation of CR systems even when the number of received samples is large. We analytically quantify the degradation in terms of throughput efficiency and show that the threshold for spectrum sensing should be determined by an exact distribution which is derived in this paper if the received signals are weak.