On the validity of Gaussian approximations to exact test statistics of energy detector based spectrum sensing for cognitive radios

Cognitive radio is an emerging technology which offers dynamic and opportunistic access to under utilized spectrum resources. Energy detection is the dominantly used spectrum sensing approach owing to its low computational complexity and ability to identify spectrum holes without requiring a priori knowledge of primary transmission characteristics. In this paper, we discuss the challenges in using the Gaussian approximation to the exact test statistics of energy detector. More importantly, we present an in depth analysis on the validity of the Gaussian approximation under practical sensing scenarios. Specifically, in addition to the number of observed samples, the roles of signal to noise ratio and performance constraints in terms of required probability of detection or allowed false-alarm rate are highlighted when the Gaussian approximation is used instead of the exact test statistics.