Analytical and learning-based spectrum sensing over channels with both fading and shadowing

In this paper, sensing performance of an energy detector (ED) for local and collaborative detection scenarios is investigated in unreliable environments dominated by multipath fading and shadowing effects. The channel is modeled by using K_G distribution for Nakagami-m multipath fading and lognormal shadowing. Novel analytical expressions are firstly derived for the average detection probability for both fading and fading/shadowing cases. The analysis is then extended to the conventional fusion strategies i.e. decision fusion and data fusion. The performance of decision fusion scheme under the generalized k-out-of-n fusion rule has been investigated. In data fusion method, the analytical expressions are derived for two combining schemes including maximal ratio combining (MRC) and square law combining (SLC). Further, a reliable fusion scheme based on a learning algorithm is proposed. In this fusion mechanism, the Least Mean Square (LMS) algorithm is utilized to enhance reliability of the final decision regarding presence or absence of primary user (PU). The analytical results are validated by numerical computations and Monte-Carlo simulations along with the performance of the proposed learning-based fusion scheme.