Bayesian estimators for cooperative spectrum sensing in cognitive radio networks

In this paper we consider centralized cooperative spectrum sensing (SS) techniques for cognitive radio networks using energy detector scheme. In light of the requirements imposed by centralized SS methods such as Maximum Ratio Combining (MRC), namely the estimation and transmission of the signal-to-noise ratio (SNR) on each secondary user, as well as the transmission of the exact energy level to the fusion center, we aim to analyze and compare alternative Bayesian techniques to tackle the trade-off between operation overhead and detection performance. Based on the statistics of classic energy detection scheme, we consider three Bayesian SS estimators, the Weighted Bayesian (WB) estimator, the Gaussian Mixture (GM) estimator and a Naive Bayes (NB) classifier. We compare the results of these techniques with well established cooperative SS schemes, such as the MRC, AND and OR rule. Through the use of Monte Carlo simulations, our results shows that the Bayesian techniques evaluated offer similar performance in terms of area under the ROC curve (AUC) regarding the optimum MRC technique, while requiring less operation overhead for the secondary users.