Cooperative Spectrum Sensing with Partial CSI

Spectrum sensing is mandatory in Cognitive Radio systems, and is used in order to identify spectrum opportunities, and to guarantee that it does not cause unacceptable interference to the license owner. Since a single sensor may be in fading or shadowing, cooperative sensing among multiple sensors which experience uncorrelated fading is required to guarantee reliable sensing performance. In this paper we develop efficient centralized statistical algorithms for cooperative spectrum sensing in a cooperative based cognitive radio network. We consider a stochastic model where data signals are sent from the Primary user Base Station (PBS). The data is intercepted via a set of sensors and re-transmitted to Secondary user Base Station (SBS). The SBS has only partial Channel State Information (CSI) knowledge of the wireless channels. In order to obtain the optimal decision rule based on Likelihood Ratio Test (LRT), the marginal likelihood under each hypothesis needs to be evaluated pointwise. Therefore, evaluation of the LRT cannot be performed analytically due to the intractability of the integrals which involve multidimensional marginalisation. Instead, we present algorithms to perform approximation of the marginal likelihood, and obtain sub-optimal solutions. Performance is evaluated via analytic bounds and via numerical simulations.