Joint Local Quantization and Linear Cooperation in Spectrum Sensing for Cognitive Radio Networks

In designing cognitive radio networks (CRNs), protecting the license holders from harmful interference while maintaining acceptable quality-of-service (QoS) levels for the secondary users is a challenge effectively mitigated by cooperative spectrum sensing schemes. In this paper, cooperative spectrum sensing in CRNs is studied as a three-phase process composed of local sensing, reporting, and decision/data fusion. Then, a significant tradeoff in designing the reporting phase, i.e., the effect of the number of bits used in local sensing quantization on the overall sensing performance, is identified and formulated. In addition, a novel approach is proposed to jointly optimize the linear soft-combining scheme at the fusion phase with the number of quantization bits used by each sensing node at the reporting phase. The proposed optimization is represented using the conventional false alarm and missed detection probabilities, in the form of a mixed-integer nonlinear programming (MINLP) problem. The solution is developed as a branch-and-bound procedure based on convex hull relaxation, and a low-complexity suboptimal approach is also provided. Finally, the performance improvement associated with the proposed joint optimization scheme, which is due to better exploitation of spatial/user diversities in CRNs, is demonstrated by a set of illustrative simulation results.