Quickest Spectrum Sensing over Correlated Channels

Quickest spectrum sensing seeks to optimize a balance between two opposing performance measures in spectrum sensing, one being the delay in identifying spectrum opportunities, and the other being the quality of the decision. The existing spectrum sensing approaches formed based on quickest detection theory rely on the assumption that the occupancy states of different spectrum bands over a wideband spectrum are statistically independent. This is an assumption that cannot be met in practice, especially in broadband communication schemes in which radio channels are dynamically grouped in bundles and allocated to different users based on the users' traffic needs. As a result of such channel grouping and allocation the occupancy states of the channels, especially the adjacent ones, are correlated. This paper, in contrast to the existing literature on quickest spectrum sensing, considers a wideband spectrum in which the occupancy states of different channels follow a pre-specified dependency kernel. The objective is to design the quickest spectrum sensing approach for identifying spectrum holes, which aims to minimize the average delay in identifying spectrum opportunities while assuring, in parallel, certain guarantees on the quality of the decision. The closed-form expression of the optimal sensing scheme are delineated and are shown to have low computational complexity.