Reconstructing geographical-spectral pattern in cognitive radio networks

Abstract

The geographical-spectral pattern of interruptions from primary users within an area is important for upper layer issues like routing and congestion control in cognitive radio networks. The pattern can be considered as an image and can be recovered from reports of secondary users, like random samples for reconstructing an image. Gibbs random fields are used to model the image by employing an energy function to incorporate correlations between neighboring pixels and a priori hyperparameters. Bayesian compressed sensing is then used to reconstruct the image based on the assumption that the image is sparse in a certain transform domain. The performance of the image reconstruction is demonstrated by numerical simulations.