Distributed Bayesian Compressive Sensing Using Gibbs Sampler

Bayesian Compressive Sensing (BCS) observes sparse signal from the statistics viewpoint. In BCS, a Bayesian hierarchy is established utilizing Bayesian inference, thus gives the reconstruction algorithm plenty of robust and flexibility. When dealing with distributed scenario, Bayesian hierarchy is also an effective method. Not only can statistic model be built on the sparse signal itself, but also the inter-correlation between distributed signals can be exploited from statistic viewpoint. Based on BCS, connection between distributed signals is studied and utilized. By adopting spike and slab model, a Bayesian hierarchy including inter-correlation is established to model the distributed compressive sensing (DCS) architecture. With the help of Gibbs sampler, the hierarchy becomes solvable. Simulation results show it has a manifest promotion to reconstruction performance.