Multi-Task Bayesian Compressive Sensing Exploiting Intra-Task Dependency

In this letter, we propose a multi-task compressive sensing algorithm for the reconstruction of clustered sparse entries based on hierarchical Bayesian framework. By extending a paired spike-and-slab prior to a general multi-task model, the proposed algorithm has the capability of modeling both inter-task and intra-task dependencies of the observation data. The latter is achieved by imposing a clustered prior on non-zero entries and finds applications in radar where targets exhibit spatial extent. Simulation results verify that the proposed algorithm outperforms state-of-the-art group sparse Bayesian learning algorithms.