Bayesian Compressive Sensing in Radar Systems

Compressive Sensing (CS) is presented in a Bayesian framework for realistic radar cases whose likelihood or priors are usually non-Gaussian. Its sparse-signal processing is modelbased and detection-driven, and also done numerically using Monte-Carlo methods. This approach aims for the stochastic description of sparse solutions, and the flexibility to use any prior information on signals or on data acquisition, as well as any distribution of noise or clutter, without the need for a closed analytic form of the Bayesian solution. This flexible Bayesian CS is shown by comparison with its closed-form predecessors.