Speaker : Professor Alexander Powell( Vanderbilt University)
Title : Compressed sensing and Sobolev duals of Gaussian random matrices
Time : 2012-06-25 (Mon) 14:10 -
Place : Seminar Room 722, Institute of Mathematics (NTU Campus)
Abstract: Compressed sensing addresses the problem of representing sparse signals in high dimensions using a small number of linear measurements. We shall consider the problem of quantization (analog-to-digital conversion) for compressed sensing measurements. We show that the class of sigma-delta algorithms performs well in this setting and we prove error bounds to quantify the improvement over standard methods. Our approach crucially makes use of special non-canonical dual frames, known as Sobolev duals, associated to Gaussian random matrices.