par Determe, Jean-François 
;Louveaux, Jérôme;Jacques, Laurent;Horlin, François
Référence (19-20 May, 2016: Louvain-la-Neuve, Belgium), Proc. of the 37th WIC Symposium on Information Theory in the Benelux, WIC
Publication Publié, 2016-05
;Louveaux, Jérôme;Jacques, Laurent;Horlin, François Référence (19-20 May, 2016: Louvain-la-Neuve, Belgium), Proc. of the 37th WIC Symposium on Information Theory in the Benelux, WIC
Publication Publié, 2016-05
Publication dans des actes
| Résumé : | This paper addresses the problem of recovering several sparse signals acquiredby means of a noisy linear measurement process returning fewer observationsthan the dimension of the sparse signals of interest. The proposed signal modelassumes that the noise is additive and Gaussian. Within the aforementionedframework, theoretical developments making use of the theory of compressivesensing show that sparse signals with similar supports can be jointly and reliablyrecovered by means of the greedy algorithm entitled simultaneous orthogonalmatching pursuit (SOMP) provided that the linear measurements are appropriatelydesigned. A variant of SOMP weighting each measurement vector accordingto its noise level is then presented. Finally, simulations confirm the benefits ofweighting the measurement vectors in SOMP and show that the optimal weightspredicted by the theory match the empirical ones under proper conditions. |
