Recovery guarantees for mixed norm l-(p1, p2) block sparse representations

Fardin Afdideh 1, * Ronald Phlypo 1 Christian Jutten 1
* Auteur correspondant
1 GIPSA-VIBS - VIBS
GIPSA-DIS - Département Images et Signal
Abstract : In this work, we propose theoretical and algorithmic-independent recovery conditions which guarantee the uniqueness of block sparse recovery in general dictionaries through a general mixed norm optimization problem. These conditions are derived using the proposed block uncertainty principles and block null space property, based on some newly defined characterizations of block spark, and (p, p)-block mutual incoherence. We show that there is improvement in the recovery condition when exploiting the block structure of the representation. In addition, the proposed recovery condition extends the similar results for block sparse setting by generalizing the criterion for determining the active blocks, generalizing the block sparse recovery condition, and relaxing some constraints on blocks such as linear independency of the columns.
Type de document :
Communication dans un congrès
European Signal Processing Conference (EUSIPCO), Aug 2017, Budapest, Hungary. IEEE, 2016, Signal Processing Conference (EUSIPCO), 2016 24th European. 〈10.1109/EUSIPCO.2016.7760274〉
Liste complète des métadonnées

http://hal.univ-grenoble-alpes.fr/hal-01658328
Contributeur : Ronald Phlypo <>
Soumis le : jeudi 7 décembre 2017 - 15:16:33
Dernière modification le : lundi 9 avril 2018 - 12:22:44

Identifiants

Collections

Citation

Fardin Afdideh, Ronald Phlypo, Christian Jutten. Recovery guarantees for mixed norm l-(p1, p2) block sparse representations. European Signal Processing Conference (EUSIPCO), Aug 2017, Budapest, Hungary. IEEE, 2016, Signal Processing Conference (EUSIPCO), 2016 24th European. 〈10.1109/EUSIPCO.2016.7760274〉. 〈hal-01658328〉

Partager

Métriques

Consultations de la notice

90