A review of cosparse signal recovery methods applied to sound source localization

National audienceThis work aims at comparing several state-of-the-art methods for cosparse signal recovery, in the context of sound source localization. We assess the performance of ve cosparse recovery algorithms: Greedy Analysis Structured Pursuit, l1 and joint l1,2 minimization, Structured Analysis Iterative Hard Thresholding and Structured Analysis Hard Thresholding Pursuit. In addition, we evaluate the performance of these methods against the sparse synthesis paradigm, solved with corresponding joint l1,2 minimization method. For this evaluation, the chosen applicative showcase is sound source localization from simulated measurements of the acoustic pressure eld.L'objectif de cet article est de comparer plusieurs m ethodes de l' etat de l'art pour la reconstruction coparcimonieuse de signaux, dans le contexte de la localisation de sources sonores. Nous evaluons les performances de cinq algorithmes de reconstruction coparcimonieuse : l'algorithme de "Greedy Analysis Structured Pursuit", les minimisations l1 et l1,2 jointe, ainsi que les algorithmes "Structured Analysis Iterative Hard Thresholding" et "Structured Analysis Hard Thresholding Pursuit". Nous comparons egalement ces algorithmes a l'approche de parcimonie a la synth ese, que nous r esolvons par la minimisation jointe l1,2 correspondante. Nous illustrons nos r esultats dans le cadre d'une application a la localisation de sources sonores, r ealise sur des simulations de mesures de champs de pression acoustique