– M. Kowalski (L2S) : Social Sparsity : application to audio inpainting.

Carte non disponible

Date/heure
Date(s) - 28/03/2013
14 h 00 min - 15 h 00 min

Catégories Pas de Catégories


Social Sparsity: application to audio inpainting. By Matthieu Kowalski, L2S. Abstract: Audio inpainting problem is under consideration, using iterative thresholding algorithms build on the “social sparsity” principle. First, we present new shrinkage operators, allowing one to take into account the neighborhood of time-frequency coefficients. Then, the audio declipping problem is formulated as a unconstrained convex optimization problem, but taking into account an inportant hypothesis of audio declipping: reconstructed samples must be greater than the clipping threshold. The structured thresholding operators, such as the windowed group-Lasso or the persistent empirical wiener, are embedded into iterative algorithms, and we show on experimental results the SNR improvement compared to a more conventional Lasso. We also compare the results to the state of the art audio declipping. Download slides

– M. Kowalski (L2S) : Social Sparsity : application to audio inpainting.

Carte non disponible

Date/heure
Date(s) - 28/03/2013
14 h 00 min - 15 h 00 min

Catégories Pas de Catégories


Social Sparsity : application to audio inpainting.\n\nBy Matthieu Kowalski, L2S.\n\nAbstract :\nAudio inpainting problem is under consideration, using iterative\nthresholding algorithms build on the “social sparsity” principle.\nFirst, we present new shrinkage operators, allowing one to take into\naccount the neighborhood of time-frequency coefficients. Then, the\naudio declipping problem is formulated as a unconstrained convex\noptimization problem, but taking into account an inportant hypothesis\nof audio declipping : reconstructed samples must be greater than the\nclipping threshold. The structured thresholding operators, such as the\nwindowed group-Lasso or the persistent empirical wiener, are embedded\ninto iterative algorithms, and we show on experimental results the SNR\nimprovement compared to a more conventional Lasso. We also compare the\nresults to the state of the art audio declipping.\n\nDownload slides[