Skip to Main content Skip to Navigation
Preprints, Working Papers, ...

Étude d'un algorithme d'optimisation pour le fading temps-fréquence

Marina Krémé 1, 2 Bruno Torrésani 1 
2 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication
Inria Nancy - Grand Est, LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
Abstract : We address the problem of filtering localized time-frequency components in signals. The problem is formulated as a minimization of an appropriate quadratic form, which involves a data fidelity term on the short-time Fourier transform outside the support of the undesirable component and an energy penalty term inside the support. We study two resolution methods: a spectral method and a quasi-Newton-like method (BFGS). The latter involve operators called Gabor multipliers. We exploit random projection techniques to approximate these operators. We study and compare the theoretical complexity of a classical BFGS with a BFGS combined with random projections. We compare the computation time of these two methods on several audio signals. We also compare their computation time to that of the spectral method. The performances of all these approaches are evaluated and compared on several audio signals.
Complete list of metadata
Contributor : AMA MARINA KREME Connect in order to contact the contributor
Submitted on : Wednesday, June 22, 2022 - 3:52:13 PM
Last modification on : Saturday, June 25, 2022 - 3:32:08 AM


Files produced by the author(s)


  • HAL Id : hal-03701278, version 1


Marina Krémé, Bruno Torrésani. Étude d'un algorithme d'optimisation pour le fading temps-fréquence. 2022. ⟨hal-03701278⟩



Record views


Files downloads