Sparsity-based audio declipping methods: selected overview, new algorithms, and large-scale evaluation

Recent advances in audio declipping have substan- tially improved the state of the art. Yet, practitioners need guidelines to choose a method, and while existing benchmarks have been instrumental in advancing the field, larger-scale exper- iments are needed to guide such choices. First, we show that the clipping levels in existing small-scale benchmarks are moderate and call for benchmarks with more perceptually significant clipping levels. We then propose a general algorithmic framework for declipping that covers existing and new combinations of variants of state-of-the-art techniques exploiting time-frequency sparsity: synthesis vs. analysis sparsity, with plain or structured sparsity. Finally, we systematically compare these combinations and a selection of state-of-the-art methods. Using a large-scale numerical benchmark and a smaller scale formal listening test, we provide guidelines for various clipping levels, both for speech and various musical genres. The code is made publicly available for the purpose of reproducible research and benchmarking.

Keywords

audio declipping time-frequency structured sparsity listening test sparsity

Domains

Sound [cs.SD] Signal and Image Processing

Fichier principal

main.pdf (1.39 Mo)

figures/experiments/DeclippingRedundancy2CHAMBER.pdf (5.77 Ko)

figures/experiments/DeclippingRedundancy2JAZZ.pdf (5.77 Ko)

figures/experiments/DeclippingRedundancy2ORCHESTRA.pdf (5.78 Ko)

figures/experiments/DeclippingRedundancy2PEAQ.pdf (7.47 Ko)

figures/experiments/DeclippingRedundancy2POP.pdf (5.98 Ko)

figures/experiments/DeclippingRedundancy2SPEECH.pdf (5.74 Ko)

figures/experiments/DeclippingRedundancy2SPEECHPESQ.pdf (7.06 Ko)

figures/experiments/DeclippingRedundancy2SPEECHSTOI.pdf (7.29 Ko)

figures/experiments/DeclippingRedundancy2VOCALS.pdf (5.77 Ko)

figures/experiments/LegendSMALL.pdf (1.85 Ko)

figures/experiments/Mushra.pdf (5.91 Ko)

figures/experiments/PlainCosparseITER.pdf (36.32 Ko)

figures/experiments/PlainSparseITER.pdf (34.85 Ko)

figures/experiments/SMALLMusicSDR.pdf (5.45 Ko)

figures/experiments/SMALLPeaq.pdf (7.69 Ko)

figures/experiments/SMALLPesq.pdf (7.77 Ko)

figures/experiments/SMALLSpeechSDR.pdf (5.61 Ko)

figures/experiments/SMALLStoi.pdf (7.04 Ko)

figures/experiments/SocialCosparseITER.pdf (33.37 Ko)

figures/experiments/SocialSparseITER.pdf (33.57 Ko)

figures/others/SpectrogramTonal.pdf (451.63 Ko)

figures/others/SpectrogramTransient.pdf (357.2 Ko)

figures/quantifying/musicClipVsPEAQ.pdf (4.92 Ko)

figures/quantifying/musicClipVsSDR.pdf (5.31 Ko)

figures/quantifying/musicSDRVsPEAQ.pdf (5.17 Ko)

figures/quantifying/speechClipVsPESQ.pdf (5.14 Ko)

figures/quantifying/speechClipVsSDR.pdf (5.33 Ko)

figures/quantifying/speechSDRVsPESQ.pdf (5.32 Ko)

figures/response/Ranking.pdf (5.56 Ko)

Origin	Files produced by the author(s)

Clément Gaultier : Connect in order to contact the contributor

https://inria.hal.science/hal-02611226

Submitted on : Thursday, January 28, 2021-2:21:59 PM

Last modification on : Wednesday, September 4, 2024-5:34:07 PM

Dates and versions

hal-02611226 , version 1 (18-05-2020)

hal-02611226 , version 2 (30-11-2020)

hal-02611226 , version 3 (28-01-2021)

Identifiers

HAL Id : hal-02611226 , version 3
ARXIV : 2005.10228
DOI : 10.1109/TASLP.2021.3059264

Cite

Clément Gaultier, Srđan Kitić, Rémi Gribonval, Nancy Bertin. Sparsity-based audio declipping methods: selected overview, new algorithms, and large-scale evaluation. IEEE/ACM Transactions on Audio, Speech and Language Processing, 2021, 29, pp.1174-1187. ⟨10.1109/TASLP.2021.3059264⟩. ⟨hal-02611226v3⟩

Export

BibTeX XML-TEI Dublin Core DC Terms EndNote DataCite

Collections

ENS-LYON UNIV-LYON3 UNIV-RENNES1 UGA CNRS INRIA UNIV-LYON1 UNIV-LYON2 INSA-LYON INSA-RENNES IRISA CENTRALESUPELEC INRIA2 UR1-MATH-STIC UR1-UFR-ISTIC UNIV-RENNES INSA-GROUPE UDL UR1-MATH-NUM

425 View

845 Download