Speech enhancement with variational autoencoders and alpha-stable distributions - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2019

Speech enhancement with variational autoencoders and alpha-stable distributions

Abstract

his paper focuses on single-channel semi-supervised speech en-hancement. We learn a speaker-independent deep generative speechmodel using the framework of variational autoencoders. The noisemodel remains unsupervised because we do not assume prior knowl-edge of the noisy recording environment. In this context, our con-tribution is to propose a noise model based on alpha-stable distribu-tions, instead of the more conventional Gaussian non-negative ma-trix factorization approach found in previous studies. We develop aMonte Carlo expectation-maximization algorithm for estimating themodel parameters at test time. Experimental results show the supe-riority of the proposed approach both in terms of perceptual qualityand intelligibility of the enhanced speech signal.
Fichier principal
Vignette du fichier
LSLGH-icassp2019.pdf (535.25 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02005106 , version 1 (08-02-2019)

Licence

Attribution

Identifiers

Cite

Simon Leglaive, Umut Şimşekli, Antoine Liutkus, Laurent Girin, Radu Horaud. Speech enhancement with variational autoencoders and alpha-stable distributions. ICASSP 2019 - 44th IEEE International Conference on Acoustics, Speech and Signal Processing, May 2019, Brighton, United Kingdom. pp.541-545, ⟨10.1109/ICASSP.2019.8682546⟩. ⟨hal-02005106⟩
428 View
587 Download

Altmetric

Share

Gmail Facebook X LinkedIn More