A Statistically Principled and Computationally Efficient Approach to Speech Enhancement using Variational Autoencoders : Supporting Document - Archive ouverte HAL Access content directly
Reports (Research Report) Year : 2019

A Statistically Principled and Computationally Efficient Approach to Speech Enhancement using Variational Autoencoders : Supporting Document

(1, 2) , (1, 2) , (1, 2)
1
2

Abstract

Recent studies have explored the use of deep generative models of speech spectra based of variational autoencoders (VAEs), combined with unsupervised noise models, to perform speech enhancement. These studies developed iterative algorithms involving either Gibbs sampling or gradient descent at each step, making them computationally expensive. This paper proposes a variational inference method to iteratively estimate the power spectrogram of the clean speech. Our main contribution is the analytical derivation of the variational steps in which the encoder of the pre-learned VAE can be used to estimate the variational approximation of the true posterior distribution, using the very same assumption made to train VAEs. Experiments show that the proposed method produces results on par with the aforementioned iterative methods using sampling, while decreasing the computational cost by a factor 36 to reach a given performance.
Fichier principal
Vignette du fichier
support_document_final.pdf (1.03 Mo) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02089062 , version 1 (08-04-2019)

Identifiers

  • HAL Id : hal-02089062 , version 1

Cite

Manuel Pariente, Antoine Deleforge, Emmanuel Vincent. A Statistically Principled and Computationally Efficient Approach to Speech Enhancement using Variational Autoencoders : Supporting Document. [Research Report] RR-9268, INRIA. 2019, pp.1-8. ⟨hal-02089062⟩
249 View
163 Download

Share

Gmail Facebook Twitter LinkedIn More