Alpha-stable low-rank plus residual decomposition for speech enhancement

Umut Simsekli 1 Halil Erdogan 2 Simon Leglaive 1 Antoine Liutkus 3 Roland Badeau 1 Gaël Richard 1
3 ZENITH - Scientific Data Management
LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier, CRISAM - Inria Sophia Antipolis - Méditerranée
Abstract : In this study, we propose a novel probabilistic model for separating clean speech signals from noisy mixtures by decomposing the mixture spectrograms into a structured speech part and a more flexible residual part. The main novelty in our model is that it uses a family of heavy-tailed distributions, so called the α-stable distributions, for modeling the residual signal. We develop an expectation-maximization algorithm for parameter estimation and a Monte Carlo scheme for posterior estimation of the clean speech. Our experiments show that the proposed method outperforms relevant factorization-based algorithms by a significant margin.
Document type :
Conference papers
Complete list of metadatas

Cited literature [33 references]  Display  Hide  Download

https://hal.inria.fr/hal-01714909
Contributor : Antoine Liutkus <>
Submitted on : Thursday, February 22, 2018 - 7:51:25 AM
Last modification on : Thursday, October 17, 2019 - 12:36:55 PM
Long-term archiving on : Wednesday, May 23, 2018 - 12:11:22 PM

File

2017102794510_839706_2832.pdf
Files produced by the author(s)

Identifiers

Citation

Umut Simsekli, Halil Erdogan, Simon Leglaive, Antoine Liutkus, Roland Badeau, et al.. Alpha-stable low-rank plus residual decomposition for speech enhancement. ICASSP: International Conference on Acoustics, Speech, and Signal Processing, Apr 2018, Calgary, Canada. pp.651-655, ⟨10.1109/ICASSP.2018.8461539⟩. ⟨hal-01714909⟩

Share

Metrics

Record views

523

Files downloads

433