Multichannel Audio Modeling with Elliptically Stable Tensor Decomposition

Fontaine, Mathieu; Stöter, Fabian-Robert; Şimşekli, Umut; Serizel, Romain; Badeau, Roland

doi:10.1007/978-3-319-93764-9_2

Cited by 5 publications

(3 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Overall, it is shown that the proposed approach improves the intelligibility and perceptual quality of the enhanced speech signal. Future works include extending the proposed approach to a multi-microphone setting using multivariate αstable distributions [18,20].…”

Section: Discussionmentioning

confidence: 99%

“…where α ∈]0, 2] is the characteristic exponent and σ b,f ∈ R+ is the scale parameter. As proposed in [20] for multichannel speech enhancement, this scale parameter is only frequency-dependent, it does not depend on time. For algorithmic purposes, the noise model ( 6) can be conveniently rewritten in an equivalent scale mixture of Gaussians form [25], by making use of the product property of the symmetric α-stable distribution [24]:…”

Section: Noise and Mixture Modelsmentioning

confidence: 99%

“…In particular, α-stable distributions (cf. Section 3) are becoming popular heavy-tailed models for audio modeling due to their nice theoretical properties [17,13,14,18,19,20,21].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Speech Enhancement with Variational Autoencoders and Alpha-stable Distributions

Leglaive

Şimşekli

Girin

et al. 2019

ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

This paper focuses on single-channel semi-supervised speech enhancement. We learn a speaker-independent deep generative speech model using the framework of variational autoencoders. The noise model remains unsupervised because we do not assume prior knowledge of the noisy recording environment. In this context, our contribution is to propose a noise model based on alpha-stable distributions, instead of the more conventional Gaussian non-negative matrix factorization approach found in previous studies. We develop a Monte Carlo expectation-maximization algorithm for estimating the model parameters at test time. Experimental results show the superiority of the proposed approach both in terms of perceptual quality and intelligibility of the enhanced speech signal.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Noise and Mixture Modelsmentioning

confidence: 99%

See 1 more Smart Citation

Speech Enhancement with Variational Autoencoders and Alpha-stable Distributions

Leglaive

Şimşekli

Girin

et al. 2019

ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

Generalized Fast Multichannel Nonnegative Matrix Factorization Based on Gaussian Scale Mixtures for Blind Source Separation

Fontaine

Sekiguchi

Nugraha

et al. 2022

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

This paper describes heavy-tailed extensions of a state-of-the-art versatile blind source separation method called fast multichannel nonnegative matrix factorization (FastMNMF) from a unified point of view. The common way of deriving such an extension is to replace the multivariate complex Gaussian distribution in the likelihood function with its heavy-tailed generalization, e.g., the multivariate complex Student's t and leptokurtic generalized Gaussian distributions, and tailor-make the corresponding parameter optimization algorithm. Using a wider class of heavy-tailed distributions called a Gaussian scale mixture (GSM), i.e., a mixture of Gaussian distributions whose variances are perturbed by positive random scalars called impulse variables, we propose GSM-FastMNMF and develop an expectationmaximization algorithm that works even when the probability density function of the impulse variables have no analytical expressions. We show that existing heavy-tailed FastMNMF extensions are instances of GSM-FastMNMF and derive a new instance based on the generalized hyperbolic distribution that include the normal-inverse Gaussian, Student's t, and Gaussian distributions as the special cases. Our experiments show that the normalinverse Gaussian FastMNMF outperforms the state-of-the-art FastMNMF extensions and ILRMA model in speech enhancement and separation in terms of the signal-to-distortion ratio.

show abstract