A fast EM algorithm for Gaussian model-based source separation - Archive ouverte HAL Access content directly
Conference Papers Year : 2013

A fast EM algorithm for Gaussian model-based source separation

(1, 2) , (3)
1
2
3

Abstract

We consider the FASST framework for audio source separation, which models the sources by full-rank spatial covariance matrices and multilevel nonnegative matrix factorization (NMF) spectra. The computational cost of the expectation-maximization (EM) algorithm in [1] greatly increases with the number of channels. We present alternative EM updates using discrete hidden variables which exhibit a smaller cost. We evaluate the results on mixtures of speech and real-world environmental noise taken from our DEMAND database. The proposed algorithm is several orders of magnitude faster and it provides better separation quality for two-channel mixtures in low input signal-to-noise ratio (iSNR) conditions.
Fichier principal
Vignette du fichier
thiemann_EUSIPCO13.pdf (268.57 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00840366 , version 1 (02-07-2013)

Identifiers

  • HAL Id : hal-00840366 , version 1

Cite

Joachim Thiemann, Emmanuel Vincent. A fast EM algorithm for Gaussian model-based source separation. EUSIPCO - 21st European Signal Processing Conference - 2013, Sep 2013, Marrakech, Morocco. ⟨hal-00840366⟩
405 View
400 Download

Share

Gmail Facebook Twitter LinkedIn More