Adaptive Pre-whitening Based on Parametric NMF

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceabstrakt i proceedingForskningpeer review

18 Downloads (Pure)

Resumé

Several speech processing methods assume that a clean signal is observed in white Gaussian noise (WGN). An argument against those methods is that the WGN assumption is not valid in many real acoustic scenarios. To take into account the coloured nature of the noise, a pre-whitening filter which renders the background noise closer to white can be applied. This paper introduces an adaptive pre-whitener based on a supervised non-negative matrix factorization (NMF), in which a pre-trained dictionary includes parametrized spectral information about the noise and speech sources in the form of autoregressive (AR) coefficients. Results show that the noise can get closer to white, in comparison to pre-whiteners based on conventional noise power spectral density (PSD) estimates such as minimum statistics and MMSE. A better pitch estimation accuracy can be achieved as well. Speech enhancement based on the WGN assumption shows a similar performance to the conventional enhancement which makes use of the background noise PSD estimate, which reveals that the proposed pre-whitener can preserve the signal of interest.
OriginalsprogEngelsk
Titel2019 27th European Signal Processing Conference (EUSIPCO)
StatusAccepteret/In press - 2019

Fingerprint

Power spectral density
Factorization
Speech enhancement
Speech processing
Glossaries
Acoustics
Statistics

Citer dette

Esquivel Jaramillo, A., Nielsen, J. K., & Christensen, M. G. (Accepteret/In press). Adaptive Pre-whitening Based on Parametric NMF. I 2019 27th European Signal Processing Conference (EUSIPCO)
@inbook{5c0da2e6a25a49ab8073b48d81f30330,
title = "Adaptive Pre-whitening Based on Parametric NMF",
abstract = "Several speech processing methods assume that a clean signal is observed in white Gaussian noise (WGN). An argument against those methods is that the WGN assumption is not valid in many real acoustic scenarios. To take into account the coloured nature of the noise, a pre-whitening filter which renders the background noise closer to white can be applied. This paper introduces an adaptive pre-whitener based on a supervised non-negative matrix factorization (NMF), in which a pre-trained dictionary includes parametrized spectral information about the noise and speech sources in the form of autoregressive (AR) coefficients. Results show that the noise can get closer to white, in comparison to pre-whiteners based on conventional noise power spectral density (PSD) estimates such as minimum statistics and MMSE. A better pitch estimation accuracy can be achieved as well. Speech enhancement based on the WGN assumption shows a similar performance to the conventional enhancement which makes use of the background noise PSD estimate, which reveals that the proposed pre-whitener can preserve the signal of interest.",
keywords = "pre-whitening, NMF, spectral flatness, pitch estimation, speech enhancement",
author = "{Esquivel Jaramillo}, Alfredo and Nielsen, {Jesper Kj{\ae}r} and Christensen, {Mads Gr{\ae}sb{\o}ll}",
year = "2019",
language = "English",
booktitle = "2019 27th European Signal Processing Conference (EUSIPCO)",

}

Esquivel Jaramillo, A, Nielsen, JK & Christensen, MG 2019, Adaptive Pre-whitening Based on Parametric NMF. i 2019 27th European Signal Processing Conference (EUSIPCO).

Adaptive Pre-whitening Based on Parametric NMF. / Esquivel Jaramillo, Alfredo; Nielsen, Jesper Kjær; Christensen, Mads Græsbøll.

2019 27th European Signal Processing Conference (EUSIPCO). 2019.

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceabstrakt i proceedingForskningpeer review

TY - ABST

T1 - Adaptive Pre-whitening Based on Parametric NMF

AU - Esquivel Jaramillo, Alfredo

AU - Nielsen, Jesper Kjær

AU - Christensen, Mads Græsbøll

PY - 2019

Y1 - 2019

N2 - Several speech processing methods assume that a clean signal is observed in white Gaussian noise (WGN). An argument against those methods is that the WGN assumption is not valid in many real acoustic scenarios. To take into account the coloured nature of the noise, a pre-whitening filter which renders the background noise closer to white can be applied. This paper introduces an adaptive pre-whitener based on a supervised non-negative matrix factorization (NMF), in which a pre-trained dictionary includes parametrized spectral information about the noise and speech sources in the form of autoregressive (AR) coefficients. Results show that the noise can get closer to white, in comparison to pre-whiteners based on conventional noise power spectral density (PSD) estimates such as minimum statistics and MMSE. A better pitch estimation accuracy can be achieved as well. Speech enhancement based on the WGN assumption shows a similar performance to the conventional enhancement which makes use of the background noise PSD estimate, which reveals that the proposed pre-whitener can preserve the signal of interest.

AB - Several speech processing methods assume that a clean signal is observed in white Gaussian noise (WGN). An argument against those methods is that the WGN assumption is not valid in many real acoustic scenarios. To take into account the coloured nature of the noise, a pre-whitening filter which renders the background noise closer to white can be applied. This paper introduces an adaptive pre-whitener based on a supervised non-negative matrix factorization (NMF), in which a pre-trained dictionary includes parametrized spectral information about the noise and speech sources in the form of autoregressive (AR) coefficients. Results show that the noise can get closer to white, in comparison to pre-whiteners based on conventional noise power spectral density (PSD) estimates such as minimum statistics and MMSE. A better pitch estimation accuracy can be achieved as well. Speech enhancement based on the WGN assumption shows a similar performance to the conventional enhancement which makes use of the background noise PSD estimate, which reveals that the proposed pre-whitener can preserve the signal of interest.

KW - pre-whitening

KW - NMF

KW - spectral flatness

KW - pitch estimation

KW - speech enhancement

M3 - Conference abstract in proceeding

BT - 2019 27th European Signal Processing Conference (EUSIPCO)

ER -

Esquivel Jaramillo A, Nielsen JK, Christensen MG. Adaptive Pre-whitening Based on Parametric NMF. I 2019 27th European Signal Processing Conference (EUSIPCO). 2019