Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks

Emad M Grais, Gerard Roma, Andrew JR Simpson, Mark D Plumbley

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

The sources separated by most single channel audio source separation techniques are usually distorted and each separated source contains residual signals from the other sources. To tackle this problem, we propose to enhance the separated sources to decrease the distortion and interference between the separated sources using deep neural networks (DNNs). Two different DNNs are used in this work. The first DNN is used to separate the sources from the mixed signal. The second DNN is used to enhance the separated signals. To consider the interactions between the separated sources, we propose to use a single DNN to enhance all the separated sources together. To reduce the residual signals of one source from the other separated sources (interference), we train the DNN for enhancement discriminatively to maximize the dissimilarity between the predicted sources. The experimental results show that using discriminative enhancement decreases the distortion and interference between the separated sources.
Original languageEnglish
Title of host publicationLatent Variable Analysis and Signal Separation
EditorsPetr Tichavský, Massoud Babaie-Zadeh, Olivier J.J. Michel, Nadège Thirion-Moreau
PublisherSpringer, Cham
Pages236-246
Number of pages11
ISBN (Electronic)9783319535470
ISBN (Print)9783319535463
DOIs
Publication statusPublished - 15 Feb 2017
Externally publishedYes
Event13th International Conference on Latent Variable Analysis and Signal Separation - Grenoble-Alpes University, Grenoble, France
Duration: 21 Feb 201723 Feb 2017
Conference number: 13
http://www.lva-ica-2017.com/ (Link to Conference Website)

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume10169
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference13th International Conference on Latent Variable Analysis and Signal Separation
Abbreviated titleLVA/ICA 2017
CountryFrance
CityGrenoble
Period21/02/1723/02/17
Internet address

Fingerprint

Source separation
Deep neural networks

Cite this

Grais, E. M., Roma, G., Simpson, A. JR., & Plumbley, M. D. (2017). Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks. In P. Tichavský, M. Babaie-Zadeh, O. J. J. Michel, & N. Thirion-Moreau (Eds.), Latent Variable Analysis and Signal Separation (pp. 236-246). (Lecture Notes in Computer Science; Vol. 10169). Springer, Cham. https://doi.org/10.1007/978-3-319-53547-0_23
Grais, Emad M ; Roma, Gerard ; Simpson, Andrew JR ; Plumbley, Mark D. / Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks. Latent Variable Analysis and Signal Separation. editor / Petr Tichavský ; Massoud Babaie-Zadeh ; Olivier J.J. Michel ; Nadège Thirion-Moreau. Springer, Cham, 2017. pp. 236-246 (Lecture Notes in Computer Science).
@inproceedings{1e62f770e5fd4148a87f11a2b2ef3d24,
title = "Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks",
abstract = "The sources separated by most single channel audio source separation techniques are usually distorted and each separated source contains residual signals from the other sources. To tackle this problem, we propose to enhance the separated sources to decrease the distortion and interference between the separated sources using deep neural networks (DNNs). Two different DNNs are used in this work. The first DNN is used to separate the sources from the mixed signal. The second DNN is used to enhance the separated signals. To consider the interactions between the separated sources, we propose to use a single DNN to enhance all the separated sources together. To reduce the residual signals of one source from the other separated sources (interference), we train the DNN for enhancement discriminatively to maximize the dissimilarity between the predicted sources. The experimental results show that using discriminative enhancement decreases the distortion and interference between the separated sources.",
keywords = "Audio enhancement, Deep neural networks, Discriminative training, Single channel audio source separation",
author = "Grais, {Emad M} and Gerard Roma and Simpson, {Andrew JR} and Plumbley, {Mark D}",
year = "2017",
month = "2",
day = "15",
doi = "10.1007/978-3-319-53547-0_23",
language = "English",
isbn = "9783319535463",
series = "Lecture Notes in Computer Science",
publisher = "Springer, Cham",
pages = "236--246",
editor = "Petr Tichavsk{\'y} and Massoud Babaie-Zadeh and Michel, {Olivier J.J.} and Nad{\`e}ge Thirion-Moreau",
booktitle = "Latent Variable Analysis and Signal Separation",

}

Grais, EM, Roma, G, Simpson, AJR & Plumbley, MD 2017, Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks. in P Tichavský, M Babaie-Zadeh, OJJ Michel & N Thirion-Moreau (eds), Latent Variable Analysis and Signal Separation. Lecture Notes in Computer Science, vol. 10169, Springer, Cham, pp. 236-246, 13th International Conference on Latent Variable Analysis and Signal Separation, Grenoble, France, 21/02/17. https://doi.org/10.1007/978-3-319-53547-0_23

Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks. / Grais, Emad M; Roma, Gerard; Simpson, Andrew JR; Plumbley, Mark D.

Latent Variable Analysis and Signal Separation. ed. / Petr Tichavský; Massoud Babaie-Zadeh; Olivier J.J. Michel; Nadège Thirion-Moreau. Springer, Cham, 2017. p. 236-246 (Lecture Notes in Computer Science; Vol. 10169).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks

AU - Grais, Emad M

AU - Roma, Gerard

AU - Simpson, Andrew JR

AU - Plumbley, Mark D

PY - 2017/2/15

Y1 - 2017/2/15

N2 - The sources separated by most single channel audio source separation techniques are usually distorted and each separated source contains residual signals from the other sources. To tackle this problem, we propose to enhance the separated sources to decrease the distortion and interference between the separated sources using deep neural networks (DNNs). Two different DNNs are used in this work. The first DNN is used to separate the sources from the mixed signal. The second DNN is used to enhance the separated signals. To consider the interactions between the separated sources, we propose to use a single DNN to enhance all the separated sources together. To reduce the residual signals of one source from the other separated sources (interference), we train the DNN for enhancement discriminatively to maximize the dissimilarity between the predicted sources. The experimental results show that using discriminative enhancement decreases the distortion and interference between the separated sources.

AB - The sources separated by most single channel audio source separation techniques are usually distorted and each separated source contains residual signals from the other sources. To tackle this problem, we propose to enhance the separated sources to decrease the distortion and interference between the separated sources using deep neural networks (DNNs). Two different DNNs are used in this work. The first DNN is used to separate the sources from the mixed signal. The second DNN is used to enhance the separated signals. To consider the interactions between the separated sources, we propose to use a single DNN to enhance all the separated sources together. To reduce the residual signals of one source from the other separated sources (interference), we train the DNN for enhancement discriminatively to maximize the dissimilarity between the predicted sources. The experimental results show that using discriminative enhancement decreases the distortion and interference between the separated sources.

KW - Audio enhancement

KW - Deep neural networks

KW - Discriminative training

KW - Single channel audio source separation

U2 - 10.1007/978-3-319-53547-0_23

DO - 10.1007/978-3-319-53547-0_23

M3 - Conference contribution

SN - 9783319535463

T3 - Lecture Notes in Computer Science

SP - 236

EP - 246

BT - Latent Variable Analysis and Signal Separation

A2 - Tichavský, Petr

A2 - Babaie-Zadeh, Massoud

A2 - Michel, Olivier J.J.

A2 - Thirion-Moreau, Nadège

PB - Springer, Cham

ER -

Grais EM, Roma G, Simpson AJR, Plumbley MD. Discriminative Enhancement for Single Channel Audio Source Separation Using Deep Neural Networks. In Tichavský P, Babaie-Zadeh M, Michel OJJ, Thirion-Moreau N, editors, Latent Variable Analysis and Signal Separation. Springer, Cham. 2017. p. 236-246. (Lecture Notes in Computer Science). https://doi.org/10.1007/978-3-319-53547-0_23