Weakly informed audio source separation

Kilian Schulze-Forster; Clement Doire; Gael Richard; Roland Badeau

doi:10.1109/WASPAA.2019.8937266

Weakly informed audio source separation

Kilian Schulze-Forster, Clement Doire, Gael Richard, Roland Badeau

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Prior information about the target source can improve audio source separation quality but is usually not available with the necessary level of audio alignment. This has limited its usability in the past. We propose a separation model that can nevertheless exploit such weak information for the separation task while aligning it on the mixture as a byproduct using an attention mechanism. We demonstrate the capabilities of the model on a singing voice separation task exploiting artificial side information with different levels of expressiveness. Moreover, we highlight an issue with the common separation quality assessment procedure regarding parts where targets or predictions are silent and refine a previous contribution for a more complete evaluation.

Original language	English
Title of host publication	2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	273-277
Number of pages	5
ISBN (Electronic)	9781728111230
DOIs	https://doi.org/10.1109/WASPAA.2019.8937266
Publication status	Published - 1 Oct 2019
Externally published	Yes
Event	2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019 - New Paltz, United States Duration: 20 Oct 2019 → 23 Oct 2019

Publication series

Name	IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
Volume	2019-October
ISSN (Print)	1931-1168
ISSN (Electronic)	1947-1629

Conference

Conference	2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019
Country/Territory	United States
City	New Paltz
Period	20/10/19 → 23/10/19

Keywords

attention
informed source separation
separation evaluation
singing voice separation
weak labels

Access to Document

10.1109/WASPAA.2019.8937266

Cite this

Schulze-Forster, K., Doire, C., Richard, G., & Badeau, R. (2019). Weakly informed audio source separation. In 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019 (pp. 273-277). Article 8937266 (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics; Vol. 2019-October). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/WASPAA.2019.8937266

@inproceedings{7463d6029bee4034b7382abb8f4399d1,

title = "Weakly informed audio source separation",

abstract = "Prior information about the target source can improve audio source separation quality but is usually not available with the necessary level of audio alignment. This has limited its usability in the past. We propose a separation model that can nevertheless exploit such weak information for the separation task while aligning it on the mixture as a byproduct using an attention mechanism. We demonstrate the capabilities of the model on a singing voice separation task exploiting artificial side information with different levels of expressiveness. Moreover, we highlight an issue with the common separation quality assessment procedure regarding parts where targets or predictions are silent and refine a previous contribution for a more complete evaluation.",

keywords = "attention, informed source separation, separation evaluation, singing voice separation, weak labels",

author = "Kilian Schulze-Forster and Clement Doire and Gael Richard and Roland Badeau",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019 ; Conference date: 20-10-2019 Through 23-10-2019",

year = "2019",

month = oct,

day = "1",

doi = "10.1109/WASPAA.2019.8937266",

language = "English",

series = "IEEE Workshop on Applications of Signal Processing to Audio and Acoustics",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "273--277",

booktitle = "2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019",

}

Schulze-Forster, K, Doire, C, Richard, G & Badeau, R 2019, Weakly informed audio source separation. in 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019., 8937266, IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, vol. 2019-October, Institute of Electrical and Electronics Engineers Inc., pp. 273-277, 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019, New Paltz, United States, 20/10/19. https://doi.org/10.1109/WASPAA.2019.8937266

Weakly informed audio source separation. / Schulze-Forster, Kilian; Doire, Clement; Richard, Gael et al.
2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019. Institute of Electrical and Electronics Engineers Inc., 2019. p. 273-277 8937266 (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics; Vol. 2019-October).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Weakly informed audio source separation

AU - Schulze-Forster, Kilian

AU - Doire, Clement

AU - Richard, Gael

AU - Badeau, Roland

PY - 2019/10/1

Y1 - 2019/10/1

N2 - Prior information about the target source can improve audio source separation quality but is usually not available with the necessary level of audio alignment. This has limited its usability in the past. We propose a separation model that can nevertheless exploit such weak information for the separation task while aligning it on the mixture as a byproduct using an attention mechanism. We demonstrate the capabilities of the model on a singing voice separation task exploiting artificial side information with different levels of expressiveness. Moreover, we highlight an issue with the common separation quality assessment procedure regarding parts where targets or predictions are silent and refine a previous contribution for a more complete evaluation.

AB - Prior information about the target source can improve audio source separation quality but is usually not available with the necessary level of audio alignment. This has limited its usability in the past. We propose a separation model that can nevertheless exploit such weak information for the separation task while aligning it on the mixture as a byproduct using an attention mechanism. We demonstrate the capabilities of the model on a singing voice separation task exploiting artificial side information with different levels of expressiveness. Moreover, we highlight an issue with the common separation quality assessment procedure regarding parts where targets or predictions are silent and refine a previous contribution for a more complete evaluation.

KW - attention

KW - informed source separation

KW - separation evaluation

KW - singing voice separation

KW - weak labels

U2 - 10.1109/WASPAA.2019.8937266

DO - 10.1109/WASPAA.2019.8937266

M3 - Conference contribution

AN - SCOPUS:85077999371

T3 - IEEE Workshop on Applications of Signal Processing to Audio and Acoustics

SP - 273

EP - 277

BT - 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019

Y2 - 20 October 2019 through 23 October 2019

ER -

Schulze-Forster K, Doire C, Richard G , Badeau R. Weakly informed audio source separation. In 2019 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2019. Institute of Electrical and Electronics Engineers Inc. 2019. p. 273-277. 8937266. (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics). doi: 10.1109/WASPAA.2019.8937266

Weakly informed audio source separation

Abstract

Publication series

Conference

Keywords

Access to Document

Fingerprint

Cite this