Predicting algorithm efficacy for adaptive multi-cue source separation

Ethan Manilow; Prem Seetharaman; Fatemeh Pishdadian; Bryan A Pardo

doi:10.1109/WASPAA.2017.8170038

Predicting algorithm efficacy for adaptive multi-cue source separation

Ethan Manilow, Prem Seetharaman, Fatemeh Pishdadian, Bryan A Pardo

Computer Science

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

11 Scopus citations

Abstract

Audio source separation is the process of decomposing a signal containing sounds from multiple sources into a set of signals, each from a single source. Source separation algorithms typically leverage assumptions about correlations between audio signal characteristics ('cues') and the audio sources or mixing parameters, and exploit these to do separation. We train a neural network to predict quality of source separation, as measured by Signal to Distortion Ratio, or SDR. We do this for three source separation algorithms, each leveraging a different cue-repetition, spatialization, and harmonicity/pitch proximity. Our model estimates separation quality using only the original audio mixture and separated source output by an algorithm. These estimates are reliable enough to be used to guide switching between algorithms as cues vary. Our approach for separation quality prediction can be generalized to arbitrary source separation algorithms.

Original language	English (US)
Title of host publication	2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	274-278
Number of pages	5
ISBN (Electronic)	9781538616321
DOIs	https://doi.org/10.1109/WASPAA.2017.8170038
State	Published - Dec 7 2017
Event	2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017 - New Paltz, United States Duration: Oct 15 2017 → Oct 18 2017

Publication series

Name	IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
Volume	2017-October

Other

Other	2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017
Country/Territory	United States
City	New Paltz
Period	10/15/17 → 10/18/17

Funding

∗contributed equally †This work was supported by NSF Grant 1420971.

Keywords

background
foreground
melody
prediction
repetition
singing voice separation
spatial

ASJC Scopus subject areas

Electrical and Electronic Engineering
Computer Science Applications

Access to Document

10.1109/WASPAA.2017.8170038

Cite this

Manilow, E., Seetharaman, P., Pishdadian, F., & Pardo, B. A. (2017). Predicting algorithm efficacy for adaptive multi-cue source separation. In 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017 (pp. 274-278). (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics; Vol. 2017-October). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/WASPAA.2017.8170038

Manilow, Ethan ; Seetharaman, Prem ; Pishdadian, Fatemeh et al. / Predicting algorithm efficacy for adaptive multi-cue source separation. 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 274-278 (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics).

@inproceedings{df20648f04784d8bab75f16257da4b4b,

title = "Predicting algorithm efficacy for adaptive multi-cue source separation",

abstract = "Audio source separation is the process of decomposing a signal containing sounds from multiple sources into a set of signals, each from a single source. Source separation algorithms typically leverage assumptions about correlations between audio signal characteristics ('cues') and the audio sources or mixing parameters, and exploit these to do separation. We train a neural network to predict quality of source separation, as measured by Signal to Distortion Ratio, or SDR. We do this for three source separation algorithms, each leveraging a different cue-repetition, spatialization, and harmonicity/pitch proximity. Our model estimates separation quality using only the original audio mixture and separated source output by an algorithm. These estimates are reliable enough to be used to guide switching between algorithms as cues vary. Our approach for separation quality prediction can be generalized to arbitrary source separation algorithms.",

keywords = "background, foreground, melody, prediction, repetition, singing voice separation, spatial",

author = "Ethan Manilow and Prem Seetharaman and Fatemeh Pishdadian and Pardo, {Bryan A}",

note = "Publisher Copyright: {\textcopyright} 2017 IEEE.; 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017 ; Conference date: 15-10-2017 Through 18-10-2017",

year = "2017",

month = dec,

day = "7",

doi = "10.1109/WASPAA.2017.8170038",

language = "English (US)",

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

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

pages = "274--278",

booktitle = "2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017",

address = "United States",

}

Manilow, E, Seetharaman, P, Pishdadian, F & Pardo, BA 2017, Predicting algorithm efficacy for adaptive multi-cue source separation. in 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017. IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, vol. 2017-October, Institute of Electrical and Electronics Engineers Inc., pp. 274-278, 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017, New Paltz, United States, 10/15/17. https://doi.org/10.1109/WASPAA.2017.8170038

Predicting algorithm efficacy for adaptive multi-cue source separation. / Manilow, Ethan; Seetharaman, Prem; Pishdadian, Fatemeh et al.
2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 274-278 (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics; Vol. 2017-October).

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

TY - GEN

T1 - Predicting algorithm efficacy for adaptive multi-cue source separation

AU - Manilow, Ethan

AU - Seetharaman, Prem

AU - Pishdadian, Fatemeh

AU - Pardo, Bryan A

PY - 2017/12/7

Y1 - 2017/12/7

N2 - Audio source separation is the process of decomposing a signal containing sounds from multiple sources into a set of signals, each from a single source. Source separation algorithms typically leverage assumptions about correlations between audio signal characteristics ('cues') and the audio sources or mixing parameters, and exploit these to do separation. We train a neural network to predict quality of source separation, as measured by Signal to Distortion Ratio, or SDR. We do this for three source separation algorithms, each leveraging a different cue-repetition, spatialization, and harmonicity/pitch proximity. Our model estimates separation quality using only the original audio mixture and separated source output by an algorithm. These estimates are reliable enough to be used to guide switching between algorithms as cues vary. Our approach for separation quality prediction can be generalized to arbitrary source separation algorithms.

AB - Audio source separation is the process of decomposing a signal containing sounds from multiple sources into a set of signals, each from a single source. Source separation algorithms typically leverage assumptions about correlations between audio signal characteristics ('cues') and the audio sources or mixing parameters, and exploit these to do separation. We train a neural network to predict quality of source separation, as measured by Signal to Distortion Ratio, or SDR. We do this for three source separation algorithms, each leveraging a different cue-repetition, spatialization, and harmonicity/pitch proximity. Our model estimates separation quality using only the original audio mixture and separated source output by an algorithm. These estimates are reliable enough to be used to guide switching between algorithms as cues vary. Our approach for separation quality prediction can be generalized to arbitrary source separation algorithms.

KW - background

KW - foreground

KW - melody

KW - prediction

KW - repetition

KW - singing voice separation

KW - spatial

UR - http://www.scopus.com/inward/record.url?scp=85042349506&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85042349506&partnerID=8YFLogxK

U2 - 10.1109/WASPAA.2017.8170038

DO - 10.1109/WASPAA.2017.8170038

M3 - Conference contribution

AN - SCOPUS:85042349506

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

SP - 274

EP - 278

BT - 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017

Y2 - 15 October 2017 through 18 October 2017

ER -

Manilow E, Seetharaman P, Pishdadian F, Pardo BA. Predicting algorithm efficacy for adaptive multi-cue source separation. In 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 274-278. (IEEE Workshop on Applications of Signal Processing to Audio and Acoustics). doi: 10.1109/WASPAA.2017.8170038

Predicting algorithm efficacy for adaptive multi-cue source separation

Abstract

Publication series

Other

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this