Melody extraction by contour classification

Rachel M. Bittner, Justin Salamon, Slim Essid, Juan Bello

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

Abstract

Due to the scarcity of labeled data, most melody extraction algorithms do not rely on fully data-driven processing blocks but rather on careful engineering. For example, the Melodia melody extraction algorithm employs a pitch contour selection stage that relies on a number of heuristics for selecting the melodic output. In this paper we explore the use of a discriminative model to perform purely data-driven melodic contour selection. Specifically, a discriminative binary classifier is trained to distinguish melodic from non-melodic contours. This classifier is then used to predict likelihoods for a track’s extracted contours, and these scores are decoded to generate a single melody output. The results are compared with the Melodia algorithm and with a generative model used in a previous study. We show that the discriminative model outperforms the generative model in terms of contour classification accuracy, and the melody output from our proposed system performs comparatively to Melodia. The results are complemented with error analysis and avenues for future improvements.

Original languageEnglish (US)
Title of host publicationProceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015
EditorsFrans Wiering, Meinard Muller
PublisherInternational Society for Music Information Retrieval
Pages500-506
Number of pages7
ISBN (Electronic)9788460688532
StatePublished - Jan 1 2015
Event16th International Society for Music Information Retrieval Conference, ISMIR 2015 - Malaga, Spain
Duration: Oct 26 2015Oct 30 2015

Publication series

NameProceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015

Conference

Conference16th International Society for Music Information Retrieval Conference, ISMIR 2015
CountrySpain
CityMalaga
Period10/26/1510/30/15

Fingerprint

Classifiers
Error analysis
Melody
Data-driven
Generative
Classifier
Pitch Contour
Heuristics
Melodic Contour
Error Analysis
Scarcity

ASJC Scopus subject areas

  • Music
  • Information Systems

Cite this

Bittner, R. M., Salamon, J., Essid, S., & Bello, J. (2015). Melody extraction by contour classification. In F. Wiering, & M. Muller (Eds.), Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015 (pp. 500-506). (Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015). International Society for Music Information Retrieval.

Melody extraction by contour classification. / Bittner, Rachel M.; Salamon, Justin; Essid, Slim; Bello, Juan.

Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015. ed. / Frans Wiering; Meinard Muller. International Society for Music Information Retrieval, 2015. p. 500-506 (Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015).

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

Bittner, RM, Salamon, J, Essid, S & Bello, J 2015, Melody extraction by contour classification. in F Wiering & M Muller (eds), Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015. Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015, International Society for Music Information Retrieval, pp. 500-506, 16th International Society for Music Information Retrieval Conference, ISMIR 2015, Malaga, Spain, 10/26/15.
Bittner RM, Salamon J, Essid S, Bello J. Melody extraction by contour classification. In Wiering F, Muller M, editors, Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015. International Society for Music Information Retrieval. 2015. p. 500-506. (Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015).
Bittner, Rachel M. ; Salamon, Justin ; Essid, Slim ; Bello, Juan. / Melody extraction by contour classification. Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015. editor / Frans Wiering ; Meinard Muller. International Society for Music Information Retrieval, 2015. pp. 500-506 (Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015).
@inproceedings{cf3c3a8fee91423285a765d1a49bc0c2,
title = "Melody extraction by contour classification",
abstract = "Due to the scarcity of labeled data, most melody extraction algorithms do not rely on fully data-driven processing blocks but rather on careful engineering. For example, the Melodia melody extraction algorithm employs a pitch contour selection stage that relies on a number of heuristics for selecting the melodic output. In this paper we explore the use of a discriminative model to perform purely data-driven melodic contour selection. Specifically, a discriminative binary classifier is trained to distinguish melodic from non-melodic contours. This classifier is then used to predict likelihoods for a track’s extracted contours, and these scores are decoded to generate a single melody output. The results are compared with the Melodia algorithm and with a generative model used in a previous study. We show that the discriminative model outperforms the generative model in terms of contour classification accuracy, and the melody output from our proposed system performs comparatively to Melodia. The results are complemented with error analysis and avenues for future improvements.",
author = "Bittner, {Rachel M.} and Justin Salamon and Slim Essid and Juan Bello",
year = "2015",
month = "1",
day = "1",
language = "English (US)",
series = "Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015",
publisher = "International Society for Music Information Retrieval",
pages = "500--506",
editor = "Frans Wiering and Meinard Muller",
booktitle = "Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015",

}

TY - GEN

T1 - Melody extraction by contour classification

AU - Bittner, Rachel M.

AU - Salamon, Justin

AU - Essid, Slim

AU - Bello, Juan

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Due to the scarcity of labeled data, most melody extraction algorithms do not rely on fully data-driven processing blocks but rather on careful engineering. For example, the Melodia melody extraction algorithm employs a pitch contour selection stage that relies on a number of heuristics for selecting the melodic output. In this paper we explore the use of a discriminative model to perform purely data-driven melodic contour selection. Specifically, a discriminative binary classifier is trained to distinguish melodic from non-melodic contours. This classifier is then used to predict likelihoods for a track’s extracted contours, and these scores are decoded to generate a single melody output. The results are compared with the Melodia algorithm and with a generative model used in a previous study. We show that the discriminative model outperforms the generative model in terms of contour classification accuracy, and the melody output from our proposed system performs comparatively to Melodia. The results are complemented with error analysis and avenues for future improvements.

AB - Due to the scarcity of labeled data, most melody extraction algorithms do not rely on fully data-driven processing blocks but rather on careful engineering. For example, the Melodia melody extraction algorithm employs a pitch contour selection stage that relies on a number of heuristics for selecting the melodic output. In this paper we explore the use of a discriminative model to perform purely data-driven melodic contour selection. Specifically, a discriminative binary classifier is trained to distinguish melodic from non-melodic contours. This classifier is then used to predict likelihoods for a track’s extracted contours, and these scores are decoded to generate a single melody output. The results are compared with the Melodia algorithm and with a generative model used in a previous study. We show that the discriminative model outperforms the generative model in terms of contour classification accuracy, and the melody output from our proposed system performs comparatively to Melodia. The results are complemented with error analysis and avenues for future improvements.

UR - http://www.scopus.com/inward/record.url?scp=85008455727&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85008455727&partnerID=8YFLogxK

M3 - Conference contribution

T3 - Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015

SP - 500

EP - 506

BT - Proceedings of the 16th International Society for Music Information Retrieval Conference, ISMIR 2015

A2 - Wiering, Frans

A2 - Muller, Meinard

PB - International Society for Music Information Retrieval

ER -