Using hyper-genre training to explore genre information for automatic chord estimation

Yizhao Ni; Matt Mcvicar; Raul Santos-Rodríguez; Tijl De Bie

Using hyper-genre training to explore genre information for automatic chord estimation

Yizhao Ni^*, Matt Mcvicar, Raul Santos-Rodríguez, Tijl De Bie

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

8 Citations (Scopus)

Abstract

Recently a large amount of new chord annotations have been made available. This raises hopes for further development in automatic chord estimation. While more data seems to imply better performance, a major challenge however, is the wide variety of genres covered by these new data. As a result, the genre-independent training scheme as is common today is bound to fail. In this paper we investigate various options for exploring genre information for chord estimation, while also maximally exploiting the full dataset. More specifically, we propose a hyper-genre training scheme in which each genre cluster has its own parameters, tied together by hyper parameters as a Bayesian prior. The results are promising, showing significant improvements over other prevailing training schemes.

Original language	English
Title of host publication	Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012
Pages	109-114
Number of pages	6
Publication status	Published - 2012
Event	13th International Society for Music Information Retrieval Conference, ISMIR 2012 - Porto, Portugal Duration: 8 Oct 2012 → 12 Oct 2012

Publication series

Name	Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012

Conference

Conference	13th International Society for Music Information Retrieval Conference, ISMIR 2012
Country/Territory	Portugal
City	Porto
Period	8/10/12 → 12/10/12

Bibliographical note

Copyright:
Copyright 2013 Elsevier B.V., All rights reserved.

Handle.net

Persistent link

Cite this

Ni, Y., Mcvicar, M., Santos-Rodríguez, R., & De Bie, T. (2012). Using hyper-genre training to explore genre information for automatic chord estimation. In Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012 (pp. 109-114). (Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012).

Ni, Yizhao ; Mcvicar, Matt ; Santos-Rodríguez, Raul et al. / Using hyper-genre training to explore genre information for automatic chord estimation. Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012. 2012. pp. 109-114 (Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012).

@inproceedings{e4295d373b3a4e65ab4d4f997da9abcb,

title = "Using hyper-genre training to explore genre information for automatic chord estimation",

abstract = "Recently a large amount of new chord annotations have been made available. This raises hopes for further development in automatic chord estimation. While more data seems to imply better performance, a major challenge however, is the wide variety of genres covered by these new data. As a result, the genre-independent training scheme as is common today is bound to fail. In this paper we investigate various options for exploring genre information for chord estimation, while also maximally exploiting the full dataset. More specifically, we propose a hyper-genre training scheme in which each genre cluster has its own parameters, tied together by hyper parameters as a Bayesian prior. The results are promising, showing significant improvements over other prevailing training schemes.",

author = "Yizhao Ni and Matt Mcvicar and Raul Santos-Rodr{\'i}guez and \{De Bie\}, Tijl",

note = "Copyright: Copyright 2013 Elsevier B.V., All rights reserved.; 13th International Society for Music Information Retrieval Conference, ISMIR 2012 ; Conference date: 08-10-2012 Through 12-10-2012",

year = "2012",

language = "English",

isbn = "9789727521449",

series = "Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012",

pages = "109--114",

booktitle = "Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012",

}

Ni, Y, Mcvicar, M, Santos-Rodríguez, R & De Bie, T 2012, Using hyper-genre training to explore genre information for automatic chord estimation. in Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012. Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012, pp. 109-114, 13th International Society for Music Information Retrieval Conference, ISMIR 2012, Porto, Portugal, 8/10/12.

Using hyper-genre training to explore genre information for automatic chord estimation. / Ni, Yizhao; Mcvicar, Matt; Santos-Rodríguez, Raul et al.
Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012. 2012. p. 109-114 (Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012).

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Using hyper-genre training to explore genre information for automatic chord estimation

AU - Ni, Yizhao

AU - Mcvicar, Matt

AU - Santos-Rodríguez, Raul

AU - De Bie, Tijl

PY - 2012

Y1 - 2012

N2 - Recently a large amount of new chord annotations have been made available. This raises hopes for further development in automatic chord estimation. While more data seems to imply better performance, a major challenge however, is the wide variety of genres covered by these new data. As a result, the genre-independent training scheme as is common today is bound to fail. In this paper we investigate various options for exploring genre information for chord estimation, while also maximally exploiting the full dataset. More specifically, we propose a hyper-genre training scheme in which each genre cluster has its own parameters, tied together by hyper parameters as a Bayesian prior. The results are promising, showing significant improvements over other prevailing training schemes.

AB - Recently a large amount of new chord annotations have been made available. This raises hopes for further development in automatic chord estimation. While more data seems to imply better performance, a major challenge however, is the wide variety of genres covered by these new data. As a result, the genre-independent training scheme as is common today is bound to fail. In this paper we investigate various options for exploring genre information for chord estimation, while also maximally exploiting the full dataset. More specifically, we propose a hyper-genre training scheme in which each genre cluster has its own parameters, tied together by hyper parameters as a Bayesian prior. The results are promising, showing significant improvements over other prevailing training schemes.

UR - https://www.scopus.com/pages/publications/84873453177

M3 - Conference Contribution (Conference Proceeding)

AN - SCOPUS:84873453177

SN - 9789727521449

T3 - Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012

SP - 109

EP - 114

BT - Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012

T2 - 13th International Society for Music Information Retrieval Conference, ISMIR 2012

Y2 - 8 October 2012 through 12 October 2012

ER -

Using hyper-genre training to explore genre information for automatic chord estimation

Abstract

Publication series

Conference

Bibliographical note

Handle.net

Other files and links

Fingerprint

Cite this