Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs

Jesper Højvang Jensen; Dan P. W. Ellis; Mads Græsbøll Christensen; Søren Holdt Jensen

Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs

Jesper Højvang Jensen, Dan P. W. Ellis, Mads Græsbøll Christensen, Søren Holdt Jensen

Department of Electronic Systems

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

49 Citations (Scopus)

1463 Downloads (Pure)

Abstract

In music similarity and in the related task of genre classification, a distance measure between Gaussian mixture models is frequently needed. We present a comparison of the Kullback-Leibler distance, the earth movers distance and the normalized L2 distance for this application. Although the normalized L2 distance was slightly inferior to the Kullback-Leibler distance with respect to classification performance, it has the advantage of obeying the triangle inequality, which allows for efficient searching.

Original language	English
Title of host publication	Proceedings of the 8th International Conference on Music Information Retrieval
Number of pages	2
Publisher	Austrian Computer Society
Publication date	2007
Pages	107-108
ISBN (Print)	978-3-85403-218-2
Publication status	Published - 2007
Event	International Conference on Music Information Retrieval - Vienna, Austria Duration: 23 Sept 2007 → 27 Sept 2007 Conference number: 8

Conference

Conference	International Conference on Music Information Retrieval
Number	8
Country/Territory	Austria
City	Vienna
Period	23/09/2007 → 27/09/2007

Access to Document

Jensen07ismirFinal published version, 282 KB

http://ismir2007.ismir.net/schedule.html

AUB Link

Search for the material in Aalborg University Library's search engine

IntelligentSound
Jensen, S. H., Christensen, M. G., Jensen, J. H. & Laurberg, H.
STVF
01/01/2005 → 31/12/2008
Project: Research

Cite this

@inproceedings{0be3b2703f4411dc912d000ea68e967b,

title = "Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs",

abstract = "In music similarity and in the related task of genre classification, a distance measure between Gaussian mixture models is frequently needed. We present a comparison of the Kullback-Leibler distance, the earth movers distance and the normalized L2 distance for this application. Although the normalized L2 distance was slightly inferior to the Kullback-Leibler distance with respect to classification performance, it has the advantage of obeying the triangle inequality, which allows for efficient searching.",

author = "Jensen, {Jesper H{\o}jvang} and Ellis, {Dan P. W.} and Christensen, {Mads Gr{\ae}sb{\o}ll} and Jensen, {S{\o}ren Holdt}",

year = "2007",

language = "English",

isbn = "978-3-85403-218-2",

pages = "107--108",

booktitle = "Proceedings of the 8th International Conference on Music Information Retrieval",

publisher = "Austrian Computer Society",

note = "International Conference on Music Information Retrieval ; Conference date: 23-09-2007 Through 27-09-2007",

}

Jensen, JH, Ellis, DPW, Christensen, MG & Jensen, SH 2007, Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs. in Proceedings of the 8th International Conference on Music Information Retrieval. Austrian Computer Society, pp. 107-108, International Conference on Music Information Retrieval, Vienna, Austria, 23/09/2007. <http://ismir2007.ismir.net/schedule.html>

Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs. / Jensen, Jesper Højvang; Ellis, Dan P. W.; Christensen, Mads Græsbøll et al.
Proceedings of the 8th International Conference on Music Information Retrieval. Austrian Computer Society, 2007. p. 107-108.

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs

AU - Jensen, Jesper Højvang

AU - Ellis, Dan P. W.

AU - Christensen, Mads Græsbøll

AU - Jensen, Søren Holdt

N1 - Conference code: 8

PY - 2007

Y1 - 2007

N2 - In music similarity and in the related task of genre classification, a distance measure between Gaussian mixture models is frequently needed. We present a comparison of the Kullback-Leibler distance, the earth movers distance and the normalized L2 distance for this application. Although the normalized L2 distance was slightly inferior to the Kullback-Leibler distance with respect to classification performance, it has the advantage of obeying the triangle inequality, which allows for efficient searching.

AB - In music similarity and in the related task of genre classification, a distance measure between Gaussian mixture models is frequently needed. We present a comparison of the Kullback-Leibler distance, the earth movers distance and the normalized L2 distance for this application. Although the normalized L2 distance was slightly inferior to the Kullback-Leibler distance with respect to classification performance, it has the advantage of obeying the triangle inequality, which allows for efficient searching.

M3 - Article in proceeding

SN - 978-3-85403-218-2

SP - 107

EP - 108

BT - Proceedings of the 8th International Conference on Music Information Retrieval

PB - Austrian Computer Society

T2 - International Conference on Music Information Retrieval

Y2 - 23 September 2007 through 27 September 2007

ER -

Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs

Abstract

Conference

Access to Document

AUB Link

Fingerprint

Projects

IntelligentSound

Cite this