Segmental K-means initialization for SOM-based speaker clustering

Oshry Ben-Harush; Itshak Lapidot; Hugo Guterman

Segmental K-means initialization for SOM-based speaker clustering

Oshry Ben-Harush, Itshak Lapidot, Hugo Guterman

Department of Electrical & Computer Engineering

Research output: Contribution to journal › Conference article › peer-review

3 Scopus citations

Abstract

A new approach for initial assignment of data in a speaker clustering application is presented. This approach employs Segmental K-Means clustering algorithm prior to competitive based learning. The clustering system relies on Self-Organizing Maps (SOM) for speaker modeling and as a likelihood estimator. Performance is evaluated on 108 two speaker conversations taken from LDC CALLHOME American English Speech corpus using NIST criterion and shows an improvement of 20%-30% in Cluster Error Rate (CER) relative to the randomly initialized clustering system. The number of iterations was reduced significantly, which contributes to both speed and efficiency of the clustering system.

Original language	English
Article number	4747495
Pages (from-to)	305-308
Number of pages	4
Journal	Proceedings Elmar - International Symposium Electronics in Marine
Volume	1
State	Published - 1 Dec 2008
Event	ELMAR-2008 - 50th International Symposium ELMAR 2008 - Zadar, Croatia Duration: 10 Sep 2008 → 12 Sep 2008

Keywords

Clustering
Initial Conditions
K-means
SOM
Speech

ASJC Scopus subject areas

Electrical and Electronic Engineering

Cite this

@article{9273b2af64d24153a60a560ca2f2f7ca,

title = "Segmental K-means initialization for SOM-based speaker clustering",

abstract = "A new approach for initial assignment of data in a speaker clustering application is presented. This approach employs Segmental K-Means clustering algorithm prior to competitive based learning. The clustering system relies on Self-Organizing Maps (SOM) for speaker modeling and as a likelihood estimator. Performance is evaluated on 108 two speaker conversations taken from LDC CALLHOME American English Speech corpus using NIST criterion and shows an improvement of 20%-30% in Cluster Error Rate (CER) relative to the randomly initialized clustering system. The number of iterations was reduced significantly, which contributes to both speed and efficiency of the clustering system.",

keywords = "Clustering, Initial Conditions, K-means, SOM, Speech",

author = "Oshry Ben-Harush and Itshak Lapidot and Hugo Guterman",

year = "2008",

month = dec,

day = "1",

language = "English",

volume = "1",

pages = "305--308",

journal = "Proceedings Elmar - International Symposium Electronics in Marine",

issn = "1334-2630",

publisher = "Institute of Electrical and Electronics Engineers",

note = "ELMAR-2008 - 50th International Symposium ELMAR 2008 ; Conference date: 10-09-2008 Through 12-09-2008",

}

TY - JOUR

T1 - Segmental K-means initialization for SOM-based speaker clustering

AU - Ben-Harush, Oshry

AU - Lapidot, Itshak

AU - Guterman, Hugo

PY - 2008/12/1

Y1 - 2008/12/1

N2 - A new approach for initial assignment of data in a speaker clustering application is presented. This approach employs Segmental K-Means clustering algorithm prior to competitive based learning. The clustering system relies on Self-Organizing Maps (SOM) for speaker modeling and as a likelihood estimator. Performance is evaluated on 108 two speaker conversations taken from LDC CALLHOME American English Speech corpus using NIST criterion and shows an improvement of 20%-30% in Cluster Error Rate (CER) relative to the randomly initialized clustering system. The number of iterations was reduced significantly, which contributes to both speed and efficiency of the clustering system.

AB - A new approach for initial assignment of data in a speaker clustering application is presented. This approach employs Segmental K-Means clustering algorithm prior to competitive based learning. The clustering system relies on Self-Organizing Maps (SOM) for speaker modeling and as a likelihood estimator. Performance is evaluated on 108 two speaker conversations taken from LDC CALLHOME American English Speech corpus using NIST criterion and shows an improvement of 20%-30% in Cluster Error Rate (CER) relative to the randomly initialized clustering system. The number of iterations was reduced significantly, which contributes to both speed and efficiency of the clustering system.

KW - Clustering

KW - Initial Conditions

KW - K-means

KW - SOM

KW - Speech

UR - http://www.scopus.com/inward/record.url?scp=62849098794&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:62849098794

SN - 1334-2630

VL - 1

SP - 305

EP - 308

JO - Proceedings Elmar - International Symposium Electronics in Marine

JF - Proceedings Elmar - International Symposium Electronics in Marine

M1 - 4747495

T2 - ELMAR-2008 - 50th International Symposium ELMAR 2008

Y2 - 10 September 2008 through 12 September 2008

ER -

Segmental K-means initialization for SOM-based speaker clustering

Abstract

Keywords

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this