Improving speaker detection in multi-speaker utterances through automatic purification of training data

David C. Smith; Dan Richman

Improving speaker detection in multi-speaker utterances through automatic purification of training data

David C. Smith, Dan Richman

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

Abstract

This article is concerned with automatic purification of data used in training statistical models for automatic speaker detection. It is assumed that the available data for training a model for detecting a particular speaker of interest (SOI) is contaminated by utterances from at least one other speaker. Our approach consists of three steps: (1) build a Gaussian mixture model (GMM) for the SOI, training on the contaminated training files; (2) score consecutive segments of these training files with this GMM; (3) build a new purified GMM from highest scoring segments. We apply our method to a set of SOIs from the Switch-Board I corpus (using summed conversation sides), and show that the purified GMMs are significantly more accurate than the contaminated GMMs for detecting the presence of the SOIs in test data known to contain multi-speaker utterances. This evaluation is text-independent, and no assumptions about the identity or relationship of the non-SOIs in the training and testing data are made.

Original language	English
Title of host publication	Proceedings of the IASTED International Conference on Circuits, Signals, and Systems
Editors	M.H. Rashid, M.H. Rashid
Pages	269-274
Number of pages	6
State	Published - 2003
Externally published	Yes
Event	Proceedings of the IASTED International Conference on Circuits, Signals and Systems - Cancun, Mexico Duration: May 19 2003 → May 21 2003

Publication series

Name	Proceedings of the IASTED International Conference on Circuits, Signals, and Systems

Conference

Conference	Proceedings of the IASTED International Conference on Circuits, Signals and Systems
Country/Territory	Mexico
City	Cancun
Period	05/19/03 → 05/21/03

Keywords

Gaussian mixture models
ROC curves
Score normalization
Speaker detection
Speech processing
Training data purification

Cite this

@inproceedings{083a677bfe854418b766b26766689f32,

title = "Improving speaker detection in multi-speaker utterances through automatic purification of training data",

abstract = "This article is concerned with automatic purification of data used in training statistical models for automatic speaker detection. It is assumed that the available data for training a model for detecting a particular speaker of interest (SOI) is contaminated by utterances from at least one other speaker. Our approach consists of three steps: (1) build a Gaussian mixture model (GMM) for the SOI, training on the contaminated training files; (2) score consecutive segments of these training files with this GMM; (3) build a new purified GMM from highest scoring segments. We apply our method to a set of SOIs from the Switch-Board I corpus (using summed conversation sides), and show that the purified GMMs are significantly more accurate than the contaminated GMMs for detecting the presence of the SOIs in test data known to contain multi-speaker utterances. This evaluation is text-independent, and no assumptions about the identity or relationship of the non-SOIs in the training and testing data are made.",

keywords = "Gaussian mixture models, ROC curves, Score normalization, Speaker detection, Speech processing, Training data purification",

author = "Smith, {David C.} and Dan Richman",

year = "2003",

language = "English",

isbn = "0889863512",

series = "Proceedings of the IASTED International Conference on Circuits, Signals, and Systems",

pages = "269--274",

editor = "M.H. Rashid and M.H. Rashid",

booktitle = "Proceedings of the IASTED International Conference on Circuits, Signals, and Systems",

note = "null ; Conference date: 19-05-2003 Through 21-05-2003",

}

Smith, DC & Richman, D 2003, Improving speaker detection in multi-speaker utterances through automatic purification of training data. in MH Rashid & MH Rashid (eds), Proceedings of the IASTED International Conference on Circuits, Signals, and Systems. Proceedings of the IASTED International Conference on Circuits, Signals, and Systems, pp. 269-274, Proceedings of the IASTED International Conference on Circuits, Signals and Systems, Cancun, Mexico, 05/19/03.

Improving speaker detection in multi-speaker utterances through automatic purification of training data. / Smith, David C.; Richman, Dan.

Proceedings of the IASTED International Conference on Circuits, Signals, and Systems. ed. / M.H. Rashid; M.H. Rashid. 2003. p. 269-274 (Proceedings of the IASTED International Conference on Circuits, Signals, and Systems).

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

TY - GEN

T1 - Improving speaker detection in multi-speaker utterances through automatic purification of training data

AU - Smith, David C.

AU - Richman, Dan

PY - 2003

Y1 - 2003

N2 - This article is concerned with automatic purification of data used in training statistical models for automatic speaker detection. It is assumed that the available data for training a model for detecting a particular speaker of interest (SOI) is contaminated by utterances from at least one other speaker. Our approach consists of three steps: (1) build a Gaussian mixture model (GMM) for the SOI, training on the contaminated training files; (2) score consecutive segments of these training files with this GMM; (3) build a new purified GMM from highest scoring segments. We apply our method to a set of SOIs from the Switch-Board I corpus (using summed conversation sides), and show that the purified GMMs are significantly more accurate than the contaminated GMMs for detecting the presence of the SOIs in test data known to contain multi-speaker utterances. This evaluation is text-independent, and no assumptions about the identity or relationship of the non-SOIs in the training and testing data are made.

AB - This article is concerned with automatic purification of data used in training statistical models for automatic speaker detection. It is assumed that the available data for training a model for detecting a particular speaker of interest (SOI) is contaminated by utterances from at least one other speaker. Our approach consists of three steps: (1) build a Gaussian mixture model (GMM) for the SOI, training on the contaminated training files; (2) score consecutive segments of these training files with this GMM; (3) build a new purified GMM from highest scoring segments. We apply our method to a set of SOIs from the Switch-Board I corpus (using summed conversation sides), and show that the purified GMMs are significantly more accurate than the contaminated GMMs for detecting the presence of the SOIs in test data known to contain multi-speaker utterances. This evaluation is text-independent, and no assumptions about the identity or relationship of the non-SOIs in the training and testing data are made.

KW - Gaussian mixture models

KW - ROC curves

KW - Score normalization

KW - Speaker detection

KW - Speech processing

KW - Training data purification

UR - http://www.scopus.com/inward/record.url?scp=1542286735&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:1542286735

SN - 0889863512

T3 - Proceedings of the IASTED International Conference on Circuits, Signals, and Systems

SP - 269

EP - 274

BT - Proceedings of the IASTED International Conference on Circuits, Signals, and Systems

A2 - Rashid, M.H.

Y2 - 19 May 2003 through 21 May 2003

ER -

Improving speaker detection in multi-speaker utterances through automatic purification of training data

Abstract

Publication series

Conference

Keywords

Other files and links

Fingerprint

Cite this