Improving discrimination of confusable words using the divergence measure

Philipos Loizou, Andreas Spanias

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

This paper describes the use of the divergence measure as a criterion for finding a transformation matrix which will map the original speech observations onto a subspace with more discriminative ability than the original. A gradient-based algorithm is also proposed to compute the transformation matrix efficiently. The subspace approach is used as a preprocessing step in a hidden Markov model (HMM) based system to enhance discrimination of acoustically similar pairs of words. This approach is compared with standard linear discriminant analysis (LDA) techniques and shown to yield as much as 4.5% improvement. The subspace approach is also applied successfully to a more general recognition problem, i.e., discrimination of K confusable words, using the average divergence measure.

Original languageEnglish (US)
Pages (from-to)1106-1111
Number of pages6
JournalJournal of the Acoustical Society of America
Volume101
Issue number2
DOIs
StatePublished - Feb 1997

Fingerprint

discrimination
divergence
preprocessing
gradients
Divergence
Discrimination
Linear Discriminant Analysis
Hidden Markov Model

ASJC Scopus subject areas

  • Acoustics and Ultrasonics

Cite this

Improving discrimination of confusable words using the divergence measure. / Loizou, Philipos; Spanias, Andreas.

In: Journal of the Acoustical Society of America, Vol. 101, No. 2, 02.1997, p. 1106-1111.

Research output: Contribution to journalArticle

@article{1ac4443ec26d430398dfd8080f51282b,
title = "Improving discrimination of confusable words using the divergence measure",
abstract = "This paper describes the use of the divergence measure as a criterion for finding a transformation matrix which will map the original speech observations onto a subspace with more discriminative ability than the original. A gradient-based algorithm is also proposed to compute the transformation matrix efficiently. The subspace approach is used as a preprocessing step in a hidden Markov model (HMM) based system to enhance discrimination of acoustically similar pairs of words. This approach is compared with standard linear discriminant analysis (LDA) techniques and shown to yield as much as 4.5{\%} improvement. The subspace approach is also applied successfully to a more general recognition problem, i.e., discrimination of K confusable words, using the average divergence measure.",
author = "Philipos Loizou and Andreas Spanias",
year = "1997",
month = "2",
doi = "10.1121/1.418032",
language = "English (US)",
volume = "101",
pages = "1106--1111",
journal = "Journal of the Acoustical Society of America",
issn = "0001-4966",
publisher = "Acoustical Society of America",
number = "2",

}

TY - JOUR

T1 - Improving discrimination of confusable words using the divergence measure

AU - Loizou, Philipos

AU - Spanias, Andreas

PY - 1997/2

Y1 - 1997/2

N2 - This paper describes the use of the divergence measure as a criterion for finding a transformation matrix which will map the original speech observations onto a subspace with more discriminative ability than the original. A gradient-based algorithm is also proposed to compute the transformation matrix efficiently. The subspace approach is used as a preprocessing step in a hidden Markov model (HMM) based system to enhance discrimination of acoustically similar pairs of words. This approach is compared with standard linear discriminant analysis (LDA) techniques and shown to yield as much as 4.5% improvement. The subspace approach is also applied successfully to a more general recognition problem, i.e., discrimination of K confusable words, using the average divergence measure.

AB - This paper describes the use of the divergence measure as a criterion for finding a transformation matrix which will map the original speech observations onto a subspace with more discriminative ability than the original. A gradient-based algorithm is also proposed to compute the transformation matrix efficiently. The subspace approach is used as a preprocessing step in a hidden Markov model (HMM) based system to enhance discrimination of acoustically similar pairs of words. This approach is compared with standard linear discriminant analysis (LDA) techniques and shown to yield as much as 4.5% improvement. The subspace approach is also applied successfully to a more general recognition problem, i.e., discrimination of K confusable words, using the average divergence measure.

UR - http://www.scopus.com/inward/record.url?scp=0031081709&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031081709&partnerID=8YFLogxK

U2 - 10.1121/1.418032

DO - 10.1121/1.418032

M3 - Article

AN - SCOPUS:0031081709

VL - 101

SP - 1106

EP - 1111

JO - Journal of the Acoustical Society of America

JF - Journal of the Acoustical Society of America

SN - 0001-4966

IS - 2

ER -