SVM-based combination of confidence measures in speech recognition

Shi Lei Huang; Jing Ming Kuang; Xiang Xie

SVM-based combination of confidence measures in speech recognition

Shi Lei Huang^*
, Jing Ming Kuang
, Xiang Xie

^*Corresponding author for this work

School of Information and Electronics

Beijing Institute of Technology

Research output: Contribution to journal › Article › peer-review

1 Citation (Scopus)

Abstract

Discusses the combination of phone-level confidence measures using SVM (support vector machines) in speech recognition. Phone-level confidence scores are computed using transitional method while word-level confidence scores are computed from phone-level confidence scores using support vector machines. Experiments on speaker-independent mandarin isolated word recognition showed that γ_EER (equal error rates, EER) with the proposed method are lower than those with traditional methods (the γ_EER dropped from 28.14% in baseline to 23.71% in the proposed method), and the complexity of the system increased only slightly.

Original language	English
Pages (from-to)	255-259
Number of pages	5
Journal	Beijing Ligong Daxue Xuebao/Transaction of Beijing Institute of Technology
Volume	27
Issue number	3
Publication status	Published - Mar 2007

Keywords

Confidence measure
Speech recognition
Support vector machines

Cite this

@article{7a89dd6b16554be8a69d970da57ceab6,

title = "SVM-based combination of confidence measures in speech recognition",

abstract = "Discusses the combination of phone-level confidence measures using SVM (support vector machines) in speech recognition. Phone-level confidence scores are computed using transitional method while word-level confidence scores are computed from phone-level confidence scores using support vector machines. Experiments on speaker-independent mandarin isolated word recognition showed that γEER (equal error rates, EER) with the proposed method are lower than those with traditional methods (the γEER dropped from 28.14\% in baseline to 23.71\% in the proposed method), and the complexity of the system increased only slightly.",

keywords = "Confidence measure, Speech recognition, Support vector machines",

author = "Huang, \{Shi Lei\} and Kuang, \{Jing Ming\} and Xiang Xie",

year = "2007",

month = mar,

language = "English",

volume = "27",

pages = "255--259",

journal = "Beijing Ligong Daxue Xuebao/Transaction of Beijing Institute of Technology",

issn = "1001-0645",

publisher = "Beijing Institute of Technology",

number = "3",

}

TY - JOUR

T1 - SVM-based combination of confidence measures in speech recognition

AU - Huang, Shi Lei

AU - Kuang, Jing Ming

AU - Xie, Xiang

PY - 2007/3

Y1 - 2007/3

N2 - Discusses the combination of phone-level confidence measures using SVM (support vector machines) in speech recognition. Phone-level confidence scores are computed using transitional method while word-level confidence scores are computed from phone-level confidence scores using support vector machines. Experiments on speaker-independent mandarin isolated word recognition showed that γEER (equal error rates, EER) with the proposed method are lower than those with traditional methods (the γEER dropped from 28.14% in baseline to 23.71% in the proposed method), and the complexity of the system increased only slightly.

AB - Discusses the combination of phone-level confidence measures using SVM (support vector machines) in speech recognition. Phone-level confidence scores are computed using transitional method while word-level confidence scores are computed from phone-level confidence scores using support vector machines. Experiments on speaker-independent mandarin isolated word recognition showed that γEER (equal error rates, EER) with the proposed method are lower than those with traditional methods (the γEER dropped from 28.14% in baseline to 23.71% in the proposed method), and the complexity of the system increased only slightly.

KW - Confidence measure

KW - Speech recognition

KW - Support vector machines

UR - https://www.scopus.com/pages/publications/34248178897

M3 - Article

AN - SCOPUS:34248178897

SN - 1001-0645

VL - 27

SP - 255

EP - 259

JO - Beijing Ligong Daxue Xuebao/Transaction of Beijing Institute of Technology

JF - Beijing Ligong Daxue Xuebao/Transaction of Beijing Institute of Technology

IS - 3

ER -

SVM-based combination of confidence measures in speech recognition

Abstract

Keywords

Other files and links

Fingerprint

Cite this