Bayesian combining of confidence measures is proposed for speech recognition. Bayesian combining is achieved by the estimation of joint pdf of confidence feature vector in correct and incorrect hypothesis classes. In addition, the adaptation of a confidence score using the pdf is presented. The proposed methods reduced the classification error rate by 18% from the conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Tae-Yoon KIM, Hanseok KO, "Bayesian Confidence Scoring and Adaptation Techniques for Speech Recognition" in IEICE TRANSACTIONS on Communications,
vol. E88-B, no. 4, pp. 1756-1759, April 2005, doi: 10.1093/ietcom/e88-b.4.1756.
Abstract: Bayesian combining of confidence measures is proposed for speech recognition. Bayesian combining is achieved by the estimation of joint pdf of confidence feature vector in correct and incorrect hypothesis classes. In addition, the adaptation of a confidence score using the pdf is presented. The proposed methods reduced the classification error rate by 18% from the conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.
URL: https://global.ieice.org/en_transactions/communications/10.1093/ietcom/e88-b.4.1756/_p
Copy
@ARTICLE{e88-b_4_1756,
author={Tae-Yoon KIM, Hanseok KO, },
journal={IEICE TRANSACTIONS on Communications},
title={Bayesian Confidence Scoring and Adaptation Techniques for Speech Recognition},
year={2005},
volume={E88-B},
number={4},
pages={1756-1759},
abstract={Bayesian combining of confidence measures is proposed for speech recognition. Bayesian combining is achieved by the estimation of joint pdf of confidence feature vector in correct and incorrect hypothesis classes. In addition, the adaptation of a confidence score using the pdf is presented. The proposed methods reduced the classification error rate by 18% from the conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.},
keywords={},
doi={10.1093/ietcom/e88-b.4.1756},
ISSN={},
month={April},}
Copy
TY - JOUR
TI - Bayesian Confidence Scoring and Adaptation Techniques for Speech Recognition
T2 - IEICE TRANSACTIONS on Communications
SP - 1756
EP - 1759
AU - Tae-Yoon KIM
AU - Hanseok KO
PY - 2005
DO - 10.1093/ietcom/e88-b.4.1756
JO - IEICE TRANSACTIONS on Communications
SN -
VL - E88-B
IS - 4
JA - IEICE TRANSACTIONS on Communications
Y1 - April 2005
AB - Bayesian combining of confidence measures is proposed for speech recognition. Bayesian combining is achieved by the estimation of joint pdf of confidence feature vector in correct and incorrect hypothesis classes. In addition, the adaptation of a confidence score using the pdf is presented. The proposed methods reduced the classification error rate by 18% from the conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.
ER -