Tied Mixture modeling optimization for Korean-digit in the embedded ASR system

Kihyeon Kim, Hanseok Ko

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In the embedded Automatic Speech Recognition (ASR) system, Semi-Continuous Hidden Markov Model (SCHMM) or Tied-Mixture (TM) model is one of the most promising acoustic modeling methods that solve the size problem of the existing Continuous Hidden Markov Model (CHMM) while minimizing the recognition performance degradation. Moreover, for a general isolated word task, context dependent models such as tri-phones are used to guarantee high recognition performance of the embedded system. However, to use the models constructed only in this way alone cannot be sufficient to render improved recognition rate in Korean-digit speech task where a large mutual similarity exists. Hence, we construct new dedicated HMM's for all or parts of Korean-digit that has exclusive states using the same Gaussian pool of previous tri-phone models. This remedial action allows the structure of entire HMMs maintained while minimizing the occupied memory space. Representative experiments are expected to reduce word-error-rate on the Korean-digit task by about 86% in comparison with using only general tri-phone models.

Original languageEnglish
Title of host publication2004 IEEE International Symposium on Consumer Electronics - Proceedings
Pages595-599
Number of pages5
Publication statusPublished - 2004 Dec 27
Event2004 IEEE International Symposium on Consumer Electronics - Proceedings - Reading, United Kingdom
Duration: 2004 Sep 12004 Sep 3

Other

Other2004 IEEE International Symposium on Consumer Electronics - Proceedings
CountryUnited Kingdom
CityReading
Period04/9/104/9/3

Fingerprint

Speech recognition
Hidden Markov models
Embedded systems
Acoustics
Data storage equipment
Degradation
Experiments

Keywords

  • Embedded ASR System
  • Exclusive HMM's
  • Korean Digits
  • Tied Mixture Model

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Kim, K., & Ko, H. (2004). Tied Mixture modeling optimization for Korean-digit in the embedded ASR system. In 2004 IEEE International Symposium on Consumer Electronics - Proceedings (pp. 595-599)

Tied Mixture modeling optimization for Korean-digit in the embedded ASR system. / Kim, Kihyeon; Ko, Hanseok.

2004 IEEE International Symposium on Consumer Electronics - Proceedings. 2004. p. 595-599.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Kim, K & Ko, H 2004, Tied Mixture modeling optimization for Korean-digit in the embedded ASR system. in 2004 IEEE International Symposium on Consumer Electronics - Proceedings. pp. 595-599, 2004 IEEE International Symposium on Consumer Electronics - Proceedings, Reading, United Kingdom, 04/9/1.
Kim K, Ko H. Tied Mixture modeling optimization for Korean-digit in the embedded ASR system. In 2004 IEEE International Symposium on Consumer Electronics - Proceedings. 2004. p. 595-599
Kim, Kihyeon ; Ko, Hanseok. / Tied Mixture modeling optimization for Korean-digit in the embedded ASR system. 2004 IEEE International Symposium on Consumer Electronics - Proceedings. 2004. pp. 595-599
@inproceedings{e15086710fe641bdbc707b1096bcba23,
title = "Tied Mixture modeling optimization for Korean-digit in the embedded ASR system",
abstract = "In the embedded Automatic Speech Recognition (ASR) system, Semi-Continuous Hidden Markov Model (SCHMM) or Tied-Mixture (TM) model is one of the most promising acoustic modeling methods that solve the size problem of the existing Continuous Hidden Markov Model (CHMM) while minimizing the recognition performance degradation. Moreover, for a general isolated word task, context dependent models such as tri-phones are used to guarantee high recognition performance of the embedded system. However, to use the models constructed only in this way alone cannot be sufficient to render improved recognition rate in Korean-digit speech task where a large mutual similarity exists. Hence, we construct new dedicated HMM's for all or parts of Korean-digit that has exclusive states using the same Gaussian pool of previous tri-phone models. This remedial action allows the structure of entire HMMs maintained while minimizing the occupied memory space. Representative experiments are expected to reduce word-error-rate on the Korean-digit task by about 86{\%} in comparison with using only general tri-phone models.",
keywords = "Embedded ASR System, Exclusive HMM's, Korean Digits, Tied Mixture Model",
author = "Kihyeon Kim and Hanseok Ko",
year = "2004",
month = "12",
day = "27",
language = "English",
isbn = "0780385276",
pages = "595--599",
booktitle = "2004 IEEE International Symposium on Consumer Electronics - Proceedings",

}

TY - GEN

T1 - Tied Mixture modeling optimization for Korean-digit in the embedded ASR system

AU - Kim, Kihyeon

AU - Ko, Hanseok

PY - 2004/12/27

Y1 - 2004/12/27

N2 - In the embedded Automatic Speech Recognition (ASR) system, Semi-Continuous Hidden Markov Model (SCHMM) or Tied-Mixture (TM) model is one of the most promising acoustic modeling methods that solve the size problem of the existing Continuous Hidden Markov Model (CHMM) while minimizing the recognition performance degradation. Moreover, for a general isolated word task, context dependent models such as tri-phones are used to guarantee high recognition performance of the embedded system. However, to use the models constructed only in this way alone cannot be sufficient to render improved recognition rate in Korean-digit speech task where a large mutual similarity exists. Hence, we construct new dedicated HMM's for all or parts of Korean-digit that has exclusive states using the same Gaussian pool of previous tri-phone models. This remedial action allows the structure of entire HMMs maintained while minimizing the occupied memory space. Representative experiments are expected to reduce word-error-rate on the Korean-digit task by about 86% in comparison with using only general tri-phone models.

AB - In the embedded Automatic Speech Recognition (ASR) system, Semi-Continuous Hidden Markov Model (SCHMM) or Tied-Mixture (TM) model is one of the most promising acoustic modeling methods that solve the size problem of the existing Continuous Hidden Markov Model (CHMM) while minimizing the recognition performance degradation. Moreover, for a general isolated word task, context dependent models such as tri-phones are used to guarantee high recognition performance of the embedded system. However, to use the models constructed only in this way alone cannot be sufficient to render improved recognition rate in Korean-digit speech task where a large mutual similarity exists. Hence, we construct new dedicated HMM's for all or parts of Korean-digit that has exclusive states using the same Gaussian pool of previous tri-phone models. This remedial action allows the structure of entire HMMs maintained while minimizing the occupied memory space. Representative experiments are expected to reduce word-error-rate on the Korean-digit task by about 86% in comparison with using only general tri-phone models.

KW - Embedded ASR System

KW - Exclusive HMM's

KW - Korean Digits

KW - Tied Mixture Model

UR - http://www.scopus.com/inward/record.url?scp=10444277350&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=10444277350&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:10444277350

SN - 0780385276

SP - 595

EP - 599

BT - 2004 IEEE International Symposium on Consumer Electronics - Proceedings

ER -