Development of Thai word recognition system for esophageal speaker using model adaptation based on HMM

Conference proceedings article

Authors/Editors

Strategic Research Themes

No matching items found.

Publication Details

Author list: Sabayjai P., Boonpranuk P., Polwisate W., Kayasith P.

Publisher: Hindawi

Publication year: 2008

Start page: 70

End page: 73

Number of pages: 4

ISBN: 9789810803681

eISSN: 1745-4557

URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-82455209367&partnerID=40&md5=f5423322afb9d15086b0fbfd340e0e7d

Languages: English-Great Britain (EN-GB)

Abstract

The paper represents a method to develop an automatic speech recognition system for Thai esophageal speaker. Using model adaptation approach, a speech recognition built by normal speech corpus is modified by esophageal speech model based on Hidden Markov Models (HMMs). The aim of the paper is to improve the recognition rate of esophageal speech using two different approach; model adaptation technique and three cepstral normalization technique. The experimental results show that the model adaptation technique can significantly improve the recognition rate of esophageal speech from 12.6% up to 78.1%. Moreover, the recognition rate is raised up to 93.0% when a cepstral normalization technique (CMMN) is combined to the model adaption.

Keywords

Feature normalization, Hidden Markov Models, Model adaptation