Minimum error rate training for PHMM-based text recognition

Authors
Yen, CC Kuo, SS Lee, CH
Citation
Cc. Yen et al., Minimum error rate training for PHMM-based text recognition, IEEE IM PR, 8(8), 1999, pp. 1120-1124
Citations number
10
Categorie Soggetti
Eletrical & Eletronics Engineeing
Journal title
IEEE TRANSACTIONS ON IMAGE PROCESSING
ISSN journal
10577149 → ACNP
Volume
8
Issue
8
Year of publication
1999
Pages
1120 - 1124
Database
ISI
SICI code
1057-7149(199908)8:8<1120:MERTFP>2.0.ZU;2-L
Abstract
In this work, discriminative training is studied to improve the performance of our pseudo two-dimensional (2-D) hidden Markov model (PHMM) based text recognition system. The aim of this discriminative training is to adjust mo del parameters to directly minimize the classification error rate. Experime ntal results have shown great reduction in recognition error rate even for PHMM's already well-trained using conventional maximum likelihood (ML) appr oaches.