Title :
Minimum error rate training for PHMM-based text recognition
Author :
Yen, Chinching ; Kuo, Shyh-shiaw ; Lee, Chin-Hui
Author_Institution :
Lucent Technol., AT&T Bell Labs., Holmdel, NJ, USA
fDate :
8/1/1999 12:00:00 AM
Abstract :
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 model parameters to directly minimize the classification error rate. Experimental results have shown great reduction in recognition error rate even for PHMMs already well-trained using conventional maximum likelihood (ML) approaches
Keywords :
classification; error statistics; hidden Markov models; image recognition; MLE; PHMM-based text recognition; classification error rate minimisation; discriminative training; experimental results; maximum likelihood estimation; minimum error rate training; model parameters adjustment; performance; pseudo 2D hidden Markov model; recognition error rate reduction; Document image processing; Error analysis; Hidden Markov models; Image recognition; Maximum likelihood estimation; Parameter estimation; Pattern classification; Pattern recognition; Text recognition; Two dimensional displays;
Journal_Title :
Image Processing, IEEE Transactions on
