Minimum error rate training for PHMM-based text recognition

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 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.

[1]  Chris A. Glasbey,et al.  An Analysis of Histogram-Based Thresholding Algorithms , 1993, CVGIP Graph. Model. Image Process..

[2]  Henry S. Baird,et al.  Document image defect models , 1995 .

[3]  Biing-Hwang Juang,et al.  New discriminative training algorithms based on the generalized probabilistic descent method , 1991, Neural Networks for Signal Processing Proceedings of the 1991 IEEE Workshop.

[4]  Biing-Hwang Juang,et al.  Discriminative learning for minimum error classification [pattern recognition] , 1992, IEEE Trans. Signal Process..

[5]  Shyh-shiaw Kuo,et al.  Degraded Gray-Scale Text Recognition Using Pseudo-2D Hidden Markov Models and N-Best Hypotheses , 1995, CVGIP Graph. Model. Image Process..

[6]  Josef Kittler,et al.  Minimum error thresholding , 1986, Pattern Recognit..

[7]  Oscar E. Agazzi,et al.  Pseudo two-dimensional hidden Markov models for document recognition , 1993, AT&T Technical Journal.

[8]  Chin-Hui Lee,et al.  Segmental GPD training of HMM based speech recognizer , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[9]  Biing-Hwang Juang,et al.  Minimum error rate training based on N-best string models , 1993, 1993 IEEE International Conference on Acoustics, Speech, and Signal Processing.