Discriminative Training for HMM-Based Offline Handwritten Character Recognition

Authors:
Roongroj Nopsuwanchai;Dan Povey
Affiliations:
-;-
Venue:
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
Year:
2003

Citing 3
Cited 0

On the Advantages of Polar and Log-Polar Mapping for Direct Estimation of Time-To-Impact from Optical Flow

IEEE Transactions on Pattern Analysis and Machine Intelligence
Offline Cursive Word Recognition using Continuous Density Hidden Markov Models Trained with PCA or ICA Features

ICPR '02 Proceedings of the 16 th International Conference on Pattern Recognition (ICPR'02) Volume 3 - Volume 3
Minimum classification error training of hidden Markov models for handwriting recognition

ICASSP '01 Proceedings of the Acoustics, Speech, and Signal Processing, 2001. on IEEE International Conference - Volume 03

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper we report the use of discriminative trainingand other techniques to improve performance in a HMM-basedisolated handwritten character recognition system.The discriminative training is Maximum Mutual Information(MMI) training; we also improve results by using compositeimages which are the concatenation of the raw images,rotated and polar transformed versions of them; andwe describe a technique called block-based Principal ComponentAnalysis (PCA). For effective discriminative trainingwe need to increase the size of our training database, whichwe do by eroding and dilating the images to give a three-foldincrease in training data. Although these techniquesare tested using isolated Thai characters, both MMI andblock-based PCA are applicable to the more difficult task ofcursive handwriting recognition.