Data-driven design of HMM topology for online handwriting recognition

Authors:
Jay J. Lee;Jahwan Kim;Jin H. Kim
Affiliations:
KAIST, Taejon, Korea;KAIST, Taejon, Korea;KAIST, Taejon, Korea
Venue:
Hidden Markov models
Year:
2001

Citing 6
Cited 1

Dynamic programming inference of Markov networks from finite sets of sample strings

IEEE Transactions on Pattern Analysis and Machine Intelligence
Fundamentals of speech recognition

Fundamentals of speech recognition
HMM Based On-Line Handwriting Recognition

IEEE Transactions on Pattern Analysis and Machine Intelligence
Ligature Modeling for Online Cursive Script Recognition

IEEE Transactions on Pattern Analysis and Machine Intelligence
The acoustic-modeling problem in automatic speech recognition

The acoustic-modeling problem in automatic speech recognition
A Comparison Between Continuous and Discrete Density Hidden Markov Models for Cursive Handwriting Recognition

ICPR '96 Proceedings of the 13th International Conference on Pattern Recognition - Volume 2

Adaptive online multi-stroke sketch recognition based on hidden markov model

ICMLC'05 Proceedings of the 4th international conference on Advances in Machine Learning and Cybernetics

Quantified Score

Hi-index	0.00

Visualization

Abstract

Although HMM is widely used for online handwriting recognition,there is no simple and well-established method of designing the HMMtopology. We propose a data-driven systematic method to design HMMtopology. Data samples in a single pattern class are structurallysimplified into a sequence of straight-line segments. Then theresulting multiple models of the class are combined to form anarchitecture of a multiple parallel-path HMM, which behavesas single HMM. To avoid excessive growing of the number of thestates, parameter trying is applied such that structural similarityamong patterns is reflected. Experiments on online Hangulrecognition showed about 19% of error reductions, compared to theintuitive deisgn method.