Title :
Minimum classification error training for online handwritten word recognition
Author_Institution :
IBM Thomas J. Watson Res. Center, Yorktown Heights, NY, USA
Abstract :
We describe an application of the minimum classification error (MCE) training criterion to online unconstrained-style word recognition. The described system uses allograph-HMMs to handle writer variability. The result, on vocabularies of 5k to 10k, shows that MCE training achieves around 17% word error rate reduction when compared to the baseline maximum likelihood system.
Keywords :
Bayes methods; decision theory; handwritten character recognition; hidden Markov models; optimisation; parameter estimation; probability; allograph-HMMs; baseline maximum likelihood system; hidden Markov modeling; minimum classification error training; online handwritten word recognition; online unconstrained-style word recognition; writer variability; Error analysis; Handheld computers; Handwriting recognition; Hidden Markov models; Personal digital assistants; Shape; Signal processing; Speech recognition; Vocabulary; Writing;
Conference_Titel :
Frontiers in Handwriting Recognition, 2002. Proceedings. Eighth International Workshop on
Print_ISBN :
0-7695-1692-0
DOI :
10.1109/IWFHR.2002.1030885
