Hidden Markov models applied to on-line handwritten isolated character recognition

Hidden Markov models are used to model the generation of handwritten, isolated characters. Models are trained on examples using the Baum-Welch optimization routine. Then, given the models for the alphabet, unknown characters can be classified using maximum-likelihood classification. Experiments have been conducted, and an average error rate of 6.9% was achieved over the alphabet consisting of the lowercase English alphabet.