Research, Theory And Algorithm Of Two-dimensional Hidden Markov Model-based Character Recognition

Hidden Markov models are statistic models, whose shortened form is HMMs. It contains two parts: a state Markov chain, which is a Markov process that can't be observed directly; a stochastic process can be observed directly, which is related to the states. In 1960s L.E.Baum brought hidden Markov models forward and in 1970s these models were applied into automatic speech recognition by Jenik and some other people and developed to be one of the most efficient techniques in the field. Now, they were widely used in gene relative analysis and gene recognition, character recognition, image processing and target tracking etc. There are three problems needed to be solved by using Hidden Markov models ,which is training, decoding and recognizing. The answers for those three problems consist of the theory of hidden Markov models. Parameter estimation is the core problem of the training process.Parameter Estimation and the applications in handwriting character recognition of 2-D stationary HMMs were discussed in this paper. Main contributions are local optimal decoding algorithm of Hidden Markov model that was applied into off-line handwriting character recognition and parameter estimations. An abridged algorithm and parameter estimation of Hidden Markov models applied in character recognition were presented under simple hypothesis. Some problems of 2-D HMMs solved by using fuzzy measure were discussed and advantages were stated in this paper.