We describe several approaches for the application of hidden Markov models to the recognition of handwritten words. In all approaches the words are described by strings of symbols. The descriptions differ only with respect to the size of the vocabulary to be recognized. We can define two distinct cases: in the first, the vocabulary is small and constant; in the second, the vocabulary is limited, but dynamic in the sense that it is a varying subset of an open one. We also describe an application of hidden Markov models to the representation of contextual knowledge and propose some strategies for rejecting unreliable word interpretations, in particular, when the word corresponding to the image does not necessarily belong to the lexicon.
Download Full PDF Version (Non-Commercial Use)