Hidden Markov Models for Character Recognition

Multifont and handwritten character recognition systems have not been successfully implemented to date because of the varability of characters and the difficulty of incorporating context in the classification process. Hierarchical systems are very useful in making context-based decisions but their knowledge sources cannot easily incorporate both general knowledge about objects and at the same time knowledge about object instantiations. In this paper we present a hierarchical system for character recognition with hidden Markov model knowledge sources which solve both the context sensitivity problem and the character instantiation problem. Our system achieves 97-99% accuracy using a two level architecture and has been implemented using a systolic array, thus permitting real time (1 ms per character) multifont and multisize printed character recognition as well as handwriting recognition.