Highly accurate recognition of printed Korean characters through an improved two-stage classification method

This paper presents a recognition system which obtains a recognition rate higher than 99% for the printed Korean characters of multifont and multisize. We recognize a given input by first identifying the character type of the input and then recognizing its constituent graphemes. In order to improve the performance we incorporated three new ideas in our system: the expansion of the subimage areas used by the grapheme classifiers, an algorithm to accurately segment the horizontal vowel’s subimage areas, and a validation process to evaluate the result of the type classifier. Through experiments we confirmed that our system performs well in a multi-font and multi-size environment and that those three ideas actually contributed to improve the performance significantly.