Recognition of Korean Characters

초록

We propose a character recognition algorithm for Korean characters by employing attribute dependent regular programmed grammar(ADRPG). The difficulty of this problem from the fact that each Korean character has many strokes. Generally, the recognition problem is more difficult if there are more strokes in a character. The ADRPG can easily incorporates attribute information in character pattern, such as position, angle, length, and branch points to recognize characters with many strokes. The proposed algorithm is tested with printed Hangul characters. In the first experiment with high quality printed document having 4,206 characters, the proposed method recognized characters 100% correctly after improving syntactic rules by training. With the same syntactic rules used in the first experiment, the algorithm is tested with a document having 3,092 characters, some of them have defects. In the second experiment, it achieved 95.1% of correct recognition.