Structure based modern Korean character set partitioning and pre-classification method of korean character recognition

With respect to practical Korean documents, structure characteristics and structure classification of Korean character and a new pre-classification method of character recognition were studied in this paper. Firstly, the concept of character structure distance and its simple calculation method were proposed to describe the difference between structures, according to the unique linearizability of each character and the occurrence of basic graphemes. Furthermore, based on ID3 algorithm and the information gain of graphemes, the decision tree of structure classification was constructed. Finally, a decision tree based pre-classification algorithm was designed for printed character recognition. The experimental results show that basic graphemes v₂, c₃, v₃ and v₁ or c₄ possess high contribution for character pre-classification, and the proposed method has reliable theoretical foundation and effective classification performance.