On-line handwritten Kanji character recognition using hypothesis generation in the space of hierarchical knowledge

Abstract

For online handwritten Kanji recognition, a new approach that cyclically generates a more concrete hypothesis from the current hypothesis by using hierarchically represented knowledge and that has achieved high recognition rate in reduced processing time is described. The properties of a characteristic stroke are represented in the form of fuzzy rules. Characteristic strokes, each of which is close to part of an input pattern stroke, are found by fuzzy inference, where each input pattern stroke is investigated by using fuzzy rulers.<>