A new similarity measurement method for fuzzy-attribute graph matching and its application to handwritten character recognition

Abstract

For problems of pattern recognition and classification procedures, the concept of fuzziness is usually applied when patterns are mapped into the feature space. To handle such fuzzy concepts, the authors extend the attributed graph to the fuzzy-attribute graph (FAG) by making the attributes fuzzy. The attributed graph is widely used as a straightforward representation of structural patterns. The vertices of the graph represent pattern primitives describing the pattern while the arc is the relation between these primitives. The authors give a formal definition of FAGs and introduce a new method of similarity measurement for FAGs to solve the problem of graph matching. A handwritten numeral recognizer was developed to evaluate the effectiveness of the proposed algorithm, and it gave an accuracy of 93% in terms of correct classification.<>