A fuzzy-attributed graph approach to handwritten character recognition

A handwritten character recognition system is described. It includes a simple but effective thinning algorithm to produce skeletons from raster-scanned images which contain characters to be recognized. Based on this provision of skeletons, each character is decomposed into a number of connected strokes. Each stroke is classified as a primitive according to a fuzzy similarity criterion. When the structural information like the relations between contiguous primitives and the attributes of each primitive are extracted, a fuzzy-attributed graph can then be constructed to represent the character. A method is proposed for measuring the similarity between two fuzzy-attributed graphs of a known character class and an unknown character. In the recognition stage, when this similarity measure is applied, an input character can be correctly classified. Experimental results show that an accuracy rate of around 93% is achieved by this approach