On the intra-category clustering to make multidictionary patterns for multidictionary template matching method

The applications of a template-matching method for handprinted Japanese kanji character recognition are discussed. Various clustering methods for the intracategory clustering to make the optimal multiple templates are examined. The efficiencies of each method with handprinted kanji characters, using the ratio of variances, are evaluated, and a character recognition experiment are executed with the multiple templates made by the best clustering method.<
>