A New Similarity Evaluation Function for Writer Recognition of Chinese Character

Abstract

In new local arc method which we have proposed before, the author has been specified from curvature distributions of three different kinds of characters. This time, a new method to obtain similarities from a superposition of four similarity values obtained from two chord lengths and two different kinds of characters, was tried in order to apply the new local arc method to the Chinese character (Simplified characters). As the result, the writer recognition became possible at the accuracy which could not be realized by the simple new local arc method. And by this method, writer recognition became possible by fewer kinds of characters.