A Novel Approach for Rotation Free Online Handwritten Chinese Character Recognition

Abstract

This paper presents a method for rotation free online handwritten Chinese character recognition (RFOHCCR). Given a skew online handwritten character sample, two orientation correction steps, including angle rectification according to the starting point, angle readjustment based on principal direction axes, are first performed to rectify the skew angle of the sample. Then 8-directional feature is extracted and the character is classified using the classifier trained by artificially rotated samples. Experiments on 863 online Chinese character dataset and SCUT-COUCH dataset show the effectiveness of the proposed approach.