A SEGMENT SEQUENCE MATCHING METHOD FOR ON LINE CHINESE CHARACTER RECOGNITION

A method based on segment sequence matching is proposed for on line Chinese character recognition. The stroke number, strode order, and stroke position are selected as the three primary recognizing features. Firstly, in order to lessen the dispersion in class, a non linear shape normalization algorithm line density equalization is used to equalize the stroke position. Then, dynamic programming algorithm is used to look for a optimal matching path between a candidate segment sequence and an unknown segment sequence; The matching segment set and lacked and reduced segments are obtained based on the optimal matching path; The distance between an unknown character and every candidate is calculated. Finally, the candidates whose distances are ordered pre ten are deeply determined using the segment matching set and selecting other features, so the final recognizing result is obtained. The experimental results show that, the proposed method could yield good recognizing effect.