A new classifier for handwritten Chinese character recognition using 2-dimensional functional relationship model

This paper presents a new classification method for online handwritten Chinese character recognition (HCCR). For classification, a similarity measure is established via statistical technique which calculates the coefficient of determination (R_p ²) for 2-dimensional unreplicated linear functional relationship (MULFR) model between the trajectory pattern of input character and character in database, according to which the recognition result is determined. The principle of the proposed method makes R_p ² very robust against size and position variation as well as stroke shape deformation, without normalization. The efficiency of our proposed method is studied by the experimental result, showing that the proposed method still remains a promising recognition rate even without undergoing normalization if compared to city block distance with deviation (CBDD) and minimum distance (MD) classifier: a high recognition rate of 94% with reduced processing time up to 77.85%.