Recognition of Handwritten Chinese Character Based on Least Square Support Vector Machine

Abstract

Recognition of handwritten Chinese charater has been applied to diversified fields in terms of industrial demands as well as in daily life, since transformation from handwritten charaters into computer-processible binary digits inevitably bring people convenience and joy. However such ubiquitous facility suffers drawbacks within current recognition schema, such as complex training process, low recognition accuracy and slow identification. In light of these dissatisfation, a novel recognition method is proposed to hadle Chinese characters, which is based on the least square support vector machine. This approach evades solving traditional QP problem in the stage of machine learning where the training is time consuming. It, however, works in a way that transforms the recognition constraints into a series of generalized inequitions. Test results show that the proposed method enjoys better recognition acccuracy compared with existent approaches.