Study on Chinese handwriting identification based on texture analysis

Abstract

As a kind of behavior-based personal identification techniques, automated Chinese handwriting identification becomes a hot topic in pattern recognition and machine learning research area. There are lots of key issues worthy researching. In this paper, the Chinese handwriting identification technology based on texture analysis is discussed. Firstly, a practical Chinese handwriting image samples library CHSL2007 is established for the comparison of exist algorithms and further research. Then the methods of feature extraction based on texture analysis are explored and the pairwise SVM classifier is utilized. The experiment results of texture analysis based on Gabor filter is compared with DB6 wavelet filter and demonstrate that the former is more suitable for handwriting identification on CHSL2007. Finally, the sheet recognition rate is defined and can be arrived at above 99.50% for CHSL2007.