Off-line Handwritten Chinese Character Recognition Based On Four Corner Structural Feature

The development of off-line handwritten Chinese character recognition in recent 20 years illustrates two facts: the demand and the feasibility of off-line handwritten Chinese character recognition technology. But the transshape of Chinese character has been the major factor restricting of the development .In this paper ,we propose a method of off-line handwritten Chinese character recognition based on four corner feature.The strategy of rough classification and detailed identification is adopted in the recognition process. Firstly, a thinning algorithm is used to get the skeleton of the Chinese character and optimization process is done on the skeleton based on the feature point information .Then ,Chinese character rough classification is achieved by four corner structure features. The traditional four-corner-code indexing system for Chinese characters can map the character to the code one to one with the stroke primitive. In this paper, stroke primitives those easy to be extracted are redesigned to extract four corner features based on the most steady 3 features which are inflexion, crossing point and gradient. After the rough classification , characters in each set have the similar structure in four corners but very different in whole which is not like the traditional similar characters .Lastly ,some simple features that can imply the structure of the character are used as recognition features in detailed identification phase. the character classification is carried out by the minimum distance classifier.The experiment on 500 Chinese characters from GB2312 shows that our approach can achieve a satisfied result in recognition for handwritten Chinese characters.