A Study Of Off-line Handwritten Chinese Character Recognition System Based On Rough Set And Variable Granular Theorem

Off-line handwritten Chinese character recognition is a super multi-classpattern recognition problem, which is considered one of the difficulties in the fieldof pattern recognition currently. There are a lot of limitations that using a singlefeature to represent the handwritten Chinese character. It is one of the methods toimprove the completeness of feature representation by multi-feature combination,but it caused multiplicity of fearures and uncertainties of recognition. In this paper,based on the rough set theory and variable granular theorem, a kind of decisioninformation system of off-line handwritten Chinese character recognition isconstructed, the actual attributes of the sample characters are used to guide thetraining process, a method of off-line handwritten Chinese character recognition isexplored.The main research works of this paper are as follows:1. A decision information system of off-line handwritten Chinese characterrecognition is constructed. Based on the general domain of information system inrough set theory, the condition attributes are made of multi-feature combination ofoff-line handwritten Chinese character, and the decision attributes are made of thereal property of Chinese character samples.2. A hierarchical reduction approach of handwritten Chinese characterattributes is given. According to the granular theorem, the knowledge granularentropy of feature attributes, the relative granular entropy and the attributesignificance are defined. The feature attributes are classified by the significance,that the representation of feature attributes in single layer and single granularitybecome multiple layers and multiple granularities. A hierarchical reductionapproach is designed, that the weak points of the single granularity that fixed andreduction is not fine are made up.3. Based on the D-S evidence theory, a rules fusion recognition method isproposed. To solve the problem that decision rules can not be matched uniquely, thedecision rules after reduction are fused by the D-S evidence theory, so the decisioninformation system can be more generalized, and the off-line handwritten Chinese character recognition rate can be improved.In this paper,30classes handwritten Chinese character images are chosenfrom ‘SCUT-IRAC HCCLIB’ handwritten Chinese character samples database,there are40character images of every class. The method of off-line handwrittenChinese character recognition which proposed in this paper is tested and verifiedon the MATLAB software. The experiment results show that the method is feasibleand effective.