The Research Of Handwritten Chinese Characters Handwriting Identification

Handwritten Chinese characters handwriting identification is a very important researchdirection in the field of pattern recognition. It has broad application prospects in the field ofpublic security, finance, etc and so on. Although there are many kinds of handwritingidentification methods at present, but it is difficult to find one can ensure reliability andaccuracy at the same time. So in this paper, using the information fusion theory organiccombined three methods of identification.First of all, this paper proposes the handwriting image block based on the analysis ofmoments feature. And then by extracting the center of gravity feature and shape feature fromeach handwriting block, and comprehensive each handwriting image block’s center of gravityfeature and shape feature and get the feature vector. And then to calculate each the distancebetween the feature vector of the handwriting under test and the sample handwriting imagefeature vector to do handwriting identification. Secondly, the texture features of thehandwriting image can describe a person’s character of writing habit. And there are a lot ofresearches show that the importance of handwriting texture analysis in handwritingidentification. In this paper, the Local Binary Pattern histogram as the texture feature, appliedto the handwriting identification, and obtained the very good identification results.Finally, put forward a new method which fuse the identification method based onmoment features and Local Binary Pattern histogram in the decision-making level, in theresult by adopting the idea of the weighted fusion. Information fusion can comprehensive theidentification results feature from different angles, and has good adaptability. Considering thereliability of the results of the first few, each sort results for the different fusion weights. Theexperimental results show that fusion the identification method based on moment features andthe identification method based on Local Binary Pattern histogram, the identification resultsaccuracy is higher than both used alone. The first five selected identification accuracy canreach96.7%, and this method has broader application scope.