Study Of Off-line Handwritten Chinese Character Recognition Based On Muui-features Pruned Binary Tree

Offline handwritten Chinese character recognition technology is animportant research topic in the field of pattern recognition。 As Offlinehandwritten Chinese character has a number of large, complex structure andrange of fonts and writing arbitrary, Offline handwritten Chinese characterrecognition has been one of the most difficult problems of the field of Chinesecharacter recognition. Support vector machine (SVM) is a new machinelearning algorithm which is proposed according to statistical learning theory.It has already achieved good effects in the aspects of pattern recognition andregressive analysis, etc.Offline handwritten Chinese character as a researchobject, A dynamic pruned binary tree method of off-line handwritten Chinesecharacter recognition based on Multi-Features SVM is discussed in thispaper,Providing a new idea for handwritten Chinese character recognition.This paper mainly focuses on the following points:(1)On the basis of the statistical characteristics and of Offlinehandwritten Chinese character offline Chinese characters and structuralcharacteristics of Offline handwritten Chinese character Chinese characters,Todefine the pixel density characteristics of the off-line handwritten Chinesecharacter. Morover, it is Divided into the characteristics of the overall pixeldensity, stroke (horizontal,vertical,oblique) pixel density characteristics,Giventhe method of access to the features of handwritten Chinese characters.(2)To provide a method of Coarse classification of offline handwrittenChinese character based on Multi-Features SVM dynamic pruned binarytree.Based on researching the generalized density characteristics of offlinehandwritten Chinese characters. Based on Multi-Features of offlinehandwritten Chinese characters, SVM dynamic pruned binary tree isbuilded,Morover, To provide a method of Coarse classification of offlinehandwritten Chinese character based on Multi-Features SVM dynamic prunedbinary tree.(3)To provide a method of off-line handwritten Chinese character SVM dynamic pruned coarse classification. Based on the analysis of the feature ofpercentage of black pixels of Chinese character, statistical feature of basicstrokes and structural feature, dynamic pruned binary tree based the abovethree features is constructed, and an algorithm of off-line handwritten Chinesecharacter coarse classification based on SVM dynamic pruned binary tree.By pruning the character set unrelating to the features of Chinese character tobe classified, the speed of coarse classification is increased, with avoidingmiss-classification and lost-classification as much as possible.In this paper, SCUT-IRAC HCCLIB handwritten Chinese charactersample is selected as the laboratorial data. MATLAB7.0are used as thesimulation software, After verifying the method of the dynamic pruned binarytree SVM coarse classification and then meticulous classification, it is provedthat this method is effective.