Research On Off-line Handwritten Chinese Characters Recognition Based On SVM

Off-line handwritten Chinese characters recognition is a popular research point in pattern recognition field, and it can be widely used in many areas. But it is very difficult in the characters recognition system because of the big Chinese characters set, many similar characters and distortion in the handwritten Chinese characters. Support Vector Machines (SVM) is a new machine learning method proposed according to statistical learning theory, which has gained better effect in pattern recognition, regression analysis. So, the research of off-line handwritten Chinese characters has some certain theory meaning and practical value.In this paper, the method of classification recognition combined with binary tree and SVM is used to do the off-line handwritten Chinese characters recognition. The main works are as follows:(1) Handwritten Chinese characters image pre-treatment and feature extraction. The image gradation, smoothing, size normalization of manuscript character images and several feature extraction methods which are used widely are introduced in this paper.(2) Construct the binary tree based on SVM which is to do the handwritten Chinese characters recognition. A method of binary tree based on SVM are presented to solve the problems of multi-classification in off-line handwritten Chinese characters recognition. The methods make coarse classification successfully depending on the theory of binary and SVM and the decomposition of structure of Chinese characters.(3) Research on handwritten Chinese characters recognition. Extract the features of strokes density, direction and wavelet grid and inosculate the features. By the inosculation, it became a new feature date. Make the new feature date as the input of the SVM. The classification based on"one against rest"SVM is adopted to recognition of same character structure.(4) In the simulation, the handwritten Chinese characters in SCUT-IRAC are chosen and make the MATLAB7.0 as simulation tool。The results of simulation indicate:The method of classification recognition combined with binary tree SVM (coarse classification) and"one against rest"SVM (final classification) can make the advantage of the binary tree which have a rapid classification rate and SVM which have a high precision. In the problem of off-line handwritten Chinese character classification, it has a good result.