Research Of Handwritten Chinese Characters Recognition Based On Varialbe Precision Rough Set

Off-line handwritten Chinese characters recognition is one of the most popular research points in pattern recognition field; it will be widely used in many fields. Because of such factors:too many types of handwritten Chinese character strokes, large differences between different people's writing. Feature selection and classification are still the important topics of handwritten character recognition. By studied the relationship between variable precision rough set theory and handwritten Chinese characters recognition, off-line handwritten Chinese character recognition decision information system was constructed in this paper. The condition attributes were defined as the feature attributes of handwritten Chinese character images, decision attribute was defined as the real property of Chinese character samples. Try using the feature reduction and model inference methods of variable precision rough set theory to study and solve the feature selection and classification problems of handwritten Chinese character recognition.The following are the major contents in the thesis:1. One kind handwritten Chinese character recognition decision information system was constructed based on the concept of information system in generalized domain defined by rough set. Variety features of handwritten Chinese characters were fused into a multi-dimensional feature vector of handwritten Chinese characters as condition attributes, the true value of Chinese character sample was taken as decision attribute.2. The definitions of two kinds handwritten Chinese character attribute significance were given. Handwritten Chinese character feature attribute approximate reduction algorithm was designed based on variable precision rough set. It can reduce the redundant feature attributes of handwritten Chinese character effectively.3. As the existence of uncertainty in the process of handwritten Chinese character recognition, one rules fusion recognition method based on weighted confidence was presented. In three different situations, handwritten Chinese character recognition strategy was given, to improving the generalization ability of decision rules. The rejection recognition rate of Chinese character was reduced.4. Handwritten Chinese characters in SCUT-IRAC were selected. The recognition methods of handwritten Chinese character images based on variable precision rough set were validated. The experiment results shown that those methods are effective.