Off-line Handwritten Chinese Character Recognition Based On Multilayer And Overlapped Meshes

The off-line handwritten Chinese character recognition is an important topic in computer intellectualization domain. With the rapid development of character recognition,it has made considerable breakthrough in the technological application. But how to design the stronger recognition method has been the difficulty of affecting its development. This paper conducts the research in view of the off-line handwritten Chinese character recognition method. We propose the method that utilizing overlapped and multilayer meshes to extract strokes to classify the characters, and carrying on the recognition by use of the number of periphery-stroke-type-penetration and the gravity-neighborhood-stroke-penetration .In view of the fact that the number of character is massive and its structure is complex, we utilize the pattern of rough classification and detailed identification . we first carry on the false refinement to the handwritten Chinese character, and utilize the fuzzy theory to extract strokes. After revising and merging the strokes, we constrcut feature vectors through multilayer and overlapped meshes to classify the characters in consider of the stroke relations of various levels meshes. The rough classfication uses the strategy of the level machine-made. It belongs to father-son or brother relations between each kind. Detailed identification constructs feature vectors by use of the number of periphery- stroke- type- penetration and the gravity-neighborhood-stroke-penetration . It utilizes minimum range sorter based on Euclidian distance to carry on detailed identification. If the similarity between the character and every characters of one kind is small, we compare it with the characters of its father kind. It has confirmed that the method proposed in this paper is valid through the experiments on different samples.