The Research Of Sub-Strokes And Feature Extraction Based On Character Recognition

Off-line Chinese character recognition (OCCR),as one of the technologies to automatic input Chinese characters, is an important interface for Chinese information processing. Having been studied for more than forty years, many achievements have been gained in both theory and technology. Those achievements are far beyond the solution to all problems in OCCR; so, farther research and development in both of them must be made. Offline hand-written Chinese character recognition is more difficult and becomes the main subject in OCCR.The dissertation review a great many of basic principles and methods for Chinese character recognition extensively and analyze the performance evaluation of different approaches. The paper describes basic process of computer character recognition system, gives some common-used Chinese character recognition methods, The main improvements to available character recognition methods are:1,A novel method of stroke extraction is proposed.By analysing the position and the amount of skeletion points generated by morphology skeleton algorithm, The relationship is discovered between the radius of skeleton points and the deformation occurring in stroke extracting. From this discovery, the stroke extraction algorithm is developed to overcome deformation and burr.2,A novel method of Chinese handwritten recognition using complex feature is proposed.The complex feature is composed of the global feature extracted by Hough transformation and the local feature extracted by elastic mesh technique, so the complex feature presents both character's global feature and local feature. Experiments show that this complex feature have higher recognition rate than single elastic mesh feature.Finally, a practical processing system of paper check developed by us is introduced,including the applying environment,primary function and the main technology used in the system.Author's main job is to process the painting number area and to recognize the name character by using the complex feature which extracted by Hough transformation and elastic mesh technique .