Off-line Handwritten Chinese Character Recognition And Study Based On Strokes

The recognition of offline Chinese characters use computer to identify the characters which have been printed on paper or written down on paper automatically, it refer to many subjects, such as pattern recognition, image processing, artificial intelligence, formal language and automatic, Chinese information, the combination of mathematics, fuzzy mathematics, information theory, processing and so on, but it also refer to Psychology, language learning, bionics and so forth. The recognition of handwritten Chinese characters is an important branch of Chinese characters recognition, it is also an indispensable category in pattern recognition and artificial intelligence.At present, the recognition technology of on-line handwritten Chinese characters has made rapid progress and has a good recognition effect, it can basically meet the needs of users in real time and accuracy. For example, recognition technology has achieved remarkable results in some specific applications, such as reading email address automatically, processing bank checks and bills. However, the application scope of on-line handwritten identification technology is relatively narrow, which also have greater restrictions on the Chinese characters writing. So it cannot meet the basic needs of users, because in our daily life, a lot of handwritten copy table and handwritten document need to be input to computer, how to put them into the computer in high efficiency, and make them into digital information, in addition, huge historical documents need to be handled, if put characters to computer one by one, it is a great project which will cost too much in man-power, material and money.In addition, handwritten Chinese characters have complex structure, too much similar characters, great characters set, more comprehensive shape, and different writing style with different people. Despite the off-line handwritten Chinese characters recognition has experienced decades of years, but there is still no mature products, technology still need to be developed, and it is still the focus of home and abroad, it is also a challenging problem in Chinese characters recognition technology. But the printed-Chinese characters recognition technology has walked out of the laboratory, and has widely used in many aspect, on-line handwritten Chinese characters recognition has become more mature and get to commercialization.Based on the research which is focus on off-line handwritten Chinese characters recognition, a way of off-line handwritten Chinese characters recognition with stroke has been put forward. Because most of Chinese characters cannot leave without four strokes, the four kinds of stroke are horizontal line, top-down vertical line, left-downward slope line, short pausing stroke. The proportion of stroke in handwriting Chinese characters is33.94%,16.77%,9.78%and39.51%. although the characters have different shape and size, but the stroke of characters is relatively stable, simple statistical feature extraction and classification algorithm had been used to Chinese characters handwritten recognition, but it cannot solve the difficulty of handwritten Chinese characters recognition radically.This paper adopts the off-line handwritten Chinese characters recognition, it is divided into three steps, which are preprocessing, extracting feature, recognizing the extraction feature. Firstly, preprocessing the handwritten Chinese characters, the preprocessing based on six steps, image gray value of two, smoothing and de-noising, image segmentation, normalization and refinement. Through the preprocessed of handwritten Chinese characters image, it effectively maintain the original image information of Chinese characters and handwritten character, weaken or reduce all kinds of interference factors which existed in the original image, so as to optimize the original image effect. Of course, with no doubt, the result of preprocessing will directly affect the efficiency of feature extraction.The second step is feature extraction. This paper adopts the following steps:1extracting stroke bifurcation point of handwriting Chinese characters which have been preprocessed;2extracting strokes inflection point with maximum distance method;3extracting the tilt and endpoint coordinates of stroke;4. repairing various distortion unavoidable in preprocessing;5. combining pseudo cross point in preprocessing;6.structuring the structure of handwriting Chinese characters.The third step is matching recognition. In the stage of matching stroke feature, firstly, in order to establish stroke template library, this paper take Chinese characters strokes of the handwritten samples as the features to be stored in template library, and training stroke template library, and then comparing the each unrecognized Chinese characters of handwritten samples to the stroke template library, and calculating the combined distance which is from unrecognized Chinese characters of handwritten samples to the stroke template library,finally, taking the minimum distance to recognize.