Research On Feature Combination And Similar Character Recognition For Online Chinese Handwritten Characters Recognition System

At present, On-Line Chinese Character Recognition (OLCCR) is a hotspot branch of the pattern recognition fields, and it gets a lot of success. The more and more applications in computer based on OLCCR come out. It will be widest for the application of OLCCR, because this technique accords with the nature habit when people write, and it can also solve the problem of slowness in typing and the fussiness in operating. Although the OLCCR is promising, there are still some problems in this technique.In this thesis, the methods in preprocess are introduced to deal with several kinds of primary noise, and the analysis is gave out; and an algorithm for Segment-Combined based on Deterministic Finite Automata is proposed .After the process of preprocess, this algorithm can utilize the state of segment to combine them automatically.After analyzing the reason caused to misrecognize the similar Chinese characters, a partial area matching method based on Support Vector Machine for similar Chinese characters recognizing is proposed. This algorithm can automatically find the similar characters out, and utilize the Support Vector Machine to recognize the right character.Form the result of experiment, this algorithm, combined with a segment-extract algorithm proposed by my study team based on angle, can combine segment exactly, And that those feature extracted can describe the character perfectly. And that this method used to recognize similar characters is performed effectively.