The Study Of Handwritten Chinese Character Recognition

The study of Chinese character recognition is regarded as not only a important theory meaning and practice value direction in pattern recognition field, but a final goal to the research of character recognition. Because of many complex structures of Chines characters and diverse distortion of handwritten Chinese characters, Off-line handwritten Chinese character recognition is the most difficult issue in Chinese character recognition field. At present, Off-line handwritten Chinese character recognition is being in a distinct stage of development.Based on the most difficult factor of Off-line handwritten Chinese character recognition, this thesis used mathematical morphology and wavelet transformation to handwritten Chinese character recognition. From simulating the human vision character, machine vision character is proposed.The primary contents of this thesis are:1) Handwritten Chines character image pretreatment. Based on six specific marginal points and mathematical morphology, stroke width is extracted and normalized. Distortion of handwritten Chinese characters is restrained effectively.2) Stroke extraction of handwritten Chinese characters. The method of stoke extraction is proposed by mathematical morphology. The method of eight adjoining points is used for the stroke amended.3) Reform of handwritten Chinese characters. A method of handwritten Chinese characters reform is advanced. Four directional strokes are added one by one, and it is obtained that a Chinese character image is similar to printing Chinese character. Furthermore, basal structure information of original handwritten Chinese character is reserved by the method of eight adjoining points.4) Recognition algorithm of handwritten Chinese characters. Based on stroke extracted and wavelet transformation, recognition algorithm is proposed for handwritten Chinese characters and printing Chinese character. Stroke distortion can be amended distinctly by this method. Recognition rate is improved.The experiment indicated that original handwritten Chinese character information is reserved and stroke distortion is restrained effectively by stroke width normalization and stroke reform. Handwritten Chinese characters can be recognized well and truly according to the methods in this paper.