Research On Offline Handwritten Chinese Character Stroke Feature Extraction Based On Fuzzy Partition

Offline handwritten Chinese character recognition is a major research direction ofpattern recognition, feature extraction has a very important role in character recognition.After observation of different handwriting styles of Chinese writing, we combine with theexisting structure of research methods. This paper introduced a method of offlinehandwritten Chinese character stroke features extraction based on fuzzy partition.Firstly, we use the fuzzy run-length algorithm improved from run-length algorithm toextract stroke segments. Discrete points and breakpoints are dealt with after getting strokesegments of four directions: horizontal, vertical, left-fall and right-fall. Secondly, strokefeatures are extracted such as stroke number, length, position, intersecting relation and soon. Based on the fuzzy set, the method introduces fuzzy membership to judge the fuzzyposition of strokes of Chinese character. Finally, with the combination of stroke features ofChinese characters, structure identification and similar characters recognition are impleted.Simple fuzzy rules are designed to identify the structures of Chinese characters. The causesof similar characters are analyzed and similar words are identified by combined structuralfeatures of Chinese characters.Experimental results show that the feature extraction methods used in this paper caneffectively improve the efficiency of offline handwritten Chinese character recognition.