基于候选特征笔画和多类阈值的手写汉字切分

通过分析汉字的常见结构,鉴于汉字与汉字之间的距离和构成汉字的部件之间的距离的显著差异性,提出一种基于候选特征笔画和多类阈值的手写汉字切分方法.首先从构成手写汉字的笔画集合中提取候选特征笔画,根据候选特征笔画将手写汉字预切分,然后利用基于间距阈值的部件组合规则对过切分的汉字部件进行组合,最后采用基于单字宽度阈值的粘连汉字判断规则搜索粘连汉字,对粘连汉字进行递归切分.实验表明,该方法对连续手写汉字的切分准确率较高,具有一定的实用性.

Segmentation of Handwritten Chinese Characters Based on the Candidate Characteristics Strokes and Multi-Class Threshold

With the analysis of the common structures of Chinese characters, a handwritten Chinese characters segmentation method which based on the candidate characteristics strokes and multi-class threshold is proposed in view of the significant difference between the distance between Chinese characters and the distance between the components that constitute a Chinese character. Firstly, the candidate characteristics strokes are extracted from the collection of strokes that constitute handwritten Chinese characters, the handwritten Chinese characters are pre-segmented according to the candidate characteristics strokes. Then, the over-segmentation components of Chinese characters are combined by using a components combination rule based on the spacing threshold. Finally, the adhesions Chinese characters are searched by using an adhesions Chinese character judgment rule based on the single character width threshold and recursive segmented. The experiments show that the segmentation accuracy of continuous handwritten Chinese characters is quite high by this method, having certain practical.