不规范日语假名的计算机模糊特征识别

The Fuzz Character Identifying of Anomaly Kana Character by Computer

为了克服日语假名的非模糊识别算法和特征抽取识别法出错率高、系统识别率低的缺点,文章提出了模糊特征识别原理,确定了特征抽取的隶属函数,提出了不规则日语假名基元之间的模糊关系的识别方法、并根据基元的相似度、重要度及稳定度进行了匹配决策,最终建立了能够适应性不规则日语假名内部各个基元及其四邻关系、细致描述、刻画日语假名内部结构的位置隶属函数,将不规则日语假名的结构描述简化为日语假名基元模糊关系的集合。

In order to conquer the high ratio of treading awry and low identifying in non-fuzzed anomaly arithmetic and draw-out eharaeter identifying about Kana characters, the principle of fuzzed character anomalying was lodged, and subjection function about character take outing was established, and identifying method of fuzzed relation was lodged for basic element anomalistic Kana charaeters, and matching decision-making was done based on kindred , import and stabilization degree of basic element, and the position subjection was established, finally, the characterization of unequal Kana are predigest to concourse of fuzzed relation in Kana basic element.