基于语义构件的甲骨文模糊字形的识别方法

Recognition of Fuzzy Inscription Character Based on Component for Bones or Tortoise Shells

针对殷墟甲骨文字整理过程中大量模糊字形难以识别的问题,提出了一种基于语境的统计分析和Hopfield网络相结合的模糊匹配识别方法。该方法利用语境分析生成的候选字库得到对应的甲骨文语义构件向量,然后结合基于Hopfield网络的识别结果计算待识别的甲骨文模糊字的匹配度,根据匹配度确定目标甲骨字。实验表明,该方法对部分甲骨拓片的模糊甲骨字识别效果较好。

In order to solve the problem that it is difficult to recognize fuzzy characters on Oracle Bone Inscription( OBI) of the Shang Dynasty,a fuzzy matching algorithm combining statistical analysis and Hopfield network is proposed.First,according to the OBI context statistics information,a candidate font library for the fuzzy characters is established;second,the initial recognition work is processed based on OBI character Hopfield network;last,the OBI component vector is used to calculate the similarity of fuzzy characters.According to the similarity,it can determine the target OBI character.Experiments show that the method can reach high recognition accuracy for some bones or tortoise shells which include fuzzy characters.