Abstract: Hangul is a frequently-used oriental character. Compared with English and Chinese, the Hangul has some characters. For example, large vocabulary, the similarity between class and each basic strokes are arranged in two dimensions space. Just as these special points, the recognition of online Hangul is a difficult point to this day. A method is proposed based on PCGM to recognize Jamo, and these models are used to segment Hangul into Jamos and these Jamos are used to retrain PCGM. The result of the experiment shows these methods have an outstanding effect on Hangul recognition.

Key words: Precision Constrained Gaussian Models（PCGM）, Hangul Recognition, Hangul segment

摘要： 韩文是一种常见的东方文字。相对于英文和汉字，韩文字具有大类别，类与类之间相似度极高，基本笔画单位具有以二维几何方式进行排列等特点，因此联机手写体韩文字符识别一直是一个难点。提出了基于PCGM模型的韩文字母识别，用字母的PCGM模型从韩文字分割出来，将分割出来的字母对此模型进行训练，使模型收敛稳定。实验结果表明方法对韩文字的识别有显著效果。

关键词: 精确限制高斯模型（PCGM）, 韩文字识别, 韩文字分割