Abstract: The paper proposes a method for constructing multi-template dictionary based on feature vectors distribution. Then a self-learning strategy is implemented to improve the generalization of the dictionary. The effectiveness of these algorithms is shown by experimental results.

摘要： 针对手写体汉字字形变化复杂的特点,提出了一种基于特征向量分布的多模板字典制作方法,并通过赋予该多模板字典自学习功能,来提高其泛化能力。实验证明,用这种方法制作的字典与用传统方法制作的同类字典相比,能够有效地提高系统的总体识别率。

关键词: 手写体汉字识别, 多模板字典, 特征区

CLC Number: 

• TP391.4