基于PCA的脱机手写汉字的统计模型及其应用

A PCA BASED STATISTICAL MODEL FOR OFF-LINE HANDWRITTEN CHINESE CHARACTERS AND ITS APPLICATIONS

本文通过分析传统汉字的结构模型所具有的优缺点,提出了建立脱机手写汉字统计模型的理论框架;并利用PCA技术发现大量数据规律性的能力,提出了一种基于PCA技术的脱机手写汉字的统计模型。与传统的结构模型相比,该模型避免了目前还无法解决的准确抽取结构基元的困难,通过以容易抽取的可重构的统计特征作为统计基元,并通过对统计基元变化的整体描述或者说对统计基元相互之间关系的描述,较好地建立了脱机手写汉字的统计模型。根据该模型得到的一些实验结果充分说明了其描述脱机手写汉字的有效性。

In this paper, a framework on building a statistical model for off-line handwritten Chinese characters is first proposed after the advantages and disadvantages of traditional structural models are analyzed. Then a statistical model based on principal component analysis technique is presented. In contrast to conventional structural models, this statistical model need not extract strokes exactly which is usually difficult. Instead, it uses the easily extracted statistical reconstructable features. The validity of this statistical model is demonstrated by the experiment results in this paper.