限定性手写汉字识别的一种可变形弹性匹配模型

Constrainted Handwritten Chinese Character Recognition Using a Deformable Elastic Matching Model

本文提出了一种应用于手写体汉字识别的可变形弹性匹配模型．在该模型中，我们假定：对于同一种汉字的不同手写样本，尽管由于书写风格的不同而使其形状各异，但它们却具有相同的空间拓扑结构，我们将用一组笔段置移矢量来描述不同的手写体之间的细节形状上的差异．在置移矢量场的作用下，一个可变形的模板汉字将逐步变形以趋近于输入的汉字，从而在两汉字之中寻求一种最佳的匹配．

This paper describes a deformable elastic matching approach to Handwritten Chinese Character Recognition (HCCR). We assume that different handwriting variations share thesame topological structure,but may differ in shape details. The variations between differnet handwriting characters are modeled by a set of stroke displacement vectors (SDV). According to theSDV derived,a model character is deformed gradually,in an effort to transform itself much closerto an input character. Experiments show that the proposed elastic matching model can efficientlydeal with local shape changes and variations between characters.