Abstract: This paper describes and analyzes the feathers used in handwritten digits recognition, including template match, statistical feathers, structural feathers and transformation feathers. Distance-based classifiers and neural network classifiers are also presented. It also proposes hybrid approach in combining different feathers and different classifiers. The new machine learning method——support vector Machine is introduced in the end.

摘要： 手写体数字识别特征提取方面，有模板匹配，统计特征和结构特征，在分类器设计上有基于距离的分类器和神经网络分类器等。分析和评价了这些问题后，指出今后的研究方向应在特征综合、分类器集成以及新的分类器的研究上。

关键词: 手写体数字识别, 特征提取, 分类器, 分类器集成, 支持向量机

CLC Number: 

• TP391.41