基于级联分类器的手写体数字识别研究

Research on Handwritten Numeral Recognition based on Cascade classifier

以手写体数字的识别为研究中心,研究了针对手写体数字的特征提取和特征识别。利用像素点统计特征,图像结构交截特征,小波分析特征提取出了多维特征。介绍了级联分类的思想,引入拒识率的概念。提出了基于改进BP神经网络和支持向量机级联的手写体数字识别系统。实验验证,该方法能有效提高字符分类的识别率。

Taking handwritten numeral recognition as the research center, the feature extraction and feature recognition of handwritten numerals are studied. Based on the statistical characteristics of pixels, the cross section characteristics of image structure, the feature of wavelet analysis the feature is extracted. This paper introduces the idea of cascade classification, and introduces the concept of rejection rate. A handwritten numeral recognition system based on improved BP neural network and support vector machine cascade classifiers is proposed. Experimental results show that the proposed method can effectively improve the recognition rate of the character classification.