基于主成分分析法及贝叶斯分类器的手写数字识别  

The identification of Hand-written digits based on Principal Component Analysis and Bayesian classifier

作　　者：尹东霞[1] 

机构地区：[1]山东科技大学网络与信息中心,山东青岛266590

出　　处：《大众科技》2015年第9期39-41,共3页Popular Science & Technology

摘　　要：针对目前手写数字难识别并且识别正确率低这一现象,提出了一套基于主成分分析法及贝叶斯分类器的手写数字识别方法。该方案首先利用主成分分析法减小输入数据的维数,而后把降维的数据作为训练过的贝叶斯分类器的输入,从而得到对于输入的手写数字的识别。在MNIST手写数字数据集上该方法能够达到96.35%的识别率。该仿真结果说明文章提出的手写数字识别策略能够实现对手写数字的高效的识别。To deal with the low discrimination and low accuracy of the hand-written digits,this paper proposed a hand-written digits identification method which is based on the Principal Component Analysis （PCA） and the Bayesian classifier.This scheme employs the PCA to reduces the dimension of the input data,then the dimension-reduced data is regarded as the input for the Bayesian classifier,the result of the classifier is the identified digits.This method achieves the accuracy of 96.35% on the MNIST hand-written digits data set.The case study shows that the hand-written digit identification scheme this paper proposed can identify the hand-written digits effectively.

关 键 词：手写数字识别 主成分分析法 贝叶斯分类器 

分 类 号：TM732[电气工程—电力系统及自动化]