基于因子分析的计算机打印文件鉴定  

Factor Analysis based Print Document Identification by computer

作　　者：赵克坚 廖海斌 ZHAO Kejian;LIAO Haibin(Computer Center,Xianning Central Hospital(The First Affiliated Hospital of Hubei University of Science and Technology,Xianning 437100;School of Computer Science and Technology,Hubei University of Science and Technology,Xianning 437100)

机构地区：[1]咸宁市中心医院(湖北科技学院附属第一医院)计算机中心,咸宁437100 [2]湖北科技学院计算机科学与技术学院,咸宁437100

出　　处：《计算机与数字工程》2019年第6期1555-1559,1564,共6页Computer & Digital Engineering

基　　金：无锡市科技局项目“计算机智能技术的打印文件检验系统”(编号:CSE00802);湖北省教育厅科学技术研究项目“面向视频监控的人脸多模态识别与3D建模研究”(编号:Q20172805);湖北省教育科学规划项目“基于智能教室与大数据分析的新型教学质量评估体系研究”(编号:2016GB086)资助

摘　　要：在计算机打印文件鉴定问题中,当样本和检材文件中没有相同字或相同字符极少时鉴定极其困难。针对这一问题,提出了一种基于因子分析模型的打印文件鉴定方法。首先对特征矩阵采用双线性模型进行因子分析,分离出文本因子,提取近似文本无关的打印机特征,然后采用贝叶斯决策和期望最大化算法进行分类,实验结果表明论文方法有效改善了识别率。In order to solve the problem of print document identification when the characters in samples and test material are completely different,a new method is proposed based on factor analysis. Firstly the feature matrix is analyzed using bilinear model and the printer factor is separated from the text factor. Then the new printer feature nearly independent on text is extracted. The Bayesian decision and Expectation-Maximization algorithm are used to classify the new feature matrix. The experimental results demonstrate that this method promotes the identification accuracy significantly.

关 键 词：打印文件鉴定 因子分析 双线性模型 EM算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]