【作者】 王辉；

【导师】 高隽； 汪荣贵；

【作者基本信息】 合肥工业大学 ， 计算机软件与理论， 2006， 硕士

【摘要】 支持向量机方法由于具有理论完备、全局优化、泛化性能好等特点正在成为人工智能研究领域的研究热点;而核主成分分析方法由于具有特征提取速度快,特征信息保留充分等特点,也被越来越多的研究者所重视,本文将二者相结合应用于人脸识别中,主要的工作内容如下: (1) 整理总结了国内外学术界关于统计学习理论和核主成分分析方面的研究成果,介绍统计学习理论的基本概念和支持向量机的基本原理以及核主成分分析的基本思想; (2) 研究了支持向量机、层次支持向量机、主成分分析以及核主成分分析的基本原理,分析了它们各自的优缺点,并详细阐述了核主成分分析结合层次支持向量机在图像识别中的优势。 (3) 在实验中,利用核主成分分析对人脸进行特征提取,再利用层次支持向量机对其进行识别,在得到较好的识别效果的同时,减少了人脸训练识别的时间。更多还原

【Abstract】 Because of the theoretical maturity, global optimization, excellent generalization, the support vector machine method is becoming the hot spot in the field of artificial intelligence. Kernel Principal Component Analysis method have the characteristics of feature extraction rapidity, sufficient feature information reservation, so it is taken more seriously by many researchers. This article combines the two methods in the field of face recognition. The main work of the thesis is summarized as follows:(1) Reorganized and summarized the domestic and foreign academic circles about the research results of the statistical learning theory and the kernel principal component analysis. Introduced the basic concept of statistics learning theory and the basic principle of support vector machine as well as the kernel principal component analysis basic thought.(2) Studied the basic principle of SVM, Layer-SVM, the principal component analysis as well as the kernel principal component analysis, and analyzed their good and bad points respectively, finally unified the multi-classification method of Layer-SVM and the feature extraction method of kernel principal component analysis and applied it in the experiment.(3) This thesis used the kernel principal component analysis to extract face image feature, then carries on the recognition using support vector machine to it in the experiment. Experimental results demonstrated that this method can obtain good recognition effect, also reduce the training time of human face recognition.更多还原