基于“频谱脸”和SVM的人脸识别方法  

作　　者：杨绍华 

机构地区：[1]宁夏大学信息工程学院,宁夏银川750021

出　　处：《科学技术创新》2017年第30期12-14,共3页Scientific and Technological Innovation

基　　金：宁夏自然科学基金资助项目(项目编号:NZ14052)

摘　　要：针对传统PCA的人脸识别方法在向量化人脸过程中丢失结构特征的问题,提出了一种基于频谱脸和支持向量机的人脸识别方法。首先选取人脸图像经小波分解后的低频分量作为人脸特征的描述,然后用SVM和最近邻法构建复合分类器进行分类,在ORL数据集上的实验结果表明,本文方法取得了比传统方法更好的识别性能。Aimed at the disadvantages occurred losing structure feature among facial image vectorization procedure. The algo- rithm of face recognition based on Wavelet Transform and Support Vector Machine was proposed in this papper. At first Wavelet Transformation is used to decompose face images and for choosing the LL coefficients, then SVM and the classical Nearest Neighbor method are performed on the reductions for the pattern classification. The experiment based on ORL-DATABASE show that above-mentioned＇ algorithm get better performance than traditional PCA method.

关 键 词：人脸识别 PCA 频谱脸 SVM 最近邻 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]