融合奇异值分解和最大间距准则的人脸识别方法

提出了一种融合奇异值分解(SVD)和最大间距准则鉴别分析(MMC)的人脸识别方法。对人脸图像进行奇异值分解,选取较大的一组奇异值构成特征向量,对所有训练样本按照最大间距准则鉴别分析算法计算投影矩阵,把人脸图像矩阵在投影矩阵上投影得到特征矩阵。融合决策阶段,在以上两类特征集中,分别计算待识别样本到所有训练样本的欧氏距离并对得到的两类结果进行加权融合,最后根据最近距离分类器分类。基于ORL人脸数据库上的实验结果表明算法的有效性。

Face recognition approach based on fusion of singular value decomposition and maximum margin criterion

A face recognition method based on singular value decomposition and maximum margin criterion discriminant analysis are proposed.Singular value decomposition is performed on face image matrix,and some bigger singular values are selected to form feature vector.According to maximum margin criterion,MMC feature matrices of training samples and testing samples are calculated.During fusion period,using Euclidean distance measure,the distance of testing sample to all training samples is calculated in both SVD feature set and MMC feature set respectively.Weighted fusion operation is performed on these results that have been gotten above.The nearest distance classification is used to distinguish each testing face sample.The experimental results on ORL face database indicate that the approach is efficient.