摘要
为提高人脸识别的速率和识别率,文中提出一种基于混合核函数的快速核主成分分析算法用于进行人脸识别,首先构造两种混合核函数,利用均值矢量的方法构建核矩阵,并利用文中提出的核主成分分析算法计算核矩阵的特征向量。分别在ORL和AR人脸数据库中做了相关实验,并且与传统的核主成分分析方法在识别率和算法运行时间上进行了比较,结果表明,文中所提核主成分分析方法具有较高的识别率和更短的运行时间,从而为实时地具有大数据的人脸识别系统提供技术支持。
In order to improve the speed and recognition rate of face recognition,the paper proposes a fast kernel principal component analysis based on the mixed kernel principal component analysis（ FKPCA） algorithm for face recognition. First we construct two kinds of mixed kernel functions,use the mean vector to build kernel matrix,and then use the proposed kernel principal component analysis algorithm to compute the eigenvectors of the kernel matrix.A large number of experiments are performed on ORL and AR face database,respectively,and a comparison is made between the traditional methods of kernel principal component analysis in their recognition rates and running time.The results show that the proposed kernel principal component analysis has higher recognition rate and shorter operational time,which provides supports for real-time face recognition system in the big data era.
出处
《电子科技》
2015年第2期46-50,共5页
Electronic Science and Technology
关键词
人脸识别
特征提取
主成分分析
核主成分分析
核函数
face recognition
feature extraction
principal component analysis
kernel principal component analysis
kernel function
