一种基于核局部保留投影的人脸识别方法

为了有效提取人脸的非线性结构信息，提出一种新的基于最大散度差的核判别局部保留投影方法.首先通过核函数将样本数据映射到高维特征空间，计算特征空间中样本的散度矩阵，其次将样本原始空间中的近邻图嵌入到散度矩阵，最后采用最大散度差准则进行特征提取.在PIE与Yale人脸数据库上的实验结果表明，提出的人脸识别方法最高识别率可达到99%. 

A Face Recognition Method Based on Kernel Locality Preserving Projection

To effectively extract non-linear structure information of face images, a novel method based on maximum scatter difference rule was proposed in this paper. Firstly, the kernel function was used to map original sample data into high dimensional feature space and the scatter matrices were obtained. Secondly, the nearest graphs of original sample data were embedded into scatter matrices; then the maximum scatter difference was used to extract the feature of samples. Experimental results on Pie and Yale face databases demonstrate that the proposed face recognition method is efficient and its top recognition rate can reach 99%.