基于LMP和KPCA的人脸识别

提出一种基于拉普拉斯矩阵投影变换和核主成分分析的人脸图像识别方法。对人脸图像做拉普拉斯矩阵变换，通过核主成分分析提取特征，再利用最近邻分类器进行分类。拉普拉斯矩阵变换在保持人脸图像的局部特性的前提下，有效地降低了图像维数。在ORL数据库上的实验表明，进行拉普拉斯矩阵变换后人脸识别精度相差不大，但计算量得到减少。

Face Recognition Based on Laplacian Matrix Projection and Kernel Principal Components Analysis

This paper proposes a face recognition method based on Laplacian Matrix Projection(LMP) and Kernel Principal Components Analysis(KPCA). After the Laplacian matrix transformation being used on graphs of human faces, KPCA method is used for feature extraction. Nearest Neighbor(NN) is used to construct classifier and classificate graphs of face. Smoothness features of faces are obtained and the dimensions of the pictures are reduced effectively after the Laplacian matrix transformation. Numerical experimental results in ORL database show that this face recognition method has a faster operational efficiency than the simple KPCA method while there is little difference between the recognition rates.