摘要： 提出一种运用非负矩阵分解(NMF)分组策略进行人脸识别的方法。将训练图像分组，分别对每组图像作NMF，获取每组图像的基图像构成的非负特征子空间，将训练图像和测试图像分别向各个特征子空间进行投影，将每组图像提取出的特征系数混合，根据最近邻原则进行识别。基于ORL人脸数据库上的实验证明了该方法的有效性。

关键词: 非负矩阵分解, 人脸识别, 基图像

Abstract: A method of applying Non-negative Matrix Factorization(NMF) group strategy for face recognition is put forward. It divides the training images into groups. NMF is applied to the each group’s images to obtain the non-negative subspace constructed by basic image of each group’s images, project the training images and testing images to each feature subspace. The extracted feature coefficents of each group’s images are mixed for recognition based on the nearest neighbor principle. Simulation experiments illustrate the effectivity of the method on the ORL face database.

Key words: Non-negative Matrix Factorization(NMF), face recognition, basic image

中图分类号: 

• TP391