基于非负矩阵分解（NMF）的人脸特征提取算法，NMF基本思想是找到一个线性子空间W，使的构成子空间的基本图像的像素点都是正值，而且人脸图像在子空间上的投影系数也是正数-Non-negative Matrix Factorization (NMF) of facial feature extraction algorithm, NMF basic idea is to find a linear sub-space W, so that the composition of sub-space of the basic image pixels are positive, and face image in the sub-space projection coefficient is positive