The traditional PCA method chooses a dimensionality reducing linear projection that maximizes the scattering of all projected samples.

• 传统的特征脸方法是根据PCA投影变换后，利用特征空间的总体散布矩阵最大化的特点进行分类；