Subspace methods in object/face recognition

In this paper, several new applications for subspace methods are introduced, including: (1) subspace method used for overcoming generalization/overfitting problem for applications such as face recognition, (2) multiple subspaces constructed from the original subspace to accommodate the inputs distorted by scaling, rotating, translating etc. and (3) a subspace framework based on which the original signal can be transformed into the subspace of a much higher dimension space through a nonlinear mapping. This nonlinear mapping provides an alternative to direct nonlinear classification in the original space, that is to perform a linear classification in this subspace