Face recognition based on discriminant vector angle embedding
Zhu Shan-an · Journal of Zhejiang University(Engineering Science) · 2008
A discriminant vector angle embedding(DVAE) method was presented for dimension reduction based face recognition by constructing neighborhood graphs.A graph including both positive edges and negative edges is constructed.A positive edge is put on two samples in the same class and a negative edge on two different class samples within k nearest neighbor.The measure in DVAE is the angle between two vectors instead of modulus in traditional methods,which not only exempts the estimation of the parameter t in heat weight function,but also reduces the influence of luminance difference between image samples on face recognition.When a test sample is embedded into the low-dimensional space with preserving the neighborhood vector angle,a classification called angle nearest neighbor is used for face recognition.Experiments on Yale and UMIST databases demonstrated that the proposed approach is superior to other methods in terms of recognition accuracy.