Face recognition using semi-supervised sparse discriminant neighborhood preserving embedding

Abstract

A semi-supervised dimensionality reduction algorithm called semi-supervised sparse discriminant neighborhood preserving embedding (SSDNPE) is proposed, which fuse supervised linear discriminant information and unsupervised information of sparse reconstruction and neighborhood with the way of trade-off parameter, inheriting advantages of sparsity preserving projections (SPP), linear discriminative analysis (LDA) and neighborhood preserving embedding (NPE). Experiments operated on Yale, UMIST and AR face dataset show the algorithm is more efficient. © 2013 Springer-Verlag.