ORTHOGONAL APPROXIMATELY HARMONIC PROJECTION FOR FACE RECOGNITION
Xia Sun, Ziqiang Wang, Lijun Sun · 2012
Face recognition has attracted growing attention for applications such as identity authentication and human-computer interface. However, a major challenge of face recognition is that the captured face image often lies in a high-dimensional feature space. To overcome the curse of dimensionality problem and improve the performance of face recognition, a novel manifold learning algorithm called orthogonal approximately harmonic projection (OMMP) is proposed in this paper. The OAHP algorithm is based on the harmonic projection (AHP) and explicitly considers the local geometrical structure and cluster structure of the face space. Meanwhile, the OAHP method can produce orthogonal basis vectors to preserve the metric structure of face space, which greatly enhances the discriminating power of the reduced lower-dimensional feature space. Experimental results on three face databases show that the proposed OAHP performs much better than related algorithms in terms of recognition rate.