3D face recognition using closest point coordinates and spherical vector norms

In this paper, we introduce a new feature named spherical vector norms for 3D face recognition. The proposed feature is efficient, insensitive to facial expression and contains discriminatory information of 3D face. The feature extraction method is firstly finding a set of the points with the closest distance to the standard face, denoted as closest point coordinates, and then extracting the spherical vector norms of these points. This paper combines point coordinates and spherical vector norms for improving recognition. Finally this approach is finished by Linear Discriminant Analysis (LDA) and Nearest Neighbor classifier. We have performed different experiments on the Face Recognition Grand Challenge database. It achieves the verification rate of 97.11% on All vs. All experiment at 0.1% FAR and 96.64% verification rate on Neutral vs. Expression experiment.