Constructing Structures of Facial Identities on the View Sphere Using Kernel Discriminant Analysis

Yongmin Li, Shaogang Gong and Heather Liddell

University of London

We present a novel approach to face recognition by constructing facial identity structures across views and over time, referred to as identity surfaces, in a Kernel Discriminant Analysis (KDA) feature space. This approach is aimed at addressing three challenging problems in face recognition: extracting the non-linear discriminant features, recognising faces across multiple views, and recognising moving faces over time. First, the KDA is developed to compute the most significant non-linear basis vectors with the intention of maximising the between-class variance and minimising the within-class variance. We applied KDA to the problem of multi-view face recognition, and a significant improvement has been achieved in robustness and accuracy. Second, identity surfaces are constructed to model the variance of facial appearance caused from rotation in depth. Recognition can then be conveniently performed by computing the pattern distances from the identity surfaces. Third, video-based online face recognition is performed by computing and matching object onto identity surfaces which encode the spatio-temporal dynamics of moving faces.

pdf file