Face Recognition using Modular Bilinear Discriminant Analysis

M. Visani, C. Garcia, JM. Jolion, France Telecom R&D & LIRIS, 2005

We present a new approach for face recogni-tion, named Modular Bilinear Discriminant Analysis (MBDA). In a firststep, a set of experts is created, each one being trained independentlyon specific face regions using a new supervised technique named Bili-ear Discriminant Analysis (BDA). BDA relies on the maximization of ageneralized Fisher criterion based on bilinear projections of face imagematrices. In a second step, the experts are combined to assign an identitywith a confidence measure to each of the query faces. A series of experi-ments is performed in order to evaluate and compare the effectiveness ofMBDA with respect to BDA and to the Modular Eigenspaces method.The experimental results indicate that MBDA is more effective than both BDA and the Modular Eigenspaces approach for face recognition.