Multi patches 3D facial representation for Person Authentication using AdaBoost

Lahoucine Ballihi (1, 2) , Boulbaba Ben Amor (1, 3) , Mohamed Daoudi (1, 3) , Anuj Srivastava (4)

1 LIFL - FOX MIIRE
2 GSCM-LRIT - Laboratoire de Recherche en Informatique et Télécommunications [Rabat]
3 Institut TELECOM/TELECOM Lille1
4 Department of Statistics [Tallahassee, FL]

In this paper, we investigate the(problem of 3D face authentification) use of AdaBoost-based learning lgorithm which select the relevent curves in the face.We use the shape of 3D curve for face authentication. The basic idea is the analysis of the shape of local facial set of cuves. The set of curves are extracted using level curves based representation centered on facial feature point landmarks. We apply our framework to study and compare this set through the computation of the geodesic distance between corresponding curves on different 3D facial models. AdaBoost considers each curve as a weak classifier and selects iteratively relevant curves and show that there is significant improvement in results. The effectiveness of this technique is evaluated on a subset taken from BU-3DFE(3D Facial Expression Database - Binghamton University) database. The proposed approach increase authentication performances using AdaBoost classifier compared to a simple fusion of scores from all curves.

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Lahoucine Ballihi, Boulbaba Ben Amor, Mohamed Daoudi, Anuj Srivastava. Multi patches 3D facial representation for Person Authentication using AdaBoost. I/V Communications and Mobile Network (ISVC), 2010 5th International Symposium on, Sep 2010, Rabat, Morocco. pp.1-4. ⟨hal-00665904⟩