3-D Facial Expression Representation using B-spline Statistical Shape Model

Effective representation and recognition of human faces are essential in a number of applications including human-computer interaction (HCI), biometrics or video conferencing. This paper presents initial results obtained for a novel method of 3-D facial expressions representation based on the shape space vector of the statistical shape model. The statistical shape model is constructed based on the control points of the B-spline surfaces of the training data set. The model fitting for the data is achieved by a modified iterative closest point (ICP) method with the surface deformations restricted to the estimated shape space. The proposed method is fully automated and tested on the synthetic 3-D facial data with various facial expressions. Experimental results show that the proposed 3-D facial expression representation can be potentially used for practical applications.