3D facial expression recognition using nonrigid CPD registration method

Abstract

In this paper we present a novel approach for 3D facial expression recognition based on a registration method. The used registration method, called the Coherent Point Drift (CPD), is applied to estimate complex non-linear and nonrigid transformation between 3D facial surfaces. The computed transformation allows to recover shape deformations that are induced by facial expression variations. Machine learning is applied using Dimensionality reduction methods in order to promote the computational efficiency and Support Vector Machine (SVM) for classification. The obtained experimental results show that our method achieves promising recognition rates on Bhosphorus database.