Basic face animation mode extraction by ICA

We extract the 3D positions of feature point locations for a specific person with different expressions by a method based on the 3D scan data. By using the independent component analysis for captured location coordinates, a vector space representation is formed and this space is called the basic face animation mode. Finally we propose a new parameter space for creating facial expressions of specific faces. The experimental results show such a representation affects correlation and independence of every part of the animated face so that the realism of face animation with several expressions is improved.