An approach to creating 3D animation of facial expression generation might be use of a motion capture system for dynamic 3D measurement of the target face, in which placing a lot of markers on the facial surface is a troublesome task. Another option is use of the morphable 3D face model which is built through principal component analysis of 3D data representing shape variations caused by facial expression generation. In this work, we adopted the latter approach, in which time series transformations of a novel 3D face is obtained by applying stepwise impression transfer vectors each of which gives displacement in the parametric space of 3D face model. In addition, the impression transfer vectors used in this experiment are designed by Support Vectors obtained by SVM learning.