In this paper, we present the generalization of the automatic face recognition system making it able to deal with different expressions and the presence of spectacles. The system behavior is measured on the FRGC 1.0 and the XM2VTS databases, addressing some of the current challenges highlighted by Phillips et al. (2005). We set up experiments having in the gallery, for each subject, both one still image and multi-still images with different expressions. Besides, we investigate how the image resolution affects the performances and test the scalability of the system doubling the gallery size