Key points are not available for this paper at this time.
In this paper, we present results on experiments employing active appearance model (AAM) derived facial representations, for the task of facial action recognition. Experimental results demonstrate the benefit of AAM-derived representations on a spontaneous AU database containing "real-world" variation. Additionally, we explore a number of normalization methods for these representations which increase facial action recognition performance.
Lucey et al. (Fri,) studied this question.