A unified probabilistic framework for measuring the intensity of spontaneous facial action units

Automatic facial expression analysis has received great attention in both academia and industry in the past two decades. Facial action coding system, which describes all possible facial expressions based on a set of anatomical facial muscle movements, called Action Unit (AU), is the most popularly used descriptive approach for analyzing facial expressions. In majority of the existing studies in the area of facial expression recognition, the focus has mostly been on facial action unit detection or basic facial expression recognition and there have been very few works on investigating the measuring the intensity of spontaneous facial actions. In addition, these works try to measure the intensity of facial actions statically and individually, ignoring the dependence among AUs, as well as the temporal information, which is crucial for analyzing spontaneous expression. To overcome this problem, this paper proposes a framework based on Dynamic Bayesian Network (DBN) to systematically model such relationships among spontaneous AUs for measuring their intensities. Our experimental results show improvement over image-driven methods alone in AU intensity measurement.