Bulent Sankur - Bogazici University Multimedia Communications
Date: - Location: Eurecom
Human facial expressions result from combinations of elementary facial deformations, called Facial Actions, like eyebrow rising or lip protrusion. The detection and recognition of these 44 action units is an important step in emotion understanding and classification of facial expressions. We address the facial action detection problem using curvature field of the 3D face data, compare with 2D modality and explore their fusion. Similarly head and face gestures play instrumental roles in conveying communicative and emotional messages. Interpretation of head and face gestures can lead to better human-computer interfaces. We classify gestures by using facial landmark trajectory information either via sequence analysis like HMM or via a spatio-temporal subspace analysis.
