Spatio-temporal Facial Features for HRI Scenarios

Human faces are mostly seen in actions conveying various information in our daily life communication. On the other hand, human brains are capable of extracting this essential information in a very short interval of time resulting in a better interaction with others. For example, in interactive scenarios where human beings are assisted by intelligent systems or robots, it is quite useful to extract sufficient information about the interacting person. In this paper we study biometric and soft-biometric traits of the humans by using their face information using a single feature set which is representative of person's identity, gender and facial behavior. This problem is addressed using spatio-temporal multifeatures (STMF) extracted from image sequences using a 3D face model. Further, this feature set provides robustness against varying head poses and facial expressions. Experiments have been performed on laboratory captured images and three benchmark databases under varying poses and facial expressions. The results have been discussed comparatively with different approaches.