Dynamic face recognition system in recognizing facial expressions for service robotics

Facial expression recognition is an essential issue in the field of human and computer interactions. The objective of this paper is to deal with the problem of failure for the feature points on face based on active appearance model (AAM) in fitting with different users. We train the fitting model for different users before recognition. A high-quality AAM alignment results rely on appropriate selections of initial positions. In our method, we apply partial AAM fitting on mouth and eyes. Therefore we could have a better alignment for facial features; as a result, it can implement for tracking real-world video and perform real-time alignment. To obtain stable partial AAM, we use multi-level optical flow to determine initial positions of facial feature models. Then we use the Eigen-faces dynamic face recognition system to recognize different users and select the trained fitting model in recognizing the facial expressions, we combine the methodology of facial expression recognition with Eigen-faces dynamic face recognition. In this paper, we have demonstrated our approach which can accurately recognize different user's facial expressions and emotional variation.