Facial Event Mining Using Active Shape Models and Hidden Markov Models

摘要

Human computer interaction (HCI) has become an active research area that has many potential applications over the last decade. As one form of human face-to-face communication, facial event mining plays an important role in the study of HCI, in particular, for automatic human face analysis in video. This paper proposes a new approach to facial event recognition by means of active shape models (ASMs) and hidden Markov models (HMMs). ASMs are used for continuously tracking facial features and extracting compact shape parameters as pattern features. Each type of facial event is modeled as a HMM for training and recognition. Four basic facial events are investigated in this paper. Preliminary experiments give consistent results and justify the effectiveness of this approach. It also shows the advantage of using dynamic shape features over static features for facial event mining in video.