Evaluation of spatio-temporal regional features for 3D face analysis

3D facial representations have been widely used for face recognition and facial expression recognition. Both local and global features can be extracted from either static or dynamic models in both spatial and temporal domains. 3D local features are referred to the features in regional facial areas while 3D global features are referred to the features from the entire facial region. In this paper, we address the issue of performance assessment of facial analysis in terms of global features versus local features, static models versus dynamic models, and spatial domain versus temporal domain. Based on the existing work of using 3D spatiotemporal HMM for facial analysis, we propose to extend it to a local-temporal HMM in order to provide an explicit comparison of global features and local features. A dynamic 3D facial expression database and a static facial expression database are used for experiments. The performance for six prototypic facial expression classification and face identification is analyzed and reported.