Abstract

We present a hierarchical graphical model to probabilistically estimate head pose angles from real-world videos, that leverages the tem- poral pose information over video frames. The proposed model employs a number of complementary facial features, and performs feature level, probabilistic classifier level and temporal level fusion. Extensive experi- ments are performed to analyze the pose estimation performance for dif- ferent combination of features, different levels of the proposed hierarchical model and for different face databases. Experiments show that the pro- posed head pose model improves on the current state-of-the-art for the unconstrained McGillFaces [10] and the constrained CMU Multi-PIE [14] databases, increasing the pose classification accuracy compared to the cur- rent top performing method by 19.38% and 19.89%, respectively.