Facial Emotion Recognition Using PHOG and a Hierarchical Expression Model

In this paper, we propose a fast approach to detecting human facial emotions, using a hierarchical multiple stage scheme and only the PHOG feature descriptors basing on frontal images of human faces. In this model, the facial expression is the composition of a set of expressive facial regions which can be evaluated with the trained emotional templates. Within this framework, the proposed algorithm is able to achieve acceptable detection accuracy for Cohn-Kanade dataset, with less time and space complexities compared with the approaches in other research literature, making it applicable to low cost hardware such as mobile device. In addition, the experiments illustrated that the approach presented in this paper has good robustness and extendibility.