Collaborative expression representation using peak expression and intra class variation face images for practical subject-independent emotion recognition in videos

This paper proposes a facial expression recognition (FER) method in videos. The proposed method automatically selects the peak expression face from a video sequence using closeness of the face to the neutral expression. The severely non-frontal faces and poorly aligned faces are discarded in advance to eliminate their negative effects on the peak expression face selection and FER. To reduce the effect of the facial identity in the feature extraction, we compute difference information between the peak expression face and its intra class variation (ICV) face. An ICV face is generated by combining the training faces of an expression class and looks similar to the peak expression face in identity. Because the difference information is defined as the distances of locally pooled texture features between the two faces, the feature extraction is robust to face rotation and mis-alignment. Results show that the proposed method is practical with videos containing spontaneous facial expressions and pose variations.