Facial expression recognition using facial-componentbased bag of words and PHOG descriptors

Facial expression recognition has many potential applications in areas such as human-computer interaction (HCI), emotion analysis, and synthetic face animation. This paper proposes a novel framework of facial appearance and shape information extraction for facial expression recognition. For appearance information extraction, a facial-componentbased bag of words method is presented. We segment face images into four component regions: forehead, eye-eyebrow, nose, and mouth. We then partition them into 4 × 4 sub-regions. Dense SIFT (scale-invariant feature transform) features are calculated over the sub-regions and vector quantized into 4 × 4 sets of codeword distributions. For shape information extraction, PHOG (pyramid histogram of orientated gradient) descriptors are computed on the four facial component regions to obtain the spatial distribution of edges. Multi-class SVM classifiers are applied to classify the six basic facial expressions using the facial-component-based bag of words and PHOG descriptors respectively. Then the appearance and shape information is fused at decision level to further improve the recognition rate. Our framework provides holistic characteristics for the local texture and shape features by enhancing the structure-based spatial information, and makes it possible to use the bag of words method and the local descriptors in facial expression recognition for the first time. The recognition rate achieved by the fusion of appearance and shape features at decision level using the Cohn-Kanade database is 96.33%, which outperforms the state-of-the-art research works.