Fusion of PHOG and Gabor for facial expression recognition

Facial expression is the real expression of people's inner emotions. Research on facial expression recognition can make machines better understand human emotional information, among which feature extraction is particularly important. This paper proposes a facial expression feature extraction method based on PHOG (Pyramid Histogram of Oriented Gradients) and Gabor. PHOG can better represent the gradient information and edge direction gradient information of facial expression images. In the frequency domain, Gabor uses the multi-scale filter to extract texture features of images and filter out irrelevant face information. In this paper, we first perform principal component analysis (PCA) on the two feature operators, then fuse the two feature vectors, and finally realize facial expression recognition in static images by a support vector machine (SVM).