High Performance Facial Expression Recognition System Using Facial Region Segmentation, Fusion of HOG & LBP Features and Multiclass SVM

Facial expression is an effective method of nonverbal communication and expressing feelings. A method to recognize emotions through facial expressions is proposed in this paper. After some preprocessing of the input image, the facial region is segmented into four expression regions according to the highly effective proposed segmentation method. Features from these segmented parts are extracted using a fusion of Histogram of Oriented Gradients (HOG) and Local Binary Patterns (LBP). Reduction of the dimension of the feature vector is done using Principal Component Analysis (PCA). For classifying the features and thus the expression images, multiclass Support Vector Machine (SVM) is used. The performance of the proposed method is measured using three publicly available and highly used datasets (JAFFE, CK+, RaFD). Finally, achieved performance is compared with performance on these datasets by other available methods to indicate that the proposed method succeeds in achieving state-of-the-art performance.