Facial Expression Recognition Using Histogram of Oriented Gradients with SVM-RFE Selected Features

This study is an attempt towards improving the accuracy and execution time of a facial expression recognition (FER) system. The algorithmic pipeline consists of a face detector block, followed by a facial alignment and registration, feature extraction, feature selection, and classification blocks. The proposed method utilizes histograms of oriented gradients (HOG) descriptor to extract features from expressive facial images. Support vector machine recursive feature elimination (SVM-RFE), a powerful feature selection algorithm is applied to select the most discriminant features from high-dimensional feature space. Finally, the selected features were fed to a support vector machine (SVM) classifier to determine the underlying emotions from expressive facial images. Performance of the proposed approach is validated on three publicly available FER databases namely CK+, JAFFE, and RFD using different performance metrics like recognition accuracy, precision, recall, and F1-Score. The experimental results demonstrated the effectiveness of the proposed approach in terms of both recognition accuracy and execution time.