Facial Expression Recognition Based on Significant Face Components Using Steerable Pyramid Transform

Facial expression recognition is a challenging problem in many areas such as computer vision and humancomputer interaction. To extract an effective facial features and then to classify them are the best important points of facial expression recognition process. In this article, a new automatic facial expression recognition algorithm is proposed in order to further enhance the recognition performance in terms of these two points. First, it is detected the some specific components of face, such as the mouth, eyes, eyebrows, and nose by Viola-Jones algorithm. Secondly it is extracted features by applying local Steerable Pyramid Transform (SPT) to each of facial component images. Thirdly it is used Support Vector Machines (SVM) classifiers for facial expression verification. Finally the classifier outputs are combined by decision fusion. The experiments on the Japanese Female Facial Expression (JAFFE) database and the Cohn-Kanade database show that the proposed Component based Facial Expression Recognition (CFER) algorithm improves facial expression recognition performance compared to an algorithm combining SPT and Principal Component Analysis (PCA) using whole face images to the results in the literature.