Learning sparse representations by K-SVD for facial expression classification
Zichen Wang, Ruojing Jiang, Xiaofei Jiang, Tong Zhou · 2015
Facial expression classification is a very challenging problem in machine perception, which plays an important role in many visual applications. In this paper, we use a machine-learning-based framework to address this problem. We firstly apply K-SVD to learn sparse representations of the face images in the training set in an unsupervised manner for image modeling. After that we train a SVM classifier in a supervised manner on those representations, and then the obtained classifier would be used for facial expression classification. Our experimental results show that the learned dictionaries by K-SVD can not only capture meaningful features from the faces for facial expression modeling, but also help to boost the performance of the subsequent SVM classifier in terms of classification accuracies and speeds.