Facial Action Recognition Using Multikernel Learning to Combine Heterogeneous Features
Keywords:Active appearance model (AAM), facial action unit (AU), facial expression recognition and analysis (FERA), local Gabor binary pattern (LGBP), multikernel learning.
Facial action recognition is a challenging problem because facial expressions can be ambiguous and appear differently on different individuals. In this paper, we propose a novel approach based on multikernel learning to identify facial actions. The approach is a combination of heterogeneous features derived from local binary pattern (LBP), histograms of oriented gradients (HOG), and deep convolutional neural networks (CNNs) features. The proposed method is evaluated on the Cohn-Kanade (CK+) and BU3DFE datasets. Our experimental results demonstrate that the proposed method substantially outperforms the state-of-the-art single and multi-kernel approaches with respect to accuracy. The experiments also show that the proposed method can exploit the complementarity of different features, resulting in improved recognition performance. This is evidenced by the significant gains over the single-kernel approaches.