A multi-task model for simultaneous face identification and facial expression recognition

Regarded as two independent tasks, both face identification and facial expression recognition perform poorly given small size training sets. To address this problem, we propose a multi-task facial inference model (MT-FIM) for simultaneous face identification and facial expression recognition. In particular, face identification and facial expression recognition are learnt simultaneously by extracting and utilizing appropriate shared information across them in the framework of multi-task learning, in which the shared information refers to the parameter controlling the sparsity. MT-FIM simultaneously minimizes the within-class scatter and maximizes the distance between different classes to enable the robust performance of each individual task. We conduct comprehensive experiments on three face image databases. The experimental results show that our algorithm outperforms the state-of-the-art algorithms.