Online Facial Expression Recognition Based on Finite Beta-Liouville Mixture Models

Facial expressions play a significant role in human communication and their automatic recognition has several applications especially in human-computer interaction. In this paper, we propose an approach for recognizing human facial expressions based on online variational learning of finite Beta-Liouville mixture models. Under the proposed framework, all the involved model parameters and the model complexity of the Beta-Liouville mixture model can be estimated simultaneously, in a closed-form, by avoiding over and under-fitting problems. The performance of the proposed method is evaluated through extensive empirical results and simulations on both artificial data and real facial expressions data sets.