Facial expressions are a set of symbols of great importance for human-to-human communication. Spontaneous in their nature, diverse and personal, facial expressions demand for real-time, complex, robust and adaptable facial expression recognition (FER) systems to facilitate the human-computer interaction. The last years' research efforts in the recognition of facial expressions are preparing FER systems to step into the real-life. In order to meet the before-mentioned requirements, this article surveys the work in FER since 2008, particularly adopting the discrete states emotion model in a quest for the most valuable FER work/systems. We first present the new spontaneous facial expression databases and then organize the real-time FER solutions grouped by spontaneous and posed facial expression databases. Then automatic FERs are compared and the cross-database validation method is presented. Finally, we outline FER system open issues to meet real-life challenges.