User action and facial expression recognition for error detection system in an ambient assisted environment

Emotion recognition through facial expressions represents a relevant way to understand and even predict the human behavior. Thus, it has been used in various fields such as human-robot interaction and ambient assistance. Nevertheless, it remains a challenging task since expressed emotions might be affected by different parameters such as ethnic origins, age and so on. In this paper, we introduce an efficient facial expression recognition approach based on a Convolutional Neural Network architecture. Carried experimentation on five benchmark facial expression datasets confirms the efficiency of the proposed approach with recognition rates higher than 95%. In the context of ambient assistance, we introduced an error detection module using a user action recognition from Radio Frequency IDentification tags placed on various objects of daily living. The experiments performed in a smart environment show a consistent improvement of the error detection module when including facial expression recognition. Indeed, the false positive detection rate is significantly reduced by over 20%.