Facial expressions based error detection for smart environment using deep learning

Creating innovative applications of ambient intelligence for smart environments became a real challenge. Indeed, the existing systems are considered as being obtrusive. Therefore, the ongoing trends consist in finding new methods which are less intrusive while being accurate and computer efficient. To that end, we introduce an ingenious way to assist the elderly and people with cognitive impairment in their daily life using emotions through facial expressions. Our hypothesis assumes the correlation between expressed emotions and the resulting errors in daily activities. In order to verify it, we introduce an experimentation protocol with neurotypical subjects. In this paper, we mainly focus on designing an approach for facial expression recognition using Convolutional Neural Network. The carried experiments with benchmark datasets provided promising results. In fact, the proposed approach outperforms the existing methods in terms of accuracy. Moreover, our approach has been implemented in an embedded system with limited resources. The main objective of this paper is to provide an effective approach recognizing facial expressions that will be used for our further experimentations.