A real-time emotion recognition system for disabled persons

In order to assure a safe navigation for an electric wheelchair user, both environment and user must be kept under surveillance for any potential endangering act, whether they are intentional, or unintentional. This paper proposes a real-time embedded emotion recognition system designed for an electric wheelchair to detect, exploit and evaluate the emotional state of an elder user or a user with some cognitive impairment. An RPI camera board connected to a raspberry PI2 modal B processing device is employed to capture frames from a recorded video of the user's facial expressions variation, the captured frames will then be processed using a python script to detect the face and recognize the apparent emotion. A set of various techniques are employed for face detection, facial feature extraction, and emotion classification such as HOG, regression trees, and PCA.