PhysSocial
Led by: Hmamouche Youssef, Magalie Ochs, Laurent Prévot, Thierry Chaminade
Project Description
This is an under developement project which aims to provide a system for hemodynamic brain activity prediction based on behavioral features. The behavioral features represent verbal and non-verbal variables extracted during an experience of human-human and human-robot conversations conducted on several subjects. The goal is to detect the behavioral features that are responsible for the activation of each brain area, by means of prediction. A feature selection step is performed to select the input variables for the prediction of brain activity, then the most relevant input features are those how lead to the best prediction score.
Skills required to participate
fMRI data, conversations signals processing (video, audio and eyetracking data), Machine learning, Python.