Restoration of facial expressiveness remains a major scientific and clinical challenge. The aim of this thesis project is to increment a facial mimic supervision system dedicated to the functional rehabilitation of the face using the biomechanical parameters of a numerical model of the face, deep learning and learning by reinforcement. A system for facial mimic analysis and real-time monitoring of facial characteristics will be developed. The evaluation of this system will be carried out to study its effectiveness as well as its acceptability in collaboration with the maxillofacial surgery department of Amiens University Hospital. This work is part of the work carried out in the MS2T labex (carried by the UTC) and the FIGURES team (carried by the Amiens University Hospital) concerning a decision support system to guide functional rehabilitation procedures. The restitution of facial expression and normal and symmetrical facial expression allows patients to improve their living conditions and their social identity and integration.