Computer Vision-based Method For Dynamic Assessment And Mirror Rehabilitation Of Facial Paralysis

Facial paralysis is also caled facial nerve injury,the change of the face will bring psychological anxiety and physical pain to patients,seriously affecting the normal interpersonal communication and life of patients.Timely and effective evaluation of treatment and rehabilitation training is helpful for patients with facial paralysis to recover as soon as possible and restore normal quality of life.This study found that most of the studies on facial paralysis assessment are based on the local or overall features of the face.Current facial feature point detection models use facial face image data of normal people for training,and facial paralysis will lead to patients with strange facial features and crooked mouth Angle,so using the mainstream feature extraction model will cause a large error.The lack of internal feedback of facial muscle movement after facial paralysis will affect the brain to receive relevant facial movement information,resulting in facial nerve function degradation.In addition to clinical treatment,necessary rehabilitation training should be carried out.In this study,computer vision technology and three-dimensional face reconstruction technology are combined.Compared with the mirror image rehabilitation method and emg feedback rehabilitation method,the feedback is stronger,which helps to improve the rehabilitation training effect of patients.Therefore,the main research content of this topic includes two aspects:(1)facial paralysis assessment: first,a model suitable for face detection and facial feature point detection in facial paralysis was obtained by training based on facial image data of patients with facial paralysis.Based on the high robustness of the facial feature point detection model,this study innovatively proposed a quantitative index of facial area,which is helpful for the detailed observation of facial motor function of facial paralysis patients in a series of facial expressions and actions,and the objective dynamic evaluation of the degree of facial nerve injury can be realized.(2)facial paralysis image rehabilitation training: first,the depth camera is used to obtain facial depth information and feature point information in a series of facial expressions and actions,and the shape modeling of the virtual face is carried out through feature point fitting.In order to increase the reality of virtual face,face texture information needs to be scanned,matched and mapped.Then the real face and the virtual face were combined by feature points to achieve the synchronization of facial expressions,providing mirror feedback for patients with facial paralysis,aleviating their aversion and restlessness,recording the recovery process of patients’ condition,and evaluating the rehabilitation effect of facial paralysis.