Design And Implementation Of Hand Movement Tracking And Recognition Software For Post-stroke Rehabilitation Training

Post-stroke is a disease caused by damage to the upper motor neurons,which can easily cause patients with different degrees of dysfunction.Hand post-stroke rehabilitation is mainly to prevent and deal with various neurological deficits during stroke,and hand motor function recovery is one of them.Hand movement is one of the ways to realize human-computer interaction.With the popularization of machine learning and deep learning methods in recent years,human-machine interaction technology based on hand movement tracking and recognition has made new progress.Computer vision-based hand motion tracking and recognition technology provides a new solution for hand rehabilitation training for post-stroke patients.Post-stroke patients’ hands are more special than normal people,often showing severe self-occlusion and high degree of closure,which makes it difficult to use hand motion tracking and recognition technology to assist hand rehabilitation training.For this situation,combining the requirements of hand rehabilitation training and the latest research progress of hand motion tracking and recognition technology,for the complex hand shape of post-stroke patients which can only provide a small amount of labeled data,and the requirements for real-time shape are high,we use the LSPS algorithm based on the partially supervised generation model to realize hand pose estimation and obtain the 3D joint point information of the patient’s hand in real time,and implement hand tracking.Then,summarize the hand gestures in the patient’s rehabilitation training process,calculate the relevant characteristics of the hand gestures based on the three-dimensional joint point information,realize the hand gesture recognition,and help correct the patient’s wrong hand gestures during the training process.Finally,combined with the current research progress in the field of hand rehabilitation,the doctor’s recommendations,and the observation of the hand condition of pose-stroke patients,the parameter data that can be used to evaluate the hand rehabilitation state of the patient are summarized,according to the joint point information obtained from the hand pose estimation,calculate the state parameter values and visualize the parameter data.In combination with the work above,a hand movement tracking and recognition software for post-stroke patients is realized,and hand rehabilitation training is achieved under complex hands,strong real-time performance and low cost.The proposed method,combined with a rehabilitation game system and put into use in hospitals,proves that the proposed hand rehabilitation training method achieves the expected functional goals.