Research And Implementation Of Rehabilitation Assessment System Based On Deep Learning And Motion Capture Technology

Stroke is one of the top three causes of human death today.Patients who survive the disease are often accompanied by impairments in cognitive,language,sensory and motor functions,with motor dysfunction being the most significant cause of reduced quality of life.Rehabilitation training aims to alleviate these disorders and impairments through scientific and reasonable means.The development of different rehabilitation training programs for different patients’ functional impairments plays an important role in the recovery of functional impairments.Clinically,rehabilitation scales are often used to rate the degree of impairments of patients,but they are highly susceptible to the subjectivity of physiotherapist.In the context of the development of smart medical,it has become possible to provide objective functional assessment results for stroke patients using sensor technology and deep learning.Therefore,by establishing an automated functional assessment system based on deep learning and motion capture technology can help to provide comprehensive and objective functional assessment for patients and often a new door to advance the rehabilitation process in home or community settings.Based on this,the main work of this thesis are as follows:(1)For the upper limb motor functional assessment of stroke patients,we select suitable inertial measurement units to collect motor data of patients during the functional assessment.By exploring motor patterns of patients and combining the characteristics of LSTM,we propose a slope connections for LSTM with self-attention model as an automatic upper limb functional assessment model for stroke patients.(2)Based on the establishment of the automatic upper limb functional assessment model for stroke patients,a data collection position model was proposed,followed by the collection of motor data using Xsens DOT,and 30 stroke patients were recruited in a clinical setting to verify the validity of the models by incorporating assessment items from the FMA-UE scale.The experimental results show that the mean accuracy of the model was 90.5%,and the coefficient of determination between the predicted total score and the total clinical assessment score was 0.9617.(3)Considering the functional assessment needs in home and community settings,the Unity3 D game engine was used to complete the development of the automatic upper limb functional assessment system,and the SQL Server database software was used to store the rehabilitation data and realize the viewing of historical functional assessment results and the remote writing of rehabilitation training plans.The experimental results of this thesis show that the combination of sensor technology and deep learning provides the possibility of automatic functional assessment for stroke patients,and the objective and accurate functional assessment during rehabilitation of stroke patients provides a reference theoretical basis for the development of rehabilitation training plans.