Research On Rehabilitation Motion Recognition And Judgment Methods For Stroke Patients

At present,the rehabilitation of stroke patients needs to be carried out in professional medical institutions.This will take up much more medical resource and a lot of energy for the patient.Therefore,effective and reasonable rehabilitation training for stroke patients become a social problem.Using artificial intelligence and computer aided equipment to resume training for patients is an idea.In this paper,the limb detection algorithm is used to detect the key points of the human body;the action classification algorithm is used to classify the movements;the Euclidean distance and correlation method are used to judge whether the movements are standard.The main contents of the study are as follows:First,the human limb detection algorithm is studied for the specific group target of stroke patients.This article describes the main frame structure of limb detection.Then the article starts with the existing methods of limb detection,compares the current mainstream algorithms:DeepCut algorithm,DeeperCut algorithm,Iqbal algorithm and OpenPose algorithm.Finally,the OpenPose algorithm is applied to the limb detection in this paper.Second,based on the detection of human limbs,the methods of existing motion classification and standard judgment are studied.We compare the classification effects of K-nearest neighbor algorithm,naive Bayesian algorithm,SVM classifier classification algorithm,dynamic time warping algorithm and hidden Markov model algorithm.The SVM algorithm is selected for action classification.We have improved the SVM algorithm for the problem that the action classification is not clear and the misclassification is generated.The improved content includes:preprocessing the data before the action classification;combining the principal component analysis method to process the joint point data;decomposing the continuous action into representative action segments and matching.The improved algorithm has a significant improvement in classification accuracy and speed.Finally,this paper combines the limb detection algorithm and the motion judgment algorithm to form a limb rehabilitation training system.We designed six sets of rehabilitation actions for this particular group of stroke patients.It includes training for the legs,arms and waist.The experimental results show that the patient’s limb rehabilitation training system proposed in this paper can effectively identify the patient’s movements and judge the correctness of the action.