Study On Human Motion Recognition For The Upper Limb Rehabilitation Training

Upper limb rehabilitation training can improve upper limb motor dysfunction in stroke patients.The traditional rehabilitation training should be carried out with the assistance of medical staffs and large training instruments.While this method can effectively complete the training of patients,it has high cost,poor autonomy,monotonous and difficult to evaluate the effect of rehabilitation in real time.In order to solve the above problems,this paper combines the human motion sensing technique with the rehabilitation training of stroke patients,uses the human motion sensing technology to provide the patients' rehabilitation training motion information,and studies the motion recognition problem of the training motion evaluation method in the upper limb rehabilitation training system based on SVM technology.This paper proposes a motion recognition method for upper limb rehabilitation training motions of the Brunnstrom 4-5 stage patients,and the upper limb rehabilitation training system based on this method was constructed.The main contents of this paper are as follows:1.Study on feature selectionThe generalization performance of motion recognition methods is often affected by the high-dimensional feature vectors of samples.Feature selection method can effectively improve the performance of motion recognition methods,reduce the sample dimension and the amount of computation,and improve the real-time performance of the tele rehabilitation system.However,in most studies,only consider the effect of individual candidate feature on the discernibility between classes,but ignore the joint effect of candidate features and selected features on the discernibility between classes.In order to solve this problem,a fisher score based on joint feature and support vector machine feature subset discernibility evaluation method is proposed.In this method,the concept of joint feature is proposed to introduce the joint effect of candidate features and selected features on the diversity between classes to the evaluation method.Experiments show that this method can get feature subsets with higher accuracy and smaller feature dimension.2.Study on multi-class classification motion recognition methodWhen using the decomposition and integration method to solve the multi-class classification problem with binary classifiers,there will be a problem of non separable region.In this paper,a Structure-Optimized DDAG(SODDAG-SVM)multi-class classification motion recognition method is proposed.According to the degree of separation of all kinds of class pairs,the structure of multi-class classifier of DDAGSVM is optimized.The SVM binary classifiers of easily separated class pairs guide the samples in the non separable region to the correct class.This paper use the GA method and the feature selection method proposed in this paper to optimize all kinds of SVM binary classifiers,and construct SODDAG-SVM multi-class classifiers.The experimental results show that the method can classify the samples in the non separable region in a better way,and can get better classification performance than that of MaxWins and DDAG methods.3.Build upper limb rehabilitation training system based on human body sensing technology.Based on the above feature selection and action recognition method,this paper constructs the hybrid human motion evaluation method,and combines the hybrid human motion evaluation method and the wearable human motion information collection system to construct the upper limb rehabilitation training system.The system can evaluate the patients' rehabilitation training motion information obtained by the human motion information collection system,and construct the 3D cartoon human model to teach and feedback the motion of patients to the patients.The recognition of rehabilitation training motion is an important research content in assisted rehabilitation training.The research work of this subject has important theoretical value and great practical value.It can improve the auxiliary rehabilitation training technology and alleviate the increasing demand of the society for the rehabilitation training.