Work

Study on Characteristic Evaluation and Control of an Upper Limb Rehabilitation System

Year: 2015
Type: dissertation
Abstract: Stroke is the leading cause of disability, which severely affects the activity of daily living for patients.Benefitting from the found that neurons of humans are plastic, and the motor cortex functions can be altered by individual motor experiences, some strategies for rehabilitation training have been proposed, named neurorehabilitation training.Because the training process requires intensive, long duration and high-level exercise, it brings much burden to therapists.However, with the development of robotic technology, some robots have been designed for rehabilitation.Considering the shortcoming of existing robots used in upper limb rehabilitation, in this thesis, a home-used upper limb rehabilitation training system was proposed.In order to be able to be used at home, the device used for rehabilitation training should be more compact and portable.The developed Upper Limb Exoskeleton Rehabilitation Device (ULED) was thus applied in the system.To release the burden of therapists, a self-training concept, in which patients can finish the training exercises by themselves, was proposed.In self-training, the affected arm wore ULED and followed the motion of the intact arm.The control reference was based on surface electromyography (sEMG) signals recorded from the intact arm.A motion recognition method was applied to map raw sEMG signals into control reference firstly.An autoregression (AR) model was used to extract features and a back-propagation neural network (BPNN) classifier was trained to recognize motion patterns.For the purpose of improvement of recognition Abstract II rate, a wavelet packet transform method was applied to reduce noise and the features were extracted by a muscular model.The support vector machine was used to instead of BPNN to be the classifier.The recognition rate improved 10% average.Furthermore, to conquer the inherent drawback of the motion recognition based method that it can only provide a binary-like control reference, a muscular model based continuous angle prediction method was developed to predict elbow joint angle using only sEMG signals.Another important issue for the home-used rehabilitation system is to evaluate training effect remotely.A remote force evaluation system was designed for this purpose.The evaluation system can predict human-environment contact force using only sEMG signals.The isometric downward touch and push motion were studied in this thesis.Seven muscles around upper limb were used for recording sEMG signals.Two musculoskeletal models were applied to derive dynamic equations for the two motions, respectively.The parameters involved were calibrated by a Bayesian Linear Regression (BLR) algorithm.The haptic device "Phantom Premium" was used in the remote side to represent the predicted force.The therapist can hold the handle of the Phantom to feel and evaluate the identical contact force exerted by the patient from a remote side.The proposed system was evaluated by ten healthy subjects for the self-training function and force evaluation function.The RMS error for the elbow joint prediction method was below 10° while the ABS relevant error for force prediction method was below 20%
Author: Pang, Muye
Institution: Kagawa University
Language: English
Subfield: Rehabilitation
Field: Medicine
Open Access status: green
OpenAlex ID: W7146782375
Cites: 93
Cited by:
Related to: 10