Human hand motion analysis and modelling for eating activity in rehabilitation exercise /
Patients with upper limb impairments have a reduced ability to perform their normal activities of daily living (ADL), and hence, deteriorating their quality of life. Eating is one of the fundamental activities of survival for all living beings. The robotic rehabilitation systems for people with uppe...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Kuala Lumpur :
Kulliyyah of Engineering,International Islamic University Malaysia,
2019
|
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4860 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Patients with upper limb impairments have a reduced ability to perform their normal activities of daily living (ADL), and hence, deteriorating their quality of life. Eating is one of the fundamental activities of survival for all living beings. The robotic rehabilitation systems for people with upper limb disabilities, must have the capability of assisting the patients, providing appropriate forces/torques, during various eating activities. Since, Human Upper Limb (HUL) motion is highly dexterous, a comprehensive hand motion analysis can be helpful in developing a dynamic model of the HUL, for predicting the torques during different eating activities. This research focuses on the analysis of HUL motion and the development of three-dimensional, 4 DOF mathematical model of the upper limb coupled with a robotic system, concentrating on the elbow flexion/extension, forearm pronation/supination, wrist adduction/abduction and wrist flexion/extension motions, using Kane's method and Artificial Neural Networks (ANNs), during different eating activities and using different cutlery. A motion analysis, simulation study, and experimental validation has been conducted involving five different food types and using two different types of cutlery, which are, a fork and a spoon, to study their effect on hand motion and the corresponding torques produced. Kinematic data from motion tracker sensors were utilized as inputs to the proposed models (i.e. 3D Kane's model and Nonlinear AutoRegressive eXogenous-Neural Network (NARX-NN) model) to determine the torques, while performing the different eating activities. Statistical tests like Analysis of Variance (ANOVA) and independent samples t-tests were employed to analyze data from the motion study. The results revealed that the bending motion of the index finger and the thumb varies with respect to different food characteristics and the type of cutlery used, whereas the bending motion of the middle finger remains unaffected. In addition, the contact forces exerted by the thumb-tip and index fingertip remain unaffected with respect to differing food types and the cutlery used. The motion of both wrist and elbow joints are influenced by the type of food and the type of cutlery (fork or spoon) used, considering the conditions involved in the experiment of this study. The torques predicted by the Kane's model and NARX-NN model were compared with the measurements from load cell sensors attached at each joint of the prototype robotic system. The results revealed that the torque predicted by the two models track the load cell torques closely. The 3D Kane's model and NARX model have an average RMSE of 0.05 Nm and 0.08Nm, respectively, for all eating activities. The estimation performance of both the models being comparatively equal, as such, NARX-NN model is appropriate to be used as a replacement of the dynamic Kane's mathematical model. These results verify that the proposed Kane's and NARX model, successfully model the HUL, during different eating tasks and using different types of cutlery. |
|---|---|
| Physical Description: | xix, 155 leaves : colour illustrations ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 141-146). |