Loading...

Knee Joint Torque Estimation using EMG Signals for Sharif Exoskeleton Control Applications

Ghiasi Noughaby, Amir | 2018

679 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 50981 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Vossoughi, Gholamreza
  7. Abstract:
  8. The human body has more than 600 muscles that cause movement. Disability and motor disorders are some of the problems that people may face due to some factors such as accident, spinal cord injuries, disorders and brain damage, and the presence of a problem in neurological commands due to stroke. One of the proposed methods for solving the problems of people with motor disorder is the use of exoskeletons to generate stimulus. Exoskeletones are electromechanical devices designed to assist human movement, and patients are used to correct their movements using specific motor patterns. Exoskeletones can also be used as an auxiliary device for carrying heavy loads. Many studies have been conducted to approximate the movement of exoskeletons to human natural movements.The goal of this project is to estimate the knee torque of one degree of freedom in a swing phase using the state space model based on the EMG data from the thigh muscles. The model mode equation combines muscle based on the Hill model and the direct dynamics of the joint movement, and uses kinematic variables as a function of the muscular activity of the muscles. To use the Hill model, the length and rate of change in muscle length, joint angles and EMG signal is needed to obtain the force in the joint muscle and torque. Accordingly, in order to investigate the Hill model and implement the calibration algorithm, the results of the experiments carried out at the Movafaghian Research Institute are first reported. The output of these results indicates the correctness of the model and accurate identification of the parameters for estimating the knee joint torque; the correlation coefficient between the estimated torque and the reference torque in these experiments is approximately 0.8. The estimated torque is also obtained with the lowest error value relative to the reference torque. Then the sensitivity analysis is performed for the selected parameters for calibration; all four selectable parameters show the effect of user geometric conditions and environmental conditions on the results. In the next step, the adaptation of Hill model and torque estimation algorithm are reported with different states of knee swing phases. In the following, after the experimental test bed has been prepared, the control loop is implemented to reduce the interactive effect by detecting human intent by EMG signals on the Sharif exoskeleton robot. The results show that the adopted control strategy has been able, despite the constraints, to control the proper interactive power. Then, after calibration over several days, it was found that only the parameters related to the user's geometry remained constant during different days. The robot was then controlled by an estimated torque from the Hill model. The interactive force between the user and the robot is reduced by about 25% using Hill's torque to detect intent; on the other hand, the torque required to move from the user's knees is reduced to a good degree. Finally, the analysis of muscle fatigue and its effect on the EMG signal and interactive force were investigated. It was observed that in case of muscle fatigue, the amplitude of the signal increases and the median frequency decreases; therefore, the RMS muscle signal and its MPF are considered to be suitable criteria for fatigue detection
  9. Keywords:
  10. Exoskeleton ; Electromyogram Signal ; Knee Joint ; Hill Muscle Model ; Exoskeletal System ; Torque Estimation

 Digital Object List

 Bookmark

No TOC