Force myography signal based hand gesture classification for the implementation of real- time prosthetic hand control system
This thesis aims to develop an interfacing mechanism for controlling prosthetic devices using Force Myography signal (FMG) and various hand gesture classifications. The FMG signals have been collected through three piezoelectric sensors banded around the forearm and Omega Data Acquisition (DAQ) System. The recorded data has been imported into Matlab, Simulink software for analysis and classification. The hand motion has been recorded through Virtual Motion Glove (VMG), and utilized in the system identification procedure to find out the dynamic relationship between the hand gesture and the corresponding FMG signals. Several classification and recognition models have been considered. Tree Decision Learning and Support Vector Machine (SVM) showed high accuracy results. Both of these estimated models generate above 82% of accuracy in terms of classification. The feasibility of the FMG signal for the implementation of a control system in the prosthetic hand is also tested. The result shows a high degree of accuracy in replicating the grasping gestures using threshold method. To limit and control, both the position and the amount of force applied at the fingertips of a prosthetic hand, a dynamic relationship has been established with the corresponding FMG signal through system identification method. These relationships will provide a useful foundation for the implementation and utilization of control system in an assistive device. In order to see the performance of FMG over electromyography (EMG), a comparative analysis has been performed by collecting EMG signals from the same groups of muscles. Unlike EMG, FMG signal is not affected by sweat, skin impedance, and doesn't need a reference signal.
Thesis (M.S.)--Wichita State University, College of Engineering, Dept. of Mechanical Engineering