Design and Control of a Myoelectric Prosthetic Hand using Multi-Channel Blind Source Separation Techniques

Wrist-disarticulated patients face several obstacles while performing their daily tasks such as eating a meal, opening a bottle, and so on due to the fact that they have a missing hand. Therefore, the purpose of this research is to develop a smart myoelectric prosthetic hand that can perform two gestures commonly used in these patients’ daily lives: button pushing and holding a bottle (grasping). In terms of the mechanical design, several aspects were considered to study its performance, such as the weight, size, and load it can handle. Static analysis is performed based on the Von Mises equation to inspect the structural failure of the prosthetic hand and fingers. For the myoelectric control, three blind source separation (BSS) techniques including Principal Component Analysis (PCA), Empirical Mode Decomposition combined with Independent Component Analysis (EMD-ICA), and Ensemble EMD-ICA (EEMD-ICA) were applied on surface Electromyographic (EMG) data obtained from 20 healthy subjects. BSS was used for extracting three motion-specific sources. As a result, 90% was the highest supervised machine learning classification accuracy obtained from the PCA-based separation technique using Fine Gaussian Support Vector Machine (SVM). Our future work will be extended by designing and implementing a complete prosthetic arm. We will also build the kinematic model of the system for the sake of optimizing the motion. In addition, we will classify more gestures for enabling patients to do a wider variety of daily tasks.