Integrating electromyography data with pronation and supination angles for the classification of hand movements using deep learning

Even though huge improvements have been made in recent years regarding the classification of hand gestures, changes in the wrist position still affect the accuracy of the used model, representing one of the main limitations in the domain. This thesis aims to analyze the possibility of using wrist pro-supination angles to increase the classification accuracy of hand gestures from the myoelectrical signals of the forearm. Deep neural networks, in combination with multimodal data, are tested in an innovative configuration representing a potentially useful resource for hand prosthetics.