Predicting Finger Movement Using an Ensemble Machine Learning Approach

Abstract

Debilitating brain trauma caused by injury or stroke, and other neurological disorders, can hinder a person’s ability to use their hands. Brain-Computer Interfaces (BCIs) are a subject of great interest regarding augmenting or restoring functionality to victims of this type of trauma. Motor imagery is a process in which a subject under test imagines performing an action without physically doing so [3]. Using brainwave sensors such as EEGs, the state of neural communication for that action can be recorded without the interference caused by the movement associated with it [8]. Using Machine learning classification techniques such as Support Vector Machines (SVMs), Multilayer Perception models (MLPs), and Fischer Linear Discrimination Analysis (LDA), it is possible to select for features and accurately predict upcoming movement by using motor imagery training and EEG data collection.