Towards the Amelioration of Classification Models for Evoked Potentials in Brain-Computer Interface

Abstract

Brain-Computer Interface technology has the potential to improve the lives of millions of people around the world. This study investigates how we may improve the performance of brain-computer interface for evoked potentials; we address some of the predominant challenges that deter its widespread availability and application, demonstrating ways to augment system bootstrapping and performance with the adaptation of classifiers that seem better suited to generalizing across human electroencephalographic data. This dissertation introduces ways in which deep transfer learning, together with interpretability, may ameliorate the approach to deploying pre-trained brain-computer interface systems that generalize well across users and tasks.