Generalisation in brain computer interface classification

Abstract: Brain computer interfaces (BCIs) are systems that allow users to interact with devices without relying on the neuromuscular pathways. This interaction is achieved by allowing the system to read the electrical activity of the brain and teaching it to map certain patterns of activation to certain commands. There are many applications for BCIs ranging from controlling prosthetics to gaming, but adapting both the user and the system to one another is a time and resource consuming process. Even more problematic, BCIs tend to only perform well for a single user and only for a limited time. This paper aims to investigate the accuracy of single-subject singlesession BCIs on other subjects and other sessions. To that end three different classifiers, a Support Vector Machine (SVM), Convolutional Neural Network (CNN) and Long Short-Term Memory network (LSTM) are developed and tested on a data set consisting of five subjects, two sessions for a binary classification task. Our results show that training on single-subject single-session data leads to an average cross-subject accuracy of 45-50% and an average cross-session accuracy of 50-55%. We find that there is no statistically significant difference in accuracy depending on the classifier used and discuss factors that affect generalization such as model complexity and good subjects.