Covert Visuospatial Attention (VSA) for EEG-Based Asynchronous Control of Robot

Abstract

Brain-Computer Interface (BCI) is an effective modality for direct communication between brain and computer, bypassing brain's conventional communication pathway of the nerves and muscles. The objective of this work is to demonstrate EEG-based asynchronous control to a robotic device (Sphero) using covert visuospatial attention (VSA). Alpha and beta power band features in conjunction with a forward search technique for feature selection are used as input features for training a classifier model. The selected features are used to train Linear Discriminant Analysis (LDA) model. Average validation accuracy of 71.25% and average test accuracy of 66 % are obtained from eight subjects. We also investigated the effect of inter-session variations in feature patterns and accuracy of control using a multiple online VSA session. It is observed that there is a shift of discriminative features within subject across sessions. An asynchronous control session is also done and average test accuracy of 62.7% is reported which validates our objective for the work..