Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Department of Electrical and Computer Engineering

First Advisor

Brett J. Borghetti, PhD


Electroencephalography (EEG) classification of visual search and vigilance tasks has vast potential in its benefits. In future human-machine teaming systems, EEG could act as the tool for operator state assessment, enabling AI teammates to know when to assist the operator in these tasks, with the potential to lead to increased safety of operations, better training systems for our operators, and improved operational effectiveness. This research investigates deep learning methods which utilize EEG signals to classify the efficiency of an operator's search and to classify whether an operator is in a decrement during a vigilance type task, and investigates performing these classifications for any person's EEG signals, known as an EEG cross-participant model. This research also improves the EEG cross-participant model building research field as a whole, demonstrating the necessity of proper partitioning of datasets when building EEG cross-participant models in order to avoid overestimation of model accuracy, with empirical results presented which demonstrate that improper partitioning of datasets can lead to error rates underestimated between 35% and 3900%. The results of a conducted visual search experiment are also presented, in which EEG signals were captured while participants performed a visual search task, and various techniques were tested to mitigate inefficient search to efficient search. Efficient search was found to be faster than inefficient search, resulting in a 13% speed up, and also more accurate, with a 61% reduction in error rate. Two techniques (the nudge and hint) were also found to be effective in mitigation of inefficient search, resulting in a 169% increase of efficient searches. Lastly, EEG cross-participant models are presented which utilized spectral features to classify whether or not a participant was in a vigilance decrement during an unseen vigilance type task, with models performing better than random chance with 64% accuracy (95% CI: 0.59, 0.69).

AFIT Designator


DTIC Accession Number