Date of Award

3-26-2015

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Electrical and Computer Engineering

First Advisor

Brett J. Borghetti, PhD.

Abstract

Across the DOD each task an operator is presented with has some level of difficulty associated with it. This level of difficulty over the course of the task is also known as workload, where the operator is faced with varying levels of workload as he or she attempts to complete the task. The focus of the research presented in this thesis is to determine if those changes in workload can be predicted and to determine if individuals can be classified based on performance in order to prevent an increase in workload that would cause a decline in performance in a given task. Despite many efforts to predict workload and classify individuals with machine learning, the classification and predictive ability of Electroencephalography (EEG) frequency data has not been explored at the individual EEG Frequency band level. In a 711th HPW/RCHP Human Universal Measurement and Assessment Network (HUMAN) Lab study, 14 Subjects were asked to complete two tasks over 16 scenarios, while their physiological data, including EEG frequency data, was recorded to capture the physiological changes their body went through over the course of the experiment. The research presented in this thesis focuses on EEG frequency data, and its ability to predict task performance and changes in workload. Several machine learning techniques are explored in this thesis before a final technique was chosen. This thesis contributes research to the medical and machine learning fields regarding the classification and workload prediction efficacy of EEG frequency data. Specifically, it presents a novel investigation of five EEG frequencies and their individual abilities to predict task performance and workload. It was discovered that using the Gamma EEG frequency and all EEG frequencies combined to predict task performance resulted in average classification accuracies of greater than 90%.

AFIT Designator

AFIT-ENG-MS-15-M-012

DTIC Accession Number

ADA616087

Share

COinS