Date of Award
Spring 5-18-2018
Degree Name
Master of Science (MS)
Degree Type
Thesis
Abstract
We sought to design a cooperative brain computer interface (BCI), wherein multiple users contribute brain activities that are decoded towards a common goal. We used a base design involving collection of electroencephalographic (EEG) brain activity from a low-cost consumer system (the Muse Headband), then classified the ensuing signals into different mental states as either relaxed or focused. The goal of the cooperative BCI was to have two subjects drive a cursor on the screen to some acceptance range given a prescribed path. Each subject was responsible for controlling either the direction or the displacement of the ball. EEG patterns for the respective mental states were recognized and investigated through power spectral density estimation techniques. For the classification of patterns, we deployed linear discriminant analysis and support vector machine techniques on the gamma and alpha band limited EEG power. Our design yielded an average error rate of 14 percent and an average information transfer rate of 0.8 bit/s, despite the noisy data and limited array of EEG electrodes. With sufficient training for each subject, the cursor was successfully driven to the acceptance range. Our results establish the feasibility of cooperative BCI using relatively modest hardware.
Language
English (en)
Chair
ShiNung Ching
Committee Members
Kevin Wise Jason Trobaugh
Comments
Permanent URL: https://doi.org/10.7936/K7DZ07RC