Document Type
Thesis
Date of Award
2022
Degree Name
Master of Arts (MA)
Department
Psychology
First Advisor
Franck J. Schieber
Abstract
A high percentage of road accidents is caused by driver errors. As such, a first step toward preventing these accidents is the prediction of driver errors. Modern theories of attention suggest that the potential for errors increases with the amount of mental resources demanded from an activity (Kato, Endo, and Kizuka, 2009). In other words, monitoring the amount of mental resources may provide a viable way to predict driver errors. Pupillometry, the study of pupil dimensions and their reactivity, proposes that the diameter of the pupil expands concomitantly with the amount of mental resources mobilized. Pupillometry, therefore, offers a measure of mental resources that seems well suited for the problem at hand. However, because the diameter of the pupil is also responsive to brightness, an application of pupillometry in a driving environment represents a challenge. The present study proposes to overcome this challenge by discriminating the pupillary signature of mental effort from the pupillary markers of light responses. Through a frequency-based analysis of pupillary diameters recorded in response to both light and mental activity, several viable features (phase and log of the average magnitude of short-time Fourier transform of pupil signal) and methods of analysis are highlighted. Additionally, a support-vector machine algorithm suited for recognition across intermittent time series is used for the extraction of patterns from the recorded data.
Subject Categories
Cognitive Psychology
Keywords
Cognition, Fourier Transform, Machine Learning, Mental workload, Pupillometry, Support Vector Machine
Number of Pages
138
Publisher
University of South Dakota
Recommended Citation
Niamba, Kouadio Marc-Antoine Audoin B, "A Support Vector Machine Application for the Detection of Pupillary Markers of Cognitive Workload" (2022). Dissertations and Theses. 309.
https://red.library.usd.edu/diss-thesis/309
