This study was undertaken to simulator test and evaluate a complete drowsy driver
detection system. The goal of the study was to recommend optimal specifications for a
system to be further studied in an actual vehicle. The system used a set of algorithms
developed from previously collected data and a set of previously optimized advisory
tones, advisory messages, alarm stimuli, and drowsiness countermeasures. Detection
occurred if eye closure or lane excursion exceeded predetermined thresholds. Data were
obtained from six sleep-deprived subjects who drove a motion base automobile simulator
late at night. Each subject was trained in carefully observing lane boundaries, using a
device which sounded an alarm if lane boundaries were exceeded.
The performance aspect of the system dominated the detection process. None of the
algorithms tracked well with the measures they were designed to estimate; correlations
were much lower than expected. The algorithms relied heavily on the positioning of the
vehicle relative to the lane.
next to an author's name indicates that all files or directories associated with their ETD are accessible from the Virginia Tech campus network only.
![[VT]](http://scholar.lib.vt.edu/images/ETD-db/restricted.gif)
