Experimental evaluation of subjective ratings of drowsiness and development of drowsiness definitions

Abstract

Researchers have struggled with the problem of obtaining an "accurate" operational definition of drowsiness. Drowsiness is difficult to define because it may involve many different indicators, such as different physiological measures. This thesis consists of two separate, but related, experiments to determine an optimal method of determining whether or not an individual is drowsy via physiological and observed measures. The first part of the experiment used behaviorally trained observers to rate different subjects on the level of drowsiness observed. The data collected showed that trained raters are relatively consistent when rating drowsiness.

The second part of the experiment tried to determine if there is a good physiological model to predict performance impairment due to drowsiness by collecting data on sleep deprived subjects. The subjects were given two interleaved tasks, low level and high level cognitive tasks, to perform while twenty-one performance and behavioral measures were collected. The results show that a regression model can be developed using eyelid closure measures, simple EEG measures and simple heart-rate measures to predict performance impairment due to drowsiness.