3.1 Rationale

In this work we investigate the EEG waveforms of individuals when they change from a state of normal alertness to a state of extreme sleepiness.  Here, extreme sleepiness is defined as a physiological state in which the individual struggles against sleep, attention lapses occur, and sleep eventually ensues.  The change from normal alertness to extreme sleepiness is often associated with slower reactions and reduced ability to process information, so the individual is not able to quickly respond to changing conditions in the work environment and cannot properly perform job-related tasks.  Individuals working in this condition may present a danger to themselves and to others by increasing accident risk.

There have been some recent research efforts that have investigated the relationship between human performance and sleepiness [Torsvall et al., 1988, 1989; Akerstedt et al., 1991; Makeig and Inlow, 1993; Wierwille et al., 1992; Dingus et al., 1987].  Torsvall and Akerstedt showed that ambulatory EEG and EOG recordings in a normal nighttime working environment contained several indicators related to an increase in sleepiness.  One key indicator of increased sleepiness was related to frequency shifts in the energy content of the EEG signal.  However, their results only identified the existence of frequency shifts in the EEG, but not the specific quantitative measures needed to characterize the transition from normal alertness to extreme sleepiness.  Also, their work was not generalized to a large number of subjects, and the use of statistical averaging across large time periods and different subjects obscured much of the valuable transient information contained in short-time epochs of individual EEG signals.

Makeig and Inlow also identified frequency shifts in subjects as performance decreased, however, their tests were conducted with eyes closed and revealed different types of frequency shifts than would be seen in normal working subjects.  Also, their subjects were not sleep-deprived which would have amplified any sleepiness features present in the EEG.

Wierwille and Dingus conducted sleepiness research using a driving simulator and focused on eye measurements (e.g. EOG, video, etc.) to evaluate sleepiness.  Several statistics based on eye measurements (e.g. blink frequen­cy, blink duration, SEMs) may provide some indication of drowsiness, however, as documented in the scientific and medical literature, there are also several sleepiness indicators that can be obtained from processing the EEG signal.  It is evident that useful and important information is contained in the EEG signal.  The major objective of this work is to discover this information and then develop advanced signal processing methods to extract and interpret this information.  Because extreme sleepiness features are most easily identified by testing sleep-deprived subjects in a controlled environment without engaging them in complex tasks, this will be the basis of the experimental plan in this work.  Complex tasks would obscure useful data while providing little or no additional information.  In addition, tests involving complex simulators may not reflect true subject, as subjects often do not perform as they would in actual circumstances due to the arousing affects of the simulator and the lack of true physical consequences (e.g. as evident in driving simulators).

In the first phase of this research the EEG waveforms of subjects engaged in a vigilance task will be analyzed.  The objectives are to identify those patterns associated with the transition from a state of normal alertness to a state of extreme sleepiness and to clearly characterize transitions in alertness by employing various advanced signal processing methods to the EEG signals.  Baseline alertness and baseline sleep data will be used for each individual to compare with the vigilance test data.  An analytical measure driven by the EEG data will be developed to improve the detection of sleepiness in individuals participating in the experimental study.

<PREVIOUS

NEXT>