7.2  Summary/re-cap of literature survey (previous state-of-the-art)

Many research attempts have been made at establishing correlations between EEG signals and drowsiness.  Many of these studies utilized laboratory tests while others used ambulatory data collected during actual task performance.  The preponderance of these studies looked for correlations between EEG spectral parameters and changes in task performance that can be directly attributed to excessive drowsiness.  Most researchers found some degree of correlation between energy in the traditional spectral bands of delta, theta, alpha, and beta and task performance.  Summarizing their general findings, most observed that individuals with their eyes open show a preponderance of beta activity while awake and alert.  As the subject's become drowsy, the energy shifts from the beta band down to the alpha band.  In general, for individuals with eyes open, drowsiness is associated with an increase in alpha and theta activity and a decrease in beta activity.  For individuals with eyes closed, drowsiness is associated with an increase in theta activity and a decrease in alpha activity.  The common conclusion is that drowsiness is associated with a decrease in the frequencies of the predominant energy bands.

Still, as much as is known about the correlation between the traditional frequency bands and drowsiness, these frequencies have not yet been successfully implemented into a drowsiness tracking system.  Even in the sleep scoring field which is very well established, the most disagreement between sleep scorers analyzing the same segment of sleep data occurs when scoring the stage W to stage 1 transition.  In addition, alpha activity tends to be an unreliable indicator of drowsiness and the state of eyes must be known in order to interpret many of the frequency shifts.  This complicates their use in ambulatory situations.  Most of the literature regarding spectral changes and drowsiness are qualitative and not quantitative, and there are still open issues regarding the determination of threshold values.

In short, EEG signal content was believed to be band limited from approximately 0 to 40Hz.  Higher frequencies in the EEG were thought to be the result of highly localized events such as spike activity.  No established correlations between states of consciousness (e.g. awake, drowsiness, sleep) were established with frequencies above 40Hz.  Frequencies outside of the 0 to 40Hz range were typically considered noise and were typically removed during filtering or discarded from analysis during post-processing.  EEG signals that were analyzed digitally were typical sampled at either 128 or 256Hz using 30Hz lowpass filters.  Signals above 40Hz were typically not examined.

<PREVIOUS

NEXT>