|

|
|
|
|
|
EEG R&D » High Frequency EEG » 6.6
 

6.6  Closing words

In this chapter we have outlined criteria for the development of an off-line drowsiness tracking algorithm.  Single-frequency tracking was examined and we investigated tracking frequency bins both in the traditional and high frequency ranges.  Some of the characteristics of the high frequency bands were examined and these were briefly compared to the corresponding low frequency behavior.  The high frequency bands were determined to be suitable for a drowsiness measure and a tracking algorithm was developed based on the observed correlation between energy in the high frequency bands and "drowsiness" as related to test performance.

Characteristics of the tracking algorithm were then examined under various conditions.  The measure showed stability under alert conditions and displayed continuous tracking characteristics to changes in levels of alertness.  The measure also showed stability against voluntary changes in the state of the eyes (open or closed).  Large failures on the visual test were easily detectable by changes in the measure's output value relative to the fixed thresholds.  Data from sleeping subjects was then processed to show that sleep looks the same as visual test failures which leads naturally to the conclusion that the measure is tracking the behavioral phenomena as intended.

The algorithm developed in this chapter is very basic and uses a frequency band distribution determined from a single test subject's data.  But, as simple as the measure is, the performance clearly shows sensitivity to the physiologic variable(s) related to drowsiness which we are interested in identifying and tracking.  Recall that the measure, as developed, uses a fixed set of thresholds (also determined on a single test subject) which was then applied to all test subjects.  One potential area for improvement of the algorithm may be in the development of individualized thresholds.  The next chapter will discuss this along with other ideas for improving the detection algorithm along with a general overview and summary of the major contributions and findings of this research.



Copyright © 2010 Consolidated Research of Richmond, Inc. All rights reserved.