PATTERN RECOGNITION OF EEG TO DETERMINE LEVEL OF ALERTNESS.
Technical progress rept. 2 Nov 68-14 May 69,
MCDONNELL DOUGLAS ASTRONAUTICS CO NEWPORT BEACH CALIF ASTROPOWER LAB
Pagination or Media Count:
This report documents the work accomplished during the second reporting period in applying the principles of pattern recognition technology to the analysis of EEG. Using EEG recordings, two sleep state classification systems, based on inputs derived from spectral analysis, have been designed, simulated, and tested. One system was based on an overnight sleep record of a single subject the other included, in the design data base, sleep EEG patterns taken from six subjects. The resulting pattern recognition systems were tested on sleep records from ten subjects and yielded reasonable classification of the training and test tapes. However, to reduce confusion between certain sleep stages i.e., 1 REM, 3 4, ... additional inputs may be required to supplement the basic frequency information. To assist in enhancing the classification ability of the recognition system a smoothing operation has been developed that monitors the systems output response and minimizes isolated misclassifications. In addition, a computer program to isolate and identify K complexes and sleep spindles is under development and shows considerable promise. Author
- Anatomy and Physiology