DIGITAL SIMULATION AID IN DESIGNING AN AUTOMATIC EEG ANALYZER.
Abstract
Through period analysis, analog sleep EEG information was compressed into a series of numbers representing the incidence of intervals generated by zero-crossing of the EEG and its first derivative (digital differential) for 11-min. epochs. The resulting measurement vectors served not only for preliminary assessment of the descriptors as stage discriminators, but also for subjective comparison of EEG signals between leads for the same stages of sleep. In efforts to generate decision surfaces for dividing sleep into five stages (I through IV, and rapid eye movements), multivariate linear discriminant analysis was employed. To limit the number of variables for training and for discrimination of a sleep night, data from only one subject were used. A 23-variable set appeared to be the best choice. According to research results, 85% accuracy can be obtained on any night of sleep for a subject, provided that the training set is from the same subject. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1970
- Accession Number
- AD0708432
Entities
People
- Charles S. Lessard
- Harry M. Hughes
Organizations
- United States Air Force School of Aerospace Medicine