Bandwidth Reduction of Sleep Information. Volume 2
Abstract
The report discusses the possibility of extracting sleep information from heart rate data. Eight hours of sleep EEG, EOG and electrocardiograms were recorded on FM magnetic tape for two nights for each of eight subjects. The time in milliseconds between heart beats was written on digital magnetic tape. The data were grouped into records containing 128 consecutive beat-to-beat intervals and a large number of descriptors was computed for each record. These descriptors for each record were the mean value, the sample variance, the nine- interval histogram of Z scores, and a set of Fourier transforms. Analysis of variance was used as a general guide to descriptor significance for each subject. The discriminant analysis procedure described by Rao and popularized by Cooley and Lohnes was used to sleep stage classify heart rate data. Accuracy of the procedure was determined in terms of percent correct classifications, correlation coefficient of the computerized sleep pattern with respect to the EEG hand scored pattern, and an empirically derived cost function.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1971
- Accession Number
- AD0740946
Entities
People
- A. J. Welch
- Jane Mockford
- Joanne M. Aldredge
- Philip C. Richardson
Organizations
- University of Texas at Austin