Pattern Recognition of EEG to Determine Level of Alertness.
Abstract
A modified pattern recognition approach has been designed for use on a general purpose digital computer for all night epoch by epoch sleep stage scoring. Comparison with experienced human scorers indicates an overall average agreement, for five nights, of 85 percent. On the same records, the overall average agreement for the three human scorers was 91.3 percent. There are five primary programs involved in the computer classification: (1) analog-to-digital conversion; (2) spectral analysis; (3) delta measurement; (4) pattern recognition; and (5) REM logic to identify stage REM. The stages 3 and 4 classifications are assigned by the delta measurement program. The pattern recognition program consists of two decision networks which separate stage 2, awake, and stage 1 or stage REM patterns. The stage 1 vs. REM dichotomy is performed later by the REM logic program.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1971
- Accession Number
- AD0726210
Entities
People
- William B. Martin