A Practical Method for Automatic Real-Time EEG Sleep State Analysis

Abstract
A computer system for real-time analysis of the electroencephalograph (EEG) is described. The system performs continuous analysis, with graphic, analog, and tabular outputs, and storage of selected samples on disk for off-line analysis. Implemented mostly in high-level software, it is based on a two-component model of the signal in which waves are detected by a combination zero-crossing and peak detection algorithm. Each sample is classified by a pattern recognition scheme into one of several classes on the basis of the frequency distribution of waves; the classes correspond to normal sleep-awake states. Samples are taken ten times per minute and tabulated once per minute to provide a concise quantified history which is well suited to long-term EEG studies. Alternate independent information channels allow verification of results. Samples stored on disk may be grouped and averaged for statistical comparisons of EEG signal characteristics. The state classification algorithm has been tailored to the EEG of the cat; the results of a series of 7-8 day sleep studies are presented.

This publication has 16 references indexed in Scilit: