Automatic Classification of Sleep/Wake Stages Using Two-Step System

This paper presents application of an automatic classification system on 53 animal polysomnographic recordings. A two-step automatic system is used to score the recordings into three traditional stages: wake, NREM sleep and REM sleep. In the first step of the analysis, monitored signals are analyzed using artifact identification strategy and artifact-free signals are selected. Then. 30sec epochs are classified according to relevant features extracted from available signals using artificial neural networks. The overall classification accuracy reached by the presented classification system exceeded 95%, when analyzed 53 polysomnographic recordings.