Evaluate different machine learning techniques for classifying sleep stages on single-channel EEG

Abstract

In this paper, we propose 3 different machine learning techniques such as Random Forest, Bagging and Support Vector Machine along with time domain feature for classifying sleep stages based on single-channel EEG. Whole-night polysomnograms from 25 subjects were recorded employing R&K standard. The evolved process investigated the EEG signals of (C4-A1) for sleep staging. Automatic and manual scoring results were associated on an epoch-by-epoch basis. An entire 96,000 data samples 30s sleep EEG epoch were calculated and applied for performance evaluation. The epoch-by-epoch assessment was created by classifying the EEG epochs into six stages (W/S1/S2/S3/S4/REM) according to proposed method and manual scoring. Result shows that Random Forest classifiers achieve the overall accuracy; specificity and sensitivity level of 97.73%, 96.3% and 99.51% respectively.