Multi-class machine learning based approach for sleep disorder identification using EEG signals

Accurately identifying sleep stages is crucial for the analysis and management of sleep disorders. Electroencephalogram (EEG) signals have shown promise in sleep stage identification. This research presents a novel method for classifying different sleep stages using EEG data based on a multi-class support vector machine (SVM) approach. The proposed method involves preprocessing the EEG signals to remove noise and artifacts, followed by feature extraction using wavelet transform and statistical methods. A multi-class SVM classifier is then trained to identify four sleep stages: waking, non-rapid eye movement (NREM) stage 1, non-rapid eye movement (NREM) stage 2, and rapid eye movement (REM) stage. The method is evaluated on a dataset comprising EEG signals collected from 20 healthy individuals during a full-night sleep study. The results demonstrate that the proposed technique outperforms existing EEG-based approaches, achieving anaverage classification accuracy of 86.2 for sleep stage identification. The suggested method has potential applications in clinical settings for sleep disorder evaluation and management. Additionally, it can be extended to other applications requiring the identification of various classes from EEG data.