Automated Analysis and Classification of Sleep Stages Based on Machine Learning Techniques from a Dual-Channel EEG Signal

Abstract

AbstractIn automatic sleep stage classification on dual channels of EEG signals in gender-specific subjects. This proposed method followed the certain steps to complete this study. In this study, we have considered four basic steps like (i) pre-processing (ii) feature extraction (iii) feature selection (iv) classification. In this proposed work we have also considered some existing approaches, procedures and methods in subject to detect the abnormality in sleep stages. In this research work, we have selected one public dataset of sleep study named as ISRUC-Sleep dataset, which were prepared by sleep experts in sleep medicine department of Coimbra University (CHUC).This proposed research work was carried out with input of two channels of EEG signals. To analyse the discrimination of sleep behaviours, we obtained feature extraction and feature selection steps. In this research work, we have obtained SVM and KNN classification techniques, also for comparisons of classification results; we have considered other classifier as random forest classification techniques. According to the results, SVM classification techniques are most effective for analysis and classification sleep stages of subject-16 through C4-A1 channel with an accuracy of 97.20%, similarly for O2-A1 channel, subject-77 achieved with an accuracy of 98.53%. Finally, it has shown that with the random forest classification techniques, an overall accuracy performance is best compared to SVM and KNN classification techniques with respect to both input channels.KeywordsElectroencephalogramDual channelSleep stagesClassification