EEG Feature Extraction Using Time Domain Analysis for Classifying Insomnia

Abstract

Insomnia is one of the most ubiquitous sleep disorders. It appears frequently in those who suffer from depression, stress, and anxiety. The objective of the study is to categorize the insomnia groups from normal groups using time-domain analysis. In this analysis, the statistical parameters are extracted from full EEG wave, alpha wave, beta wave, delta wave, and theta wave of two different groups based on single-channel Fp2-F4 electrode. The obtained parameters from various wave patterns are considered as time-domain features. These features are applied as input to four different classification techniques namely, linear Discriminant Analysis (LDA), Logistic Regression (LR), Gaussian- Support Vector Machine (G-SVM), and Ensemble Subspace KNN to differentiate Insomnia from normal groups. In this paper, we have estimated the classifiers performance by employing 5-fold cross-validation. The results demonstrate that Ensemble Subspace KNN has the highest Classification accuracy and specificity of value 78.3 %, and 80% respectively.