Classification of healthy and insomnia subjects based on wake-to-sleep transition

Abstract

This study is carried out with the aim of classifying healthy and insomniac subjects based on their wake-to-sleep transition (sleep onset process) features. The features were extracted from those signals using non-parametric and parametric methods in frequency domain. Wavelet transform was used to calculate non-parametric features: relative power of EEG sub bands (delta, theta, alpha, beta and gamma). After that Sleep onset reference epochs were determined using first and last intersection of delta and alpha respectively. The statistical analysis was applied on the features obtained. The data was divided into two groups: training data and testing data. Classification tree model was executed on training data to predict the healthy and insomniac groups in test data. K-fold cross-validation method was used for this estimation.