Classification of Normal and Hypoxia EEG Based on Approximate Entropy and Welch Power-Spectral-Density

This paper reports a novel method to classify EEGs from subjects under normal and hypoxia conditions, which provides a potential efficient indicator to evaluate hypoxia in real time. The EEG data are collected from 3 healthy subjects while their neurobehaviors are evaluated to assess the degree of hypoxia. Together with Approximate entropy (ApEn), the specific energy in a sub-band of 30-60 Hz of the Welch power-spectral-density (PSD) is extracted as the features. Bayesian classifier, 3-layer perceptron established by back-propagation and SVM are utilized for classification, respectively. The accuracy of Bayesian classifier is over 90.8% on test set. We compared the performance in terms of changing the architecture of the net. The accuracy of BP network reaches 94.2% on test set. Meanwhile, a SVM with Polynomial kernel revealed an accuracy over 92.5% on the test set. The experimental results show that the hypoxia EEG can be distinguished from normal one for individuals remarkably.