Feature extraction and optimisation for sleep apnea

Abstract

In this paper, the feature extraction and optimization for sleep apnea is investigated. The electrical activity of the brain along the scalp suffered from sleep apnea using Electroencephalogram (EEG) is addressed. The correlation between the EEG signals was compared to detect the features of sleep apnea. The Empirical Mode Decomposition (EMD) and Bivaiiate were adopted in this project to evaluate the extracted EEG signals. The performance of EMD has greatly improved when the number of samples was decreasing. The segmentation error analyzed in the Event Related Potential (ERP) reflected the occurrence of apnea. The delta power associated to the body autonomous system and homeostasis regulation is due to the drop of oxygen when sleep apnea happened. Using Hilbert Huang Transform, there is energy waveform in low frequencies when an apnea has happened. These can be linked to the delta power which relates to the body autonomous system and homeostasis regulation. The EMD, EEMD and Bivariate methods were compared to show key features linked with apnea for analysis purposes.