Sleep apnoea classification using multivariate Gaussian with EEG frequency bands and heart rate variability

In this paper, we propose Multivariate Gaussian Classifier in classifying features from EEG frequency bands and Heart Rate Variability (HRV) of ECG for healthy and sleep apnoea patients. It is believed brain and heart activities are altered in sleep apnoea patients. Five features (delta, theta, alpha, sigma and beta) were extracted from EEG frequency bands and three (LFnu,HFnu, LF/HF) from the HRV using spectral analysis. These features were tested on two models, the univariate and multivariate Gaussian models. By using multivariate Gaussian model, the combination of EEG and HRV features has achieved 64 % accuracy. To further improve the classification results, delta and acceleration coefficients were calculated and added to the original features. The classification results were improved to 71% using this method.