Sleep Apnea frame detection based on Empirical Mode Decomposition of delta wave extracted from wavelet of EEG signals

Abstract

In this paper, we have proposed an apnea frame detection method based on the Empirical Mode Decomposition(EMD) of wavelet reconstructed delta wave of EEG signal. The method begins with wavelet transforming an EEG frame and reconstructing the low frequency delta wave from the approximate coefficients. EMD is carried on the reconstructed delta wave to generate intrinsic mode functions(IMF). Mean rate of variation and variance in the first five IMFs of the reconstructed delta wave are extracted as features from each frame. Finally SVM classifier is used to test the performance of the proposed method. From MIT-BIH sleep apnea database, the proposed method is tested with 13 overnight polysomnographic (PSG) records. The proposed method is applied on each patient and overall patients. We found accuracy, sensitivity and specificity rate of 80.43%, 85.59% and 77.87% respectively on overall patients. In conclusion, our proposed method is an efficient method for detecting apnea and non-apnea frames when only EEG signal is available and can be a great tool for PSG Sleep Apnea diagnosis.