Automated sleep apnea classification based on statistical and spectral analysis of electrocardiogram signals

Well-being sleep is a significant segment for maintaining mental comparably as genuine flourishing. More than six-hour recordings are required to distinguish sleep apnea, which are extremely long duration recordings. It's difficult for a human to deduce the problem from electrocardiogram (ECG) readings. As a result, automated PC-based assessment is expected to detect abnormalities as early as possible. An automated framework for the classification of obstructive sleep apnea (OSA) can moreover be distinguished from the ECG Signals. From the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) polysomnographic informational collection. 18 subjects have been considered as data signals. The signal is segmented into 30 seconds and features are extracted by using the discrete wavelet transform (DWT). DWT of seven-level decomposition is applied on the segmented signal by using the wavelet 'sym3'. 12 features were extracted from each level and all of them are used to categorize the five types of sleep apnea. Random forest, k-nearest neighbor (KNN), and support vector machine (SVM) are used for classification of apnea. The random forest (RF) classifier outperformed the others with an average of accuracy (Acc) of 98.86% according to the study's findings. The experimental results show the developed model outperforms the state of art algorithms in the literature.