Machine Learning Based System for Sleep Apnea Classification using EEG Signal

Sleep Apnea is a condition in which people and animals cease breathing for a short period of time while sleeping. EEG signals are important in detecting sleep apnea because they detect and record brain activity. This study proposes a method in which the filtered EEG signal is moved for sub-band decomposition after pre-processing and disbanded into five frequency bands: Sigma, Alpha, Beta, Delta, and Theta. The features determined for each frequency band acquired from the filtered EEG signals are energy, kurtosis, MAD, and skewness, which are then imported for the classification process using Decision Tree and Random Forest ML classifiers. Physionet’s publicly available dataset is used to assess each classifier’s accuracy, sensitivity, precision, and specificity. Using the bagging technique, Random Forest method achieves the highest classification accuracy of 99.68 percent, sensitivity of 100 percent, and specificity of 99.37 percent.