| Summary: | 碩士 === 國立臺北科技大學 === 電機工程系 === 106 === Clinical research has shown that Sleep Apnea Syndrome (SAS) is a significant risk factor for many diseases. If it is not detected early and treated carefully thereafter, it will obviously affect daily life and cause cardiovascular disease. At present, there are approximately 85 % of SAS patients are of Obstructive Sleep Apnea (OSA). People who are suspected having sleep apnea must take Polysomnography (PSG) experiments in the sleep center at night. The sleep technologist and specialist will evaluate the Apnea and Hypopnea Index (AHI) for testing whether they have the SAS. The automatic detection methods have been gradually promoted due to the convenience of the patients as well as the reduction of increasing labor cost.
In this study, Electroencephalography is used to determine whether a sleep apnea termination event has occurred. The delta wave in the electroencephalogram is obtained by using wavelet transformation, and then converted into intrinsic mode functions (IMFs) through the empirical mode decomposition (EMD). Both the mean absolute variation and the variance of IMFs are used as features in the experiments. Eleven kinds of classifiers were employed for training and testing. From which, three kinds of classifiers with excellent performance were combined for decision making, including Bagging with REPTree, Adaboost with Decision Stump and Decision Table. The experimental results show that the Majority Voting can achieve 92.8 % for accuracy, 91.4 % for sensitivity and 92.9 % for specificity.
|
|---|
