| Summary: | 碩士 === 國立成功大學 === 電機工程學系碩博士班 === 97 === This thesis presents an automatic sleep stage classification and cyclic alternating pattern detection algorithm for sleep quality analysis. The proposed algorithm is applied to improve the performance of clinical therapy in obstructive sleep apnea issue. An automatic classification algorithm composed of a neural-network-based classifier, which constructed by the minimum description length (MDL) principle, is developed to classify the four types of sleep stages of electroencephalogram (EEG) signal. Subsequently, an automatic CAP detection algorithm with adaptive thresholds is utilized to evaluate the current sleep quality of the patient in the NREM stage. Finally, the abovementioned two algorithms are both applied to clinical therapy in sleep issues. The experimental results have successfully validated: 1) the proposed sleep stage classification algorithm can classify the four types of sleep stages of EEG signal efficiently; 2) the proposed CAP detection algorithm can reduce computational burden and achieve satisfactory performance; and 3) the novel therapy procedure combined with the above two algorithms can improve the performance and effectiveness of the conventional clinical therapy method in obstructive sleep apnea issue.
|
|---|
