李秀惠臺灣大學：資訊工程學研究所唐維志Tang, Wei-ChihWei-ChihTang2007-11-262018-07-052007-11-262018-07-052007http://ntur.lib.ntu.edu.tw//handle/246246/53928Polysomnography (PSG) is the must common procedures used for diagnosis of sleep states. One of important task of PSG is sleep stages scoring. Sleeping stage of each 30 second segment is determined by sleep specialist, and scoring stages manually consumes time and human resource. So many automatic sleep stages scoring system was developed. In this thesis, we proposed a feature set to replace the must common features relative frequency band energy. Our feature set includes harmonic parameters with wavelet transform, Hjorht parameters, wavelet entropy, and wavelet energy. We build an automatic sleep stages scoring system using SVM with RBF kernel using the feature set we found. The objective of this thesis is providing a better set of features form EEG signals. That can decrease the sensor numbers, and that may measure patients’ sleep state in their houses. The automatic sleep stages scoring model can help the sleep specialists save their time of scoring.Chapter 1 Introduction...............................................................................................1 1.1 Motivation................................................................................................1 1.2 Research Background and Goal.............................................................2 1.3 Organization of Thesis............................................................................6 Chapter 2 Relative Work............................................................................................7 2.1 Methods of Processing the EEG Signals......................................................7 2.2 Features of EEG Signals...............................................................................9 2.3 Automatic Sleep Stages Scoring System....................................................12 Chapter 3 Methods....................................................................................................14 3.1 Wavelet and Hilbert-Huang Transform....................................................14 3.1.1 Wavelet Transform...........................................................................14 3.1.2 Hilbert-Huang Transform................................................................17 3.2 Feature Extraction.......................................................................................19 3.2.1 Relative Frequency Band Energy...................................................20 3.2.2 Wavelet Entropy and Energy..........................................................24 3.2.3 Hjorth parameters, Domain Frequency and its Strength.............25 3.2.4 Harmonic Parameters......................................................................26 3.3 Support Vector Machine.............................................................................27 Chapter 4 Experiments and Results........................................................................30 4.1 Materials.......................................................................................................30 4.2 Feature Extraction.......................................................................................31 4.2.1 Relative Frequency Band Energy...................................................31 4.2.2 Wavelet Energy and Entropy..........................................................35 4.2.3 Hjorth Parameter, Domain Frequency and its Strength..............37 4.2.4 Harmonic Parameters......................................................................39 4.3 Prediction Model and Feature Selection...................................................44 4.4 Summary of Experiments and Performance Evaluation.........................48 Chapter 5 Conclusions and Future Works..............................................................52 Reference....................................................................................................................54461909 bytesapplication/pdfen-US睡眠分期腦波調和參數小波轉換EEGharmonic parameterssleep stageswavelet transformthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53928/1/ntu-96-R94922160-1.pdf