|
|
|
 |
|
|
|
|
|
|
CNN應用於禪定與放鬆休息腦電波之頻率空間特性分類=Classification of Zen-meditation EEG and Resting EEG Spatial-spectral properties by Convolutional Neural Network |
|
|
|
| Author |
王少宏 (著)=Wang, Shao-Hong (au.)
|
|---|
| Date | 2019 |
|---|
| Pages | 201 |
|---|
| Publisher | 國立交通大學 |
|---|
| Publisher Url |
https://www.nycu.edu.tw/
|
|---|
| Location | 新竹市, 臺灣 [Hsinchu shih, Taiwan] |
|---|
| Content type | 博碩士論文=Thesis and Dissertation |
|---|
| Language | 英文=English |
|---|
| Degree | master |
|---|
| Institution | 國立交通大學 |
|---|
| Department | 電控工程研究所 |
|---|
| Advisor | 羅佩禎 |
|---|
| Publication year | 107 |
|---|
| Keyword | 腦電波=Electroencephalograph; 連續小波轉換=continuous wavelet transform; 自組織映射網路=Self-organizing map; 卷積神經網路=Convolutional neural network; 禪定=centroid frequency |
|---|
| Abstract | 本論文中，我們利用機器學習，自組織映射網路(SOM)去對禪定及放鬆休息狀態下的腦電波加以分群，並試圖探索禪定與休息狀態下腦電波(30通道)質心頻率的空間特性，且利用SOM分群的結果對腦電波進行標記，再利用深度學習，卷積神經網路(CNN)對其產生的腦電圖(brain mapping of 30 centroid frequencies, BMFc)加以分類。為了最佳化SOM分群的結果，本論文比較不同參數設定下的分群結果，並根據分群的結果，可以確定與比較禪定與休息腦電波質心頻率空間屬性的特徵。卷積神經網路(CNN)是一種前饋、誤差反向傳播之深度學習模型，被廣泛應用於圖像的分類，針對不同的分類目的及資料集複雜度，設計者可設計不同之CNN架構。此篇論文比較了不同CNN模型在分類BMFc上之表現。禪定下腦電圖之最高分類率為94.61% (Model-4)，放鬆休息之腦電圖最高分類率為95.88% (Model-2)，禪定下腦電圖之平均分類率為88.87%，放鬆休息之腦電圖平均分類率為88.73%。
This thesis is aimed to investigate the spatial-spectral properties of 30-channel Zen-meditation and resting EEG (electroencephalograph) based on classification of brain mapping of centroid frequency by self-organizing map (SOM) and convolutional neural network (CNN) models. Input data entry is the brain mapping of 30 centroid frequencies (abbreviated as BMFc) extracted from CWT (continuous wavelet transform) coefficients of each channel. Based on the unsupervised learning scheme, SOM mainly performs the matching of the input feature vector Fc and the cluster center representing the quantitative features of the output cluster. In addition, to optimize the clustering results, it is necessary to carefully select the implementation parameters of SOM. From the clustering results, the major spatial-spectral features of Zen-meditation and resting EEG may be determined and compared. With the clustering results by SOM, we are able to label each BMFc and train the CNN models to classify the dataset of BMFc’s. CNN is a hybrid feedforward, error backpropagation model in deep learning that is mainly applied in image classification. The structure of CNN can be designed to match different dataset. This study compares the performance of different CNN models on BMFc classification. The best classification accuracy achieved for classifying Zen-meditation BMFc is 94.61% (Model -4) and for classifying resting BMFc is 95.88% (Model-2). Average classification accuracy is 88.87% for Zen-meditation BMFc and 88.73% for resting EEG.
|
|---|
| Table of contents | 摘要 i
Abstract iii
List of abbreviation ix
List of Figures xi
List of Tables xix
Chapter 1 Introduction 1
1-1 Background and Motivation 1
1-2Aims of this study 4
1-3 Scope of thesis 4
Chapter 2 Theories and Methods 5
2-1 Continuous Wavelet Transform 5
2-2 Self-Organizing Map 7
2-3 Convolution Neural Network 14
2-3-1 Convolution layer 17
2-3-2 Activation function layer 18
2-3-3 Pooling layer 20
2-3-4 Flatten layer 21
2-3-5 Full connection layer 22
2-3-6 Loss function 24
2-3-7 Backward propagation 26
2-3-7-1 Gradient Descent 26
2-3-7-2 Momentum 27
2-3-7-3 Adagrad 28
2-3-7-4 Adam 29
2-4 Evaluation of clustering performance 30
2-4-1 Pearson correlation coefficient 30
2-4-2 Cluster center 31
2-4-3 Intra-cluster distance 31
2-4-4 Intra-cluster PCC 32
2-4-5 Inter-cluster distance 33
2-4-6 Inter-cluster PCC 33
Chapter 3 Experiment Procedure 35
3-1 Signal Acquisition and Brain Mapping 35
3-2 Classification of Brain Mapping of Centroid Frequency 36
3-2-1 Outline of Complete Scheme 37
3-2-2 Centroid Frequency 38
3-2-3 Self-Organizing Map (SOM) 40
3-2-4 Convolutional Neural Network (CNN) 42
3-3 Parameters Analysis for SOM 47
3-3-1 Number of output neuron units (No) and training step (Nts) 48
3-3-2 Initial learning rate (α0) and initial neighborhood size (σ0) 50
3-3-2 Datasets for CNN 52
Chapter 4 Results and Discussion 55
4-1 Centroid frequency 55
4-1-1 Fc Properties of Resting EEG 55
4-1-2 Fc Properties of Zen-meditation EEG 62
4-1-3 Comparison between resting EEG and Zen-meditation EEG 67
4-2 Results of SOM Clustering 68
4-2-1 Result of SOM clustering for Resting EEG 68
4-2-2 Result of SOM Clustering for Zen-meditation EEG 77
4-3 Results of CNN Classification 88
4-3-1 Introduction of different CNN models 88
4-3-2 Result of Resting EEG Segment-1 89
4-3-3 Result of Zen-meditation EEG Segment-1 93
Chapter 5 Conclusions and Discussions 97
5-1 Conclusions 97
5-2 Future Work 108
Appendix A Resting EEG (0827, 2nd segment) 110
A.1 Fc Properties of Resting EEG (0827, 2nd segment) 110
A.2 Results of SOM Clustering 115
A.3 Results of CNN Classification 122
Appendix B Resting EEG (0827, 3rd segment) 126
B.1 Fc Properties of Resting EEG (0827, 3rd segment) 126
B.2 Results of SOM Clustering 132
B.3 Results of CNN Classification 138
Appendix C Zen-meditation EEG (0830, 2nd segment) 142
C.1 Fc Properties of Zen-meditation EEG (0830, 2nd segment) 142
C.2 Results of SOM Clustering 148
C.3 Results of CNN Classification 155
Appendix D Zen-meditation EEG (0830, 3rd segment) 159
D.1 Fc Properties of Zen-meditation EEG (0830, 3rd segment) 159
D.2 Results of SOM Clustering 165
D.3 Results of CNN Classification 172
Appendix E Dimension of data through the CNN model and parameters computation 176
Appendix F Formal Zen-meditation Practice 193
Appendix G Experiment setup and procedure 194
G-1 Zen-meditation EEG 194
G-2 Resting EEG 194
G-3 Signal Acquisition 194
Reference 196 |
|---|
| Hits | 280 |
|---|
| Created date | 2022.09.22 |
|---|
| Modified date | 2023.02.17 |
|---|
 |
Best viewed with Chrome, Firefox, Safari(Mac) but not supported IE
|
|
|
