Title Page
Abstract
국문 초록
Preface
Contents
Nomenclature 13
Chapter 1. Introduction 14
Chapter 2. Related Works 17
2.1. Imaigned Speech using EEG 17
2.2. Machine Learning Approaches 18
2.3. Linguistic Approaches 19
Chapter 3. Materials and Methods 20
3.1. Participants 20
3.2. Experimental Protocol 20
3.3. Data Acquisition 22
3.4. Data Preprocessing 23
3.5. Data Analysis 24
3.5.1. Classification 24
3.5.2. Relative Power Spectral Density 25
Chapter 4. Results and Discussion 27
4.1. Analysis of Classification Accuracy 27
4.2. Analysis of RPSD for Each Subgroup 28
4.3. RPSD differences for Inter-NL comparisons 31
4.4. RPSD differences for Intra-NL comparisons 33
Chapter 5. Conclusion 34
References 36
Table 3.1. Words in two languages used during the experiment and number of syllables. 22
Table 4.1. Comparison of classification accuracy in percentage for the whole frequency band. "s01" to "s06" and "s07" to "s12" are participant IDs of EN and CN, respectively. 28
Figure 1.1. Example of imagined speech-based BCI for drone swarm control. 14
Figure 3.1. Experimental paradigm for both English and Chinese tasks. Each trial consists of four imagination times for the presented words as a visual cue. 21
Figure 3.2. Experimental environment for imagined speech tasks. 23
Figure 3.3. Location of 64 channels used during the experiment. 24
Figure 4.1. Classification accuracy for each subgroup, task, and frequency band. It was averaged across participants, trials, and channels for the band groups. 27
Figure 4.2. The bar plots of the averaged RPSD. It was averaged across participants, trials, and channels for the band groups. The four subgroups were divided into five band... 29
Figure 4.3. The channel-wise RPSD in scalp topographies. Each scalp topography indicates the RPSD within the subgroups for the five different band groups. The scale bars... 30
Figure 4.4. Inter-NL differences of RPSD between subgroups in t-value scalp topographies. Each scalp topography indicates the distribution of channels with t-value using the... 31
Figure 4.5. Intra-NL differences of RPSD between subgroups in t-value scalp topographies. No statistically significant differences for two language tasks within the same NL... 33