| dc.description.abstract | 本研究採用了眼動追蹤技術和腦電圖同步量測技術，以獲取解題過程中的眼睛注視行為和大腦神經活動的記錄，比較資優學生與一般學生在面對不同題目型態時的行為和心理反應。參與者包括15位資優學生與15位一般學生，兩組學生接受了相同的問題解題任務，並在解題過程中進行眼動追蹤和腦電波的同步量測。由於不同的眼動事件通常被用來間接解釋學員的認知歷程，我們使用眼動儀的眼動事件切割腦電波，直接觀察不同眼動事件下，腦電波的反應。題目共分為純文字題(tMCQ)與文字加有資訊圖片題(pMCQ)兩種，我們使用俄亥俄州立大學物理系的iSTAR題目與國中會考題目。在iSTAR中，資優生與一般學生在pMCQ的得分率為0.651±0.117 v.s. 0.385±0.205(p<0.01)，在tMCQ的得分率為0.444±0.233 v.s. 0.333±0.222；在會考題目，資優生與一般學生在pMCQ的得分率為0.711±0.187 v.s. 0.566±0.158(p<0.05)，在tMCQ的得分率為0.933±0.081 v.s. 0.701±0.214(p<0.01)。學員在iSTAR中pMCQ的解題中，顯示圖片到圖片的掃視(Saccade of picture-to-picture in pMCQ, Spp_p)(0.245±0.118 vs. 0.184±0.122；p<0.05)與答案到答案的掃視(Saccade of answer-to- answer in pMCQ, SAA_p) (0.155±0.099 vs. 0.136±0.080；p<0.05)有顯著差異；但是在需要工作記憶及計算的會考題目，則沒有眼動差異，顯示資優生在解題過程中，藉由整合文字與圖片資訊進行建模，而一般學生則比較少利用圖片資訊，較常將視野停留在答案區。然而在腦波分析則顯示，資優生此兩個事件的前額葉顯示出比一般學生更高的θ (0.227±0.249 vs. -0.025±0.202 for SPP_p)和 β 頻段的正能量變化 (0.076±0.178 vs. -0.030 vs.0.130 for SPP_p; 0.069±0.116 vs. -0.070±0.206 for SAA_p)。在α 頻段資優生則呈現較一般學生負的變化，此現象在中央區及頂葉區的通道相當明顯，由於alpha頻段與放鬆有關，顯示資優生有更高的專注力。除了上述pMCQ中Spp_p與SAA_p事件，分析其他眼動切割腦波事件也顯示資優生與一般學生在theta 與beta都有顯著差異的變化，顯示腦電波在認知過程的分析中，具有較眼動裝置更高的敏感度。資優生在大部分眼動事件的腦波都具有更高的Theta與beta能量，以及更低的alpha能量，顯示資優生在思考推理與專注力方面，優於一般學生。
本研究探究眼動與腦波特徵的重要性，因此採用ten-fold-cross-validation對這些特徵進行支持向量機（SVM）建模，發現眼動特徵加上腦波特徵相較於僅僅使用眼動特徵可以更準確區分資優學生和普通學生（ACCEYE+EEG vs. ACCEYE are 86.6% vs.70%, 76.6% vs. 53.3%, 80%vs.56.7%, 76.6vs.50% for pMCQ_istar, tMCQ_istar, pMCQ_exam, and tMCQ_exam, respectively）；分析發現不論是iSTAR或是會考，資優生在有圖片的題目中都具有較佳的表現，並且在SVM的結果顯示，iSTAR與會考題型中pMCQ分類結果都較tMCQ佳(86.6% vs. 76.7% for iSTAR與80% vs. 76.6% for 會考)，顯示具有圖片的題目，對於學員的數理能力的鑑別度較高。值得注意的是，在眼動分析中，雖然在會考題中，pMCQ的眼動分類，在兩群學生並沒有顯著差異，但是在腦波分析上，我們觀察到theta與beta的腦波顯著差異，顯示腦電波對於學員認知歷程，提供更直接的證據。
本研究使用眼動事件切割腦電波，直接解釋在不同眼動事件下，可以更精確的分析學員在解題過程中認知處理能力的差異，相較於將整段腦波分析，無法得到顯著的成果，可以發現使用眼動事件對腦電波進行處理，得以更精確分析不同解題過程中的腦波處理差異。 | zh_TW |
| dc.description.abstract | This study employed eye-tracking technology and electroencephalogram (EEG) synchronization to record eye fixation behaviors and brain neural activity during problem-solving tasks, comparing the behavior and psychological responses of gifted students with those of average students when faced with different types of problems. The participants included 15 gifted students and 15 average students, with both groups completing the same problem-solving tasks while undergoing simultaneous eye-tracking and EEG measurements. Since different eye movement events are often used to indirectly interpret students′ cognitive processes, we utilized eye-tracking event segmentation to directly observe the EEG response during different eye movement events. The problems were divided into two types: text-only questions (tMCQ) and text questions with accompanying informational images (pMCQ). We used iSTAR problems from Ohio State University′s Physics Department and questions from Taiwan′s junior high school entrance exams. In the iSTAR questions, the gifted students and average students scored 0.651±0.117 vs. 0.385±0.205 (p<0.01) in pMCQ, and 0.444±0.233 vs. 0.333±0.222 in tMCQ; in the entrance exams, the gifted students and average students scored 0.711±0.187 vs. 0.566±0.158 (p<0.05) in pMCQ, and 0.933±0.081 vs. 0.701±0.214 (p<0.01) in tMCQ. In iSTAR′s pMCQ problem-solving, significant differences were observed in the saccade of picture-to-picture in pMCQ (Spp_p) (0.245±0.118 vs. 0.184±0.122; p<0.05) and the saccade of answer-to-answer in pMCQ (SAA_p) (0.155±0.099 vs. 0.136±0.080; p<0.05). However, no significant eye movement differences were found in the exam problems, which required working memory and calculations, suggesting that gifted students integrated text and image information to model the problem, while average students tended to focus more on the answer area. EEG analysis showed that the gifted students exhibited higher positive energy changes in the theta band (0.227±0.249 vs. -0.025±0.202 for SPP_p) and beta band (0.076±0.178 vs. -0.030±0.130 for SPP_p; 0.069±0.116 vs. -0.070±0.206 for SAA_p) in the prefrontal region compared to average students. In the alpha band, gifted students exhibited a more negative change compared to average students, especially in the central and parietal regions, indicating that gifted students had greater focus, as alpha band activity is associated with relaxation. In addition to Spp_p and SAA_p events, other eye-tracking segmented EEG events also revealed significant differences in theta and beta activity between gifted and average students, demonstrating that EEG analysis has greater sensitivity in cognitive process analysis than eye-tracking devices. Most of the EEG data from gifted students showed higher theta and beta energy and lower alpha energy in most eye movement events, indicating that gifted students excel in reasoning and focus compared to average students.
Given the importance of eye movement and EEG features in this study, we employed ten-fold cross-validation to model these features using support vector machines (SVM). The results showed that combining eye-tracking and EEG features improved the classification accuracy for distinguishing gifted students from average students compared to using only eye-tracking features (ACCEYE+EEG vs. ACCEYE are 86.6% vs. 70%, 76.6% vs. 53.3%, 80% vs. 56.7%, and 76.6% vs. 50% for pMCQ_istar, tMCQ_istar, pMCQ_exam, and tMCQ_exam, respectively). Analysis revealed that in both iSTAR and exam problems, gifted students performed better on problems with images. SVM results showed that the classification accuracy for pMCQ was better than for tMCQ in both iSTAR (86.6% vs. 76.7%) and exam problems (80% vs. 76.6%), indicating that problems with images had a higher ability to differentiate students′ mathematical abilities. Notably, in the exam problems, while there was no significant difference in eye movement classification for pMCQ between the two groups, EEG analysis showed significant differences in theta and beta waves, suggesting that EEG provides more direct evidence of students′ cognitive processes.
This study used eye-tracking event-segmented EEG to directly analyze the differences in students′ cognitive processing abilities during problem-solving. Compared to analyzing the entire EEG segment, which yielded no significant results, using eye-tracking events to process EEG allowed for more accurate analysis of the EEG processing differences during different problem-solving stages. | en_US |