陳志銘洪欽銘游智名YU CHIH-MING2019-09-042015-8-202019-09-042013http://etds.lib.ntnu.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22GN060070007H%22.&%22.id.&http://rportal.lib.ntnu.edu.tw:80/handle/20.500.12235/98892本研究利用CPT持續注意力測驗檢測受試者的注意力高低，以獲得對應之高低注意力腦波訊號，作為訓練與測試樣本，在腦波訊號處理上則採用小波轉換，據此從將不同腦波頻帶中抽取出影響注意力高低的特徵值，並利用基因演算法進行特徵選取，以有效找出影響注意力高低的主要腦波訊號特徵，再以支持向量機建立高低注意力辨識模型，據此發展出基於腦波訊號之注意力辨識系統，結果顯示本研究所發展系統的整體辨識率高達90.39%，可以有效辨識注意力高低。 本研究亦將基於腦波訊號之注意力辨識系統與具標記功能之影片撥放器進行整合，使得整合之系統可以偵測出學習者在觀看影片時注意力較為低落的影片片段，並據此進行低注意力影片片段的補救學習，以提升學習成效。經由多重實驗的驗證，顯示本研究發展之系統在辨識學習者的低注意力時間點上的準確率、召回率以及F測量值，均達一定程度的水準，並且系統偵測出的受試者低注意力時間點個數與學習成效呈現顯著的負相關性；低注意力時間點個數與基於低注意力時間點影片段之補救學習後的進步成績，亦呈現顯著的負相關。顯示本研究所發展之系統確實可以有效辨識出學習者在學習過程中的高低注意力。A Continuous Performance Test was conducted in this research to assess subjects’ attention levels. The corresponding brainwave signals collected were then used as the training and testing samples. Wavelet Transform was employed for signal processing. Features affecting the attention levels were extracted from various bands of brainwaves to perform a Genetic Algorithm for feature selection. An attention measuring model was generated with the Support Vector Machine after key features were captured and finally produced the Attention Recognition System. The System has yielded a total recognition rate of 90.39% that could effectively recognize subjects’ attention levels. A time-stamped supported video player was further integrated for learning result improvement during low-attending periods via remedial instructions. The results showed high precise rate, recall rate and F measurement value. Negative correlations of learning results and numbers of low-attending period as well as numbers of low-attending period and the range of improvement after remedial instruction were found. In sum, the Attention Recognition System can efficiently and effectively recognize individuals’ high and low attention level during the learning process.腦波訊號CPT小波轉換支持向量機基因演算法