應用於遠距教學之學習專注程度偵測研究

Abstract

本研究進行學習專注度偵測的研究，藉由專注度偵測降低因為不專注導致學習進度的落後，並且將研究應用在較需要偵測專注度的遠距教學環境。本研究提出藉由人臉偵測和機器學習判斷影片中每張影像人臉的視線位置，透過發呆偵測以及臉部位移偵測取得動作資訊，使用影像分段處理以及滑動窗口處理連續性的影像，將影片的每個區段判斷成專心或不專心的狀態。 實驗資料來源包括高中補習班補課以及大學遠距教學兩種不同類型的學習影片，實驗結果發現專心行為判定的準確度為93%，不專心行為判定的準確度為81%。由結果得知本研究方法能有效地偵測到出現不專心行為的時間，透過臉部位移偵測方法也能避免做筆記的行為被判定為不專心。This research is about attention detection of learners in online learning setting. The proposed attention detection method uses face detection and machine learning to determine the learner's sight. By continuously process each frame of the video with daze detection and facial displacement detection. By using sliding window, each video segment can be judged as attentive or inattentive state. Experiments are conducted with two data sources, self-paced learning videos in cram schools and online learning videos of university classes. Experimental results show that the proposed method has a correct attentive behavior determination of 93%, while accuracy of the inattentive behavior determination is 81%.