線上學習者分心行為偵測研究

Abstract

當我們從事重要的事情時，保持專心與集中精力至關重要。舉例來說，當駕駛車輛時，如果不專心駕駛不僅可能造成交通事故，還有可能導致人身傷亡和財物損失。再者，對於醫生來說，在進行手術時也必須保持專注，因為任何的分心都有可能導致嚴重的醫療錯誤，造成不必要的傷害。此外，在學習方面，當學生在讀書時，如果不能保持專心，可能會錯過重要的概念，影響其學習效果。本研究以線上學習者行為偵測為例，藉由電腦內建鏡頭拍攝學生上課情況，透過偵測發現課堂中常見的不專心行為。本研究提出以人臉偵測判斷影像中學生是否坐在座位上，並藉由頭部姿勢及眼睛視線判斷學生是否保持專注;同時使用哈欠偵測確認學生疲憊情況。此外，針對學生上課電腦螢幕進行場景偵測，自動辨別課程段落，進而探討學生是否專心於課堂。

本研究的實驗資料源自於本校研究所六名研究生協助拍攝的實際線上課程學習影像，通過實驗驗證各項行為偵測方法及整體可行性。實驗結果顯示，系統在整體學習專心程度偵測的準確率為 88%，由此可知，本研究方法能有效地偵測出線上學習者的專心與不專心。因此，本研究將進一步針對各受試者及各課程進行深入探討。

When we engage in important tasks, it is crucial to maintain focus and concentrate. For example, while driving, a lack of focus could not only lead to traffic accidents, but also result in personal injury and property damage. Similarly, doctors must remain focused while performing surgeries, as any distraction could cause serious medical errors and unnecessary harm. Furthermore, in the realm of learning, students who cannot concentrate while studying may miss important concepts and negatively impact their learning outcomes.This study takes online learner behavior detection as an example, using the built-in camera of a computer to capture the students' behavior during class, and detecting common inattentive behaviors in the classroom. The study proposes using face detection to determine whether students are sitting in their seats in the image, and judging whether students are maintaining focus through head posture and eye gaze. In addition, yawn detection is used to confirm the students' fatigue level. Furthermore, scene detection is applied to the computer screen of students during class to automatically identify course segments and explore whether students are focused on class.The experimental data for this study was sourced from actual online course learning videos recorded with the assistance of six graduate students from our university. The experiment aimed to validate various behavior detection methods and overall feasibility. The results indicated that the system achieved an accuracy of 88% in detecting focused learning behavior. This demonstrates that our research method is effective in detecting attentiveness and inattentiveness among online learners. As a result, we will further investigate individual participants and specific courses to gain deeper insights.