Publication

Abstract : This paper presents an automated analytics system which monitors the students attending online lectures from a remote location and provides feedback to the teacher. The classroom videos are recorded and analyzed to identify the student trends, which might not be noticed by a teacher during class hours. Student behaviors are classified into five affective states: Active, Transcribing, Unavailing, Distracted and Transition. The student faces are tracked and optical flow of each student is calculated. Displacements and head motion of students are derived using a simple method and each student is automatically mapped to a particular affective state. The state transitions are provided to the educator as a feedback for assessment. Three different experiments were conducted with sessions from online courses, and was observed that the automated system efficiently differentiates the attention level of students and helped the teachers to improve their style of instruction. The engagement states also defined the height of student interest on a topic and classified the attentiveness automatically, without exploiting human effort.