Recognizing Classroom Atmosphere with Good Attention to the Ongoing Lecture for Indexing its Archived Video

摘要

This article discusses how to recognize the atmosphere in a classroom where the students show good attention on the ongoing lecture from its archived video as a novel video indexing technology for e-learning application. We first show that facial orientations of the students in the classroom can be used as a useful visual feature obtainable from archived videos in order to recognize the classroom atmosphere with good attention from the students to the lecture, based on our analysis on the relationship between the behavior of the students and the classroom atmosphere that our humans actually recognize by watching the videos. Second, we present a method of recognizing the classroom atmosphere by estimating facial orientations of the students in each image frame of the video. In our experiment with a real archived videos, the success rate (F-measure) of the recognition was about 70%.