Detection of Dozing Students from an Ordinary Classroom Scene with CV technique

Abstract

In the recent years, a development of the Internet and information technique is enable to use IT devices in an ordinary classroom. One of them is videoing with a camera. We decided to take a video of a lecture with information about students who is taking part in it. The proposed system takes a video so that student’s face comes to the front. The system extracts a rectangle around each student from each frame in the video. It estimates state of student by using the face detection technique and the flesh color detection one. The system assumes the flesh color of Japanese people. In this paper, we describes a method detecting dozing students. A simple experiment is performed with 90-minute lecture. In the experimental result, state of student is effectively estimated. The system takes a class with a single camera and detects dozing of multiple people taken.