Detecting Inactivity Segments in E-learning Video Recordings ( DISE )

Abstract

Lecture video recordings used in e-learning typically contain segments of inactivity which are better removed in order to save storage and bandwidth load, and to decrease viewing time and effort. This problem is also common in other areas such as multimedia. This research proposes a method for detecting segments of inactivity in video sequences and marking them for further processing. The method is known as DISE. The core of our approach are two modules one for sound inactivity detection and the other for video inactivity detection. Inactivity segments detected by the sound module are forwarded to the video module for video inactivity processing. The method has been developed into a system (DISE) and tested using a collection of real lecture videos obtained from lecture recordings at University of Palestine for different courses under various recording conditions. The method performs close to human detection as it achieves accuracy of 84.4%. We believe that our method will add value to the process of recording lectures in the e-learning domain in educational institutes .