Model of information density measuring in e-learning videos

Abstract

Educational video content authors must be careful that the content is adequate for the audience and the purpose especially on the term level. On the other hand authors must be careful that the content does not cause information overloading. Viewers can experience it if the videos have too much motions and movements. Information overload has become an increasingly popular area of study and a lot of authors indicated that information density of movies and videos have risen in the last few decades. High information density of videos and movies may cause information overloading and it has been proved by Lang et al in their theoretic model named Limited Capacity Model of Motivated Mediated Message Processing. This theoretic model suggests how to measure information density in videos and movies but this method is slow and just trained observers can do this job well. In this paper the new model for information density measuring in videos and movies is described and tested. This new model is based on special usage of background subtraction algorithms and in this paper multi-layer background subtraction algorithm based on color and texture is used. This algorithm is presented by authors Yao and Odobez on CVPR Visual Surveillance Workshop in 2007. Results of our model are compared with the trained observer's results and the same video clips that have been used by Lang et al are used for testing. It can be seen that there is a significant correlation between information densities measured by trained observers and our model. Our model can be used to determine some forms of information density in videos and movies and it can be used to indicate higher possibility of information overloading.