Multimedia Scenario Extraction and Content Indexing for E-Learning

In this paper, we present the use of multimodal content analysis in the MARVEL (multimodal analysis of recorded video for e-learning) project. In this project, we record teachers giving their lectures in class and semi-automatically analyze the video-audio content in order to help transfer this lecture into a multimedia course for e-learning. We distinguish two primary goals in this application: scenario extraction (mostly from video) and content indexing (mostly from text and speech). Three objects take place in these goals: the teacher, the screen (for slide projection) and the whiteboard (for handwriting). These goals and the roles of all objects are explained in details, as well as our preliminary results. Through this application, we are giving some ideas about multimodality analysis and its formalization.