Automatic capture and presentation creation from multimedia lectures

Abstract

For more than a decade, the RIPPLES group at the University of Massachusetts Amherst has developed and deployed content delivery systems to support on-campus and distance education. This has been an effective but expensive approach because of its dependence on content captured by human operators and substantial production effort. Presentations Automatically Organized from Lectures (PAOL) is a classroom-capture system that is transparent to and places no constraints (teaching style or pedagogy) on a lecturer. PAOL is unique-no other system can capture unconstrained computer-based materials, capture whiteboard writing and drawing, and produce a speaker video without requiring preinstalled software, electronic whiteboards, or special training. PAOL uses highresolution cameras, a computer screen capture device, and a wireless microphone to automatically create multimedia Flash presentations, which include a digitally edited instructor video, enhanced images of all material presented by computer or written/drawn on a whiteboard, and an index to support navigation. The PAOL system has been used extensively during system development in several varied settings. We report on usability studies to determine the most effective presentation formats and interfaces and our efforts to capture a full course to further evaluate the impact on teaching and learning.

Leveraging Electromagnetic Polarization in a Two-Antenna Whiteboard in the Air

“…Whiteboard transcription systems.. These use cameras to record a whiteboard's contents [10,14,51]. While they function well in a controlled lecture theater setting where infrastructure can be permanently installed, they depend on sufficient ambient light, and require a mostly unobstructed line of sight from a camera placed at a sufficient distance from the whiteboard, precluding deployment in some scenarios and complicating deployment in most.…”

MANIC content delivery in a mobile application

“…Several computer vision and machine learning techniques [8,9] are used to process the captured video as well as several open source tools like FFmpeg to create videos. The lecture video is digitally enhanced to improve audio, remove noise, correct radial distortion and perspective distortion in the video frames, and improve video quality before it is converted to Adobe Flash format.…”

“…While RIPPLES tried a few approaches to automate content creation and capture, the demand for quicker and less expensive production and more flexibility in content delivery led to the development of PAOL [3]- [8], a content capture system, that is used fully in instrumented classrooms to capture video/audio of the lecturer, any content (such as slide presentations or software demonstrations) displayed on the instructors laptop, and any content written on whiteboards or projected on screens. PAOL emphasizes transparency to the lecturer, a key tenant of the RIPPLES group.…”

LeCTo: A rich lecture capture solution

“…Today there are various commercial and free lecture capture systems available focusing on different aspects of lecture capturing process: some focus on simplifying the recording process [7], [8], or automatizing camera control during recording [9], [10], some attempt to improve lecture reproduction properties by allowing the consumer to focus and manage only target region of interest in the lecture video [11] or by improving synchronization mechanisms [12], while some systems focus on hosting and indexing of lecture capture materials [13].…”