6. Summary
6.4. Rating the Prototype
In order to rate our approach of the distributed Automatic Lecture Recording software system, we look back to the constraints, limitations, and prerequisites listed in the beginning of this paper. We were concerned about aesthetic and financial constraints as well as about limitations in space and time. So, let us now go through these re- quirements and rate our approach.
6.4.1. Aesthetic Approach
In order to meet the demands of an aesthetic approach, we took the work of a human camera team as our role model. Our focus was not only to set up a distributed system where its modules bear the names of the different team members, but to enable these modules to mimic the work of their comparable human pendants.
We were successful in implementing basic cinematographic rules as well as more complex ones, namely
- respect the „line of action… while using four cameras,
- ensuring the minimal and recommended duration of a shot to fit the type of re- cording,
- avoiding to show a specific shot too often,
- avoiding to show recurring and therefore predictable sequences of shots in or- der to keep the spectator more focused on the content,
- making decisions which shot to show next, based on events in the environment registered by various sensors,
- inserting neutral shots to keep the spectator informed about the environment of the lecture,
- autonomous framing and adjusting of exposure parameters depending on the current scene,
- autonomous reaction on gesticulating and/or moving protagonists,
- arranging and conducting of shot ‡ counter-shot scenarios,
- use of transition effects and picture-in-picture effects for the video in order to react in a meaningful way on a larger number of incoming events from differ- ent sensors.
As we were able to realize quite a few of the relevant cinematographic rules, we con- clude that the aesthetic claim has been satisfied. The implementation of further cine- matographic rules by future work will continuously improve our result so far.
6.4.2. Affordable Approach
One important requirement is to minimize the costs to set up an Automatic Lecture Recording system in the lecture hall. In contrast to the huge expenditures presented in the Chapters 1.3.3 and 2.1.3, the cost for our prototype is mainly a one-time invest- ment into the equipment as the entire system needs at most one operator for starting and stopping if it is not done by the lecturer him- or herself.
Comparing the one-time cost of our system of about 11,200 ‹ for the equipment with no extra costs for staff to the investment into professional AV recording equipment of more than 20,000 ‹ and recurring costs of about 1,700 ‹ for the crew per recording day, or, as another possibility, even compared to the recurring costs for renting the equipment and the crew of about 6,500 ‹ per recording day, it is obvious that our ap- proach is much more affordable.
6.4.3. Space- and Time-Saving Approach
As our approach is based on AV streams over the network and on controlling the equipment over this network, it is no problem to place the computers in a different room than the lecture hall. While the cameras can be mounted fixed in the lecture hall, it is useful to keep the video server for the slides attached to the presentation computer
of the lecturer. Therefore, space is no issue as all necessary devices can be spread over different rooms.
Besides being frugal concerning space consumption, our approach helps to save the lecturer†s time and even that of operating staff. The entire system works autono- mously after being started manually until it is stopped manually. Thus, only at the beginning and at the end of a lecture, human action is necessary. Furthermore, as we also use our AV Mixer/Recorder computer to transcode the resulting lecture recording into many different streaming formats, and as we have automated this transcoding process and the publication on a streaming server (Lampi, Kopf & Effelsberg, 2006), the entire chain of providing students with lecture recordings is fast. Of course, it is still possible to manually perform any post-production steps (e.g., removing errors the lecturer made in class) before starting the transcoding.
Even if it is not possible to mount the equipment in a fixed way in the lecture hall, it is not very time-consuming to set it up. For our prototype, we brought all the equipment consisting of the computers for the lecturer†s presentation, the director computer, the cameramen computer, three PTZ cameras, the video server, the scan converter and all the necessary cabling into the lecture hall every time we recorded a lecture which was at least twice a week during the term. It was no problem to set the system up during the normal break between two lectures, which is 15 minutes, as we used a trolley on which most of the equipment was kept. After getting used to it, it took only about five minutes to set the system up from scratch.
6.4.4. Successful Prototype
Concluding the rating of our prototype, we assert that the initial requirements are all fulfilled. Therefore, we state that our approach is successful. When looking at our evaluation results and the evaluation results of (Rui, Gupta, Grudin, 2003) it seems to be realistic that an automated mimicry of a real camera team can still be detected even by non-professionals. In both cases humans detected the small things that differ to a real camera team and judged accordingly. Nevertheless, for all cases of lecture re- cording in which it is not feasible to employ a human camera team, our prototype can be a successful substitute. Furthermore, it is promising as a good base for future work which we will describe in the next Sections.