IBM's e-Seminar system and the World Wide Web sprouted from similar roots and share a similar raison d'etre.  And while e-Seminar is not likely to change the course of modern civilization as has the Web, its ultimate impact on video networking and e-learning may be quite far-reaching.

The World Wide Web began life at the CERN particle physics lab in Switzerland as a means of distributing research data among the facility's scientists.  It was an idealistic attempt to create a system that would make the sharing of information so easy that fruitful scientific collaboration would become a daily routine.  That same sort of utopian thinking has also been the impetus behind the development of the e-Seminar system at IBM's Thomas J. Watson Research Center in Hawthorne, New York. But while the World Wide Web's original intent was to link together only the scientists in one facility, e-Seminar aims to link together every lab in IBM's entire Research Division.  That currently includes research laboratories in eight far flung locations: Westchester County, New York (T.J. Watson); San Jose, California (Almaden); Austin, Texas; Zurich, Switzerland; Tokyo, Japan; Delhi, India; Beijing, China; and Haifa, Israel.

In their recent IBM research paper ("The e-Seminar Lecture Recording and Distribution System"), authors Arnd Steinmetz and Martin Kienzle defined e-Seminar as "an IBM-Research internal prototype platform to allow all IBM-Research employees access to videos and slides of talks, seminars, presentations, and other events at any IBM-Research campus worldwide."

This fledgling Video On Demand (VOD) system serves both as an infrastructure for the distribution of seminar-style presentations and as a platform for further research into various underlying technologies, such as video recording/encoding, data streaming, video analysis, and data visualizing.

The e-Seminar project began about two years ago when Steinmetz and Kienzle, Parviz Kermani, Bob Berbec, and other IBM scientists at the Watson Research Center began thinking about ways they could apply their research in digital video streaming and networking to the internal learning-distribution problems.  Most of the learning that goes on at the lab takes the form of live unrepeatable one-time-only seminars or lectures.  Such events are common within the Watson facility and deemed important to the corporate culture.  Any given week sees seven or eight seminar presentations, along with regularly scheduled classes and perhaps a few lectures from outside guest speakers.  As with any scheduled event, however, there are always a certain number of IBM staffers who miss these educational opportunities due to sickness or schedule conflicts.  To compensate, IBM scientists dreamed up e-Seminar.  But then they decided to take it a step further, to extend the system's reach so that these unique educational opportunities would be available to anyone in any of the IBM research centers worldwide.

Go with the Workflow
IBM's e-Seminar is a comprehensive system with functionality that runs the gamut from event scheduling to worldwide distribution and usage tracking.  For purposes of discussion, Steinmetz and Kienzle, in their paper, break the whole video streaming process down into nine components/modules that describe the system's workflow: These include: Scheduling, Recording/Encoding, Analysis, Composition, Storage, Distribution, Searching/Browsing, Streaming, and Feedback/Communication.

As might be expected, the e-Seminar system takes advantage of current IBM technology and equipment.  "We are trying to eat our own cooking with this project," says Bob Berbec, a senior programmer at the Watson Center.  The system, for example, uses Lotus Notes, Domino servers, and a DB2 database for the Scheduling module.  Doing so meant staffers didn't have to re-invent the wheel.  "The calendar was already up and running here.  We just wrote on top of it," says Parviz Kermani, Manager, Automated Media System at the Watson Center.  "We use it to suck off information.  We just take advantage of it."

At the heart of e-Seminar's Recording/Encoding module (indeed, central to the whole system) is IBM's VideoCharger.  This server technology has been getting a lot of attention in the marketplace as a core product for Video On Demand systems.  VideoCharger works with ordinary Web browsers, but it streams video using a "push" rather than the conventional "pull" approach.  This provides a more broadcast-like environment in which video streams are started by a play command and will continue until stopped.  VideoCharger server runs on AIX or Windows NT and is the key component of IBM's Content Manager solution, where it is often tied in with other IBM products such as DB2.

The other two modules into which IBM developers have poured a lot of effort are the Analysis and Composition modules, according to Kermani.  That's because they realize that VOD systems need not only to make video available, but also render it easily accessible and useful.  You have to be able to find what you want when you want it.  Sub-modules with e-Seminar's Analysis module address various indexing and searching problems.

All of these technologies/methods generate metadata, the ultimate goal being to "make the video less linear and more searchable," Kermani explains.  This task is carried further by the Composition module where metadata is gathered and incorporated into the Web-based user interface and the entry page of the portal.  The storyboarding process done in the Composition stage/module, for example, produces thumbnails that users can click on to jump to specific parts of the video stream.

Keep it Simple
Automation is key to the e-Seminar system.  Its creators realized early on that a complicated system requiring a lot of hands-on operation and support would end up not being used very often.  They also realized that hiring a professional video crew to capture events would be cost-prohibitive and that the alternative—assigning video tasks to valuable staff scientists or IT personnel—was equally unattractive.  As Parviz Kermani observes: "Humans are expensive." Therefore, he and the other members of the e-Seminar development team resolved to automate tasks whenever possible.

One of the development team's innovations was to design and build special mobile "recording carts" to hold all the necessary video capture equipment.  Essentially closable tables on wheels, the carts look a lot like caskets, Bob Berbec says.  Each cart contains a tripod-mounted video camera, an IBM VideoCharger video server with MPEG encoder, a VCR for backup, wireless microphones, a client machine to control live multicasts, networking equipment, and all connecting wiring.  Before the e-Seminar gang got wise and started automating, every event setup robbed a staffer of an hour's time, spent simply plugging in the wires to connect all these components.  But by placing all the components on a mobile cart, they need never be disconnected.  The components are set up once and remain set up.  A staffer just wheels the cart into place and plugs in the power.

Also key to e-Seminar's labor-saving strategy is the use of several AutoAuditorium systems from Foveal Systems of Madison, New Jersey.  AutoAuditorium is an automated multi-camera video production system that is designed to capture auditorium presentations without a crew.  It uses proprietary motion-detection technology and remote-controlled cameras to "track" the presenter as he moves about the stage.  At the heart of the system is a CPU running software that, in addition to controlling the tracking cameras, also intelligently controls video switching and audio mixing.

Parviz Kermani says the arrival of the AutoAudorium systems at the Watson center was "quite a delight." With AutoAuditorium, all you have to do is "enter the time and date into the e-Seminar scheduling systems, and the cameras start rolling automatically at the appointed time," says Kermani.

According to Foveal Systems owner Mike Bianchi, "In the first three months of 2001, the AutoAuditorium System recorded 47 programs at IBM.  During that same time, they used a real live crew for only three videos." Bianchi estimates that hiring a professional video crew to capture those 47 programs would have cost IBM $80,000.

Current Affairs
Today, the Watson lab continues to capture, store, and distribute five to ten seminars/lectures per week using the e-Seminar VOD system, which is in operation 24/7.  The e-Seminar archive currently holds about 250 hours of material and has a potential audience of over 3,000 reachable users, as anyone inside IBM research worldwide can access it.  The system could be used to connect all the IBM labs for real-time interaction, but this idea is rendered unmanageable by "the old time zone problem," says Berbec.

Overall, Parviz Kermani is very pleased with the success of e-Seminar there at the Watson Center.  But one disappointment is that the system hasn't spurred more interaction among the other labs in the IBM research family.  Some IBM'ers at Watson had hoped other labs would take more initiative in recording their own videos and sharing them.  Instead, the majority of the action has been centered there at the Watson facility.  This is natural, however, considering that Watson is the headquarters for the research division and the largest lab in the division.  It is the place where most of the seminars and lectures originate.  One can't expect every lab to be as active as Watson. "After all, every lab has different equipment, different budgets, different needs," says Kermani.  But participation is likely to increase as more people are exposed to the system.  It is, after all, an experimental project that is meant to evolve.

Think About the Future
Only two years old, the e-Seminar system is still growing, says Kermani, who foresees the system being used less as simply an information distribution system and more as a learning system.  "My goal is to shift it to a learning tool," says Kermani.  "I see that this system has great value for students." But e-Seminar will need a lot of tweaking to become a full-fledged distance learning system.  It's a big step to convert a communication tool into a learning tool, says Kermani.

Bob Berbec agrees.  A learning experience often requires complex interactions, he points out.  It's difficult to capture a classroom learning experience as a sequence of shots because the "live" experience entails so much more. "You can't capture the learning moment - that light bulb going off in the student's brain - by showing shots of a blackboard," says Berbec.  "You somehow have to show the whole process, capture the whole drama of the experience." It's a daunting task that may require innovations in pedagogy as well as in technology.

But these are the kind of challenges that Kermani, Berbec, Kienzle and the rest of the IBM development team will tackle over the coming months and years as e-Seminar continues to be a platform and test bed for capturing not just video, but learning as well.

(Foveal Systems; 190 Loantaka Way; Madison, NJ 07940-1910; 973/822-2085; www.autoauditorium.com )