July 2001

BY BROUGH TURNER

There's also more to video than just the conferencing market. There's broadcast video, and now video newscasts, available on demand via the Web. There's video training, including remote classroom instruction. And there's remote monitoring -- formerly by "surveillance" cameras, but now by the more friendly-sounding Web cams.

Video conferencing benefits from advances in related video markets, as well as from the continuing increase in available processing power (following Moore's law) and an even faster growth in available bandwidth. Finally, IP convergence is facilitating new services and new combinations of services via video over IP. IP has also brought new ideas to an industry that was stuck in a rut. With new processing power, new bandwidth, and new ideas, video over IP holds the promise of actually making video telephony a viable and popular form of communications. This won't happen overnight, but the stage is set for some dramatic changes over the next three to five years.

VIDEO OVER IP
Video over IP has already appeared in consumer applications. Video news clips are widely available on the Web. And major events, like the 1998 release of President Clinton's videotaped grand jury testimony, introduced many new surfers to the experience of video on the Web. Today, most broadband access services (cable and DSL) provide access to video news clips on their default home page.

Web cams, which initially provided a new image every 15 minutes, are now available with refresh rates approaching live video. Many daycare centers are installing Web cams that allow you to see your child at the center during the day. And more traditional surveillance applications -- banks, for example -- are moving to video over IP.

These video over IP applications work on today's public Internet, however, because they are unidirectional. It is possible to buffer content for many seconds, so that you receive streaming video even though the public Internet is subject to temporary delays, packet loss, retransmissions, etc. But it's not so easy for video telephony, or person-to-person, bi-directional communications. Can video over IP make video conferencing as good as being there?

VIDEO OVER IP CHALLENGES
People participating in a video conference want to see not only the person talking, but also their presentation or other supporting materials. If it is an interactive conference, then it is desirable to see multiple people. If it's a video connection for remote medical diagnosis, then multiple views are a must. To be effective, video conferencing over IP must include video over IP, voice over IP, and data sharing over IP, and must address all the challenges that exist in this multimedia environment -- quality of service (QoS), latency, and bandwidth.

QoS
An advantage of IP for video conferencing is that it supports multiple applications -- video, audio, and a slide presentation -- on a single connection, and allows the appropriate level of video equipment to be connected as needed for a particular meeting (or, say, a medical diagnosis). Putting the conferencing equipment on IP, however, means that it is on the same network with a lot of other applications. Moving these applications through the access network and onto the WAN requires the ability to give priority to delay-sensitive voice and video packets.

Many tools are available today to address this issue -- ATM, MPLS, DiffServ, and Ethernet priorities -- but, for now, they pre-suppose a private network or virtual private network (VPN) between locations. If such a VPN may be assumed, one may then configure IP differential services directly with an IP Multi-Service Unit or indirectly by mapping IP applications to Ethernet priorities in the LAN, and MPLS or ATM virtual circuits in the WAN. It's likely that direct differential IP services will emerge from data networking carriers over the next year or two, so this situation should get substantially simpler.

Latency And Frame Rate
Historically, the majority of video conferencing systems used the H.320 protocol over ISDN or Switched 56 data service. In most cases, the bandwidth was limited to 128 kbps or 112 kbps. As a result, the image was updated very slowly, and the audio was delayed to match. Typical delays were on the order of a second or so, making it hard to avoid talking over the other party. And the slow video updates were very unnatural, with jerky response when someone moved.

As silicon has improved to provide more and more processor MIPS and/or hard-coded ASIC implementations, the industry has moved to ever more sophisticated coders that compress the video (and audio) information to better utilize limited bandwidth. Today's video coders compress an individual scene and, thereafter, transmit only frame-to-frame motion differences. However, further coder enhancements are unlikely to improve performance by more than another factor of two or four, at least in the next few years. Greater magnitudes of improvement can only come from increasing the available bandwidth.

Bandwidth
Most of the visible delay in a video conference can be reduced by using high-bandwidth connections -- 1 Mbps or more as opposed to the 128 kbps typically provided by ISDN today. As with IP telephony, if latency can be reduced to less than 250 ms, the result is a much more natural system. And at 1.5 Mbps, TV-quality video is readily achievable. But how many individuals can afford low-latency, 1.5-Mbps connections?

The good news is that the growth of the Internet and of IP convergence has driven an incredible investment in packet bandwidth, to the point where backbone bandwidth is almost free. It's free in the sense that many WAN services are now distance-insensitive. It's only the access bandwidth that you pay for. Yes, there is still an enormous investment required to bring high bandwidth to the access network, but that is happening as well, and already higher quality IP-based video conferencing is showing up in corporate networks.

BEYOND THE CONFERENCE ROOM
As Moore's Law continues to ensure that more MIPS are available to make the coders even better, and as higher bandwidth connections -- such as DSL and cable modems -- become available to the home, useful video conferencing over IP will reach the individual.

But will people really use video conferencing from their homes? The first application will likely be in business-to-consumer transactions. There's a great commercial interest in allowing individuals viewing a Web site to "click to talk" to an agent. Once the technology exists that gives an agent a high probability of carrying on an acceptable voice conversation with Internet-connected consumers, then half-duplex video is bound to follow. The call center agent doesn't need to see you, but a full-duplex voice connection together with a half-duplex video connection would allow you, the consumer, to see the agent in a corner of your screen while your conversation continues, perhaps while you're viewing specific Web pages pushed to you by the agent.

This service will become available even before QoS is more generally available on the public Internet. Already, there are specific service providers who offer overlay networks that enhance Internet content delivery on behalf of their customers' Web sites. A logical extension will be for these service providers to offer a private overlay network that call centers can use to bypass the public Internet as much as possible, providing a low-latency path to within a short distance of the actual consumer. Having a smiling, talking head on the screen that can help persuade consumers to make a purchase is very appealing to businesses that sell via call centers. The larger catalog companies will pay for this technology the moment they can reach a meaningful portion of their prospective customers -- perhaps 40 percent coverage would be enough to be interesting.

But what about full-duplex video to the desktop or the home? When are individuals going to routinely use this, in an individual context at work, or to call a friend or relative?

In addition to the QoS, latency, and bandwidth issues, there are also some human elements that need to be addressed before video conferencing to the home or desktop can approach voice telephony as a natural way to converse. For example, if the camera is on top of the monitor and your eyes are on the screen (looking at the image of the person you are talking with), you are not looking at the camera. As a result, you appear insincere. You're talking to the other party, but you're not looking them in the eye. There are academic projects working to solve this by video image processing that locates the eyes within the image and modifies their appearance, so that they appear to be looking at the camera. While a complex calculation, this may be easier than finding a way to bury the camera in the middle of the monitor.

In short, while small-room conferencing is a growing market today, and individual systems are becoming affordable, explosive growth of individual video telephony is still a few years off. Related applications, like call centers with full-duplex audio and half-duplex video, and totally new applications, will provide the next wave of growth for the video telephony industry.

BEYOND THE HORIZON
Then what? In the next decade or so, I expect to see a new Humphrey Bogart movie. The technology of morphing characters -- already used in the latest Hollywood films -- will allow a modern actor's appearance and voice to be morphed into that of Humphrey Bogart. We'll see a sequel to Casablanca. Once this ability exists on the silver screen, it will be a short interval until it's available on my PC screen. I'll be able to answer a video call from the beach, but appear as though I am working hard in my office.

Given Moore's law, within twenty years, three-dimensional holographic video imaging will be available. (Remember the projected image of Princess Leia in Star Wars?). Done properly, this will allow me to converse with someone who realistically appears to be sitting across the table from me, even though we are miles apart.

But back to reality. While latency and bandwidth remain issues today, video-over-IP technology has reached the point where room conferencing systems and meetings at a distance make sense for corporations, and over the next two to five years there will be rapid conversion of business video conferencing to video over IP. We will have to wait, however, for more bandwidth in the access network and some QoS in the public Internet before we get the explosion of video that I expect to see this decade.

Brough Turner is senior vice president of technology at NMS Communications. For more information, call NMS Communications at 508-271-1000, or visit the company's Web site at www.nmss.com. E-mail to the author (addressed to brough_turner@nmss.com) is also welcome.

[ Return To The July 2001 Table Of Contents ]