A while ago Tsahi forwarded me an e-mail from Vidtel’s CEO, Scott Wharton, raising an interesting issue:
“I was at a conference the other night and I got a question from a media software engineer who is a heavy user of Skype.
He made the comment that he thought the number one complaint about video conferencing was the weird view of a person’s eyes, that caused them to be looking up (or down sometimes) but not right into the eyes of the caller.
He said all of his friends use Skype but get annoyed by this problem and then give up.
We then got into a debate about whether this was really that big a deal.
What do you think?”
Video calling using Skype. What are they looking at?! (CC Jon Ovington)
Well, I addressed the issue of eye contact, or gaze, before in my blog. There I mentioned that research has shown that gaze is one of the most important non-verbal cues.
But we’re not just talking eye contact here, but the general feel that people get from a remote camera. Tsahi addressed this one a few months ago when he discussed the rumors around iPad/iPhone front-facing camera.
Still I feel that the issue raised by Scott is very relevant, especially with the proliferation of video calling by Skype and their likes, So I decided I will dedicate an “Ask The EXPERT” post to this issue, and asked my expert on this – my colleague, Einat Yellin, who manages the 3DPresence project here at RADVISION, – - -
In Short
Scott’s engineer friend is not alone, says Einat – if you check the Wikipedia page for Video Conferencing, the number one problem mentioned is eye contact. Number two is appearance consciousness, and I mention it because I’ll get to it shortly.
Maintaining eye contact and correct spatial placement is very important, as this is what creates credibility and intimacy, and also allows people to have a conversation with multiple participants at the same time. Mistakes in eye contact comprehension can lead to grave consequences, and that uncomfortable feeling that can limit the use of the medium as a means of communications.
Maintaining eye contact means to convey the “feeling of being looked at” to a remote conferee. But when one looks directly into the camera on one side, on the other side everybody feel looked at. Alternatively, when you look past the camera, everybody will fill as though you are looking at a point next to them.
Most high-end video conferencing systems claim to deliver a “natural”, “true” or even “superior” eye contact experience, and yet they fail to deliver on their promise. This is because maintaining eye contact involves either complicated hardware designs (like putting the camera inside the display) or complicated algorithms (for fixing the view point).
Take the images above as an example. They were taken in a Telepresence room. The guy on the right screen is talking to someone. If you’re sitting on the left side of the room (the image on the left), it seems that the guy is talking to someone on the right. But if you’re the guy on the right (the image on the right), it still seems as though the guy is talking to someone else.
So yes – eye contact is a problem that needs to be solved, not only on your personal computer but in the video conferencing industry in general. BUT – and that’s a big but – as someone who uses video calling all the time, I must say that if you bother to place the camera on top of the screen, and if you invest a little time before the call to make sure that you are viewed correctly by the camera (not looking up or down), most of the annoyances will disappear, especially in a point to point call.
And anyway – what’s the alternative? If you’re not using video at all, there are no non-verbal cues at all. So it’s not a real solution, is it?
People used to complain (some still do) that the camera makes them over conscious to their appearance. And yet if you spend some time (not too much) doing video calls you will see how quickly you get used to it and forget all about that over consciousness.
Same implies for the eye contact problem. Just go ahead and use video – it will just stop bothering you after a while.
The Long Story
Maintaining eye contact is a very interesting topic. In fact, one of the goals of our 3DPresence project was solving the eye contact problem both horizontally and vertically, and so we devoted a lot of effort to study the issue.
Eye contact usually involves a “looker” who fixates on different targets that are offset relative to the center of his face (the bridge of the nose, for instance). Previous research on human sensitivity to eye contact is very limited.
In a famous research by Milton Chen, he found that a 7º offset below the camera center, eye contact is still perceived in most cases. He therefore suggested that a camera will be placed above the display, and it will support eye contact when the vertical angle between the camera and the eyes of the displayed person is smaller than 5 º.
In the project a research was conducted by the University of Eindhoven (TU/e) and we found that the disparity should be at most 1.2 º in the horizontal direction and 1.7 º in the vertical direction (at a 2m viewing distance). More so, we found that the camera should not be placed above the display, since the angle between the displayed person eyes and the camera will be too wide. This obviously poses a challenge to most commercial solutions.
Any solution to the problem involves “tricking” the view point, so that it will seem that the camera is indeed behind the display. This can be done in a few ways:
- Place the camera in the display -very complicated from an engineering/optical point of view. You need the display to not blind the camera, the camera lens to not hide the display, etc. Apple actually filed a patent on this idea a while ago, but there aren’t any commercial solutions that incorporate this method yet.
- Using optical tricks – Half transparent displays with camera behind them, mirrors that make the picture look as if it was taken from a different angle, etc. Errol Morris, the famous film maker, used such “tricks” in his personal interviews.
- Using 3D modeling and correcting the viewpoint to achieve eye contact – This is what the 3DPresence project is doing. The viewpoint is corrected both vertically and horizontally, which is very complicated, so that eye contact is maintained in any angle.
Using mirrors to maintain eye contact using a webcam. Source: Aram Bartholl
Bottom Line
I believe that just like our industry solved the problem of life-size images (with HD resolution) and network resilience (with tools like our SVC solution), eye contact will be eventually solved in commercial video conferencing products.
In the mean time, I think video calling is always better than audio only. So be aware of the eye contact problem, try and setup your camera to the best setting possible, and keep using video.
