But Tombroff adds that interpreting data from different types of hardware isn't the heaviest lifting that Softkinetic does. The software's chief value, he says, is that it can "classify the scene so we know how to find the player and remove the rest, and reconstruct the person's structure." The first half of that task involves filtering out a great deal of noise from the signal. "We need to zoom in on the important thing, which, for video games, is you, the player, and not the person next to you sitting on the couch and making fun of you." Secondly, the software creates a 3-D volume from the fuzzy cloud of points the camera detects and identifies body parts important to an application. So instead of interacting directly with the depth map produced by the camera, designers get information from Softkinetic's software about which body parts are moving and how quickly. The company has also identified sets of gestures people commonly make when trying to control a program in a particular way.
Anind Dey, an assistant professor at the Human-Computer Interaction Institute, at Carnegie Mellon University, says that Softkinetic's technology is particularly exciting because of the potential for full-body interaction. While he notes that all software of this type must mediate between doing too much for developers, which can stifle their creativity, and doing too little, which can leave them to reinvent the wheel, he is enthusiastic about its prospects. "If the technology works as they're claiming it works," says Dey, "I think it's a really exciting thing for the field, and not just for gaming." For example, Dey says, knowing a person's body position could help with applications such as health-care monitoring in the home, or other applications in the field of ubiquitous computing.
Tombroff says that Softkinetic has built 12 or 15 sample games in-house and is now working with game developers to help them understand the technology and what they can do with it.