By now, even if they’ve never used one themselves, most people are pretty familiar with the idea behind gesture control systems such as the Kinect – the user makes a movement, the device “sees” that movement, and interprets it. However, what would happen if the user was in another room, blocked from the device’s cameras and depth sensors? Well, as long as there was a Wi-Fi signal available, it wouldn’t be a problem ... at least, not if the WiSee system was being used.
Created by a University of Washington team led by Prof. Shyam Gollakota, the technology utilizes nothing but the Wi-Fi signals already present in a room, such as those emitted by smartphones or laptops. Those signals travel through the interior walls of the building, and are monitored by a centrally-located WiSee receiver – although a regular Wi-Fi router could also be adapted for the same purpose.
NEW ATLAS NEEDS YOUR SUPPORT
Upgrade to a Plus subscription today, and read the site without ads.
It's just US$19 a year.UPGRADE NOW
Should the user move even just a hand or a foot, they will cause a change in the frequency of the Wi-Fi signal, known as the Doppler frequency shift. Even though that shift is typically only in the order of several hertz, an algorithm created by the researchers allows the receiver to detect it.
Different types of movements produce different shift signatures, with the system currently able to differentiate between nine unique whole-body gestures. Multiple users can be tracked simultaneously, too, as the each of the receiver’s multiple antennas automatically tunes itself in to a specific source. In its current form, a single WiSee device can follow the movements of as many as five users without getting “confused.”
The upshot is that a WiSee user could (for example) be sitting in the bathroom, and wave their arm in order to turn down the stereo in the living room ... if the receiver had already been programmed to pair that gesture to that function.
So far, the system has been tested in a two-bedroom apartment and an office setting, with five users. It was able to correctly identify 94 percent of the 900 gestures made by those people.
Source: University of WashingtonView gallery - 2 images