You wave your arm while in the shower to lower the music volume, or you make a hand gesture in the air from bed to turn up the thermostat. That vision of a gesture-anywhere life may have taken a step closer toward reality as a result of newly revealed research on using household Wi-Fi signals instead of a camera for gestural interaction.
The research, conducted by a team at the University of Washington's computer science department, has been submitted to the 19th Annual International Conference on Mobile Computing and Networking, scheduled to take place in the fall in Miami. The research team calls the system WiSee, and it leverages Wi-Fi signals to read body movements without special sensors or cameras, such as in Microsoft's Kinect system.
Instead, WiSee utilizes an adapted router and some wireless devices connected to household devices and appliances. Lead researcher Shyam Gollakota, an assistant professor at UW in computer science and engineering, said in a statement that WiSee repurposes wireless signals "that already exist in new ways."
The team said that, while the concept is similar to Kinect and other gestural input system, it is simpler, cheaper, does not need cameras or distributed sensors, and does not require users to be in the same room as the device they're controlling, since Wi-Fi signals can travel through walls.
The system involves an intelligent receiver, which could be an adapted Wi-Fi router, that monitors all wireless transmissions from smartphones, laptops, tablets and other devices in the home. Movement by a person in this kind of field creates a slight change in the wireless signal frequencies, similar to the Doppler effect, a change in the perceived frequency of electromagnetic waves based on the positions of the source and an observer.
The resulting changes are on the order of a few hertz in Wi-Fi signals that operate at 5 gigahertz and have a bandwidth of about 20 MHz. The receiver's software can detect those tiny shifts, as well as account for devices that have stopped transmitting, such as a smartphone that's been turned off.
The software is currently designed to recognize nine body gestures, including pushing, pulling, punching and full-body bowling. On tests with five users in a two-bedroom apartment and in an office environment, the WiSee system could accurately recognize 94 percent of the gestures. The receiver has multiple antennae, each turned to a specific user's movements, to allow up to five users to perform gestural commands in the same home. The researchers plan next to work on the ability to control multiple devices with one gesture.
'Complementary' to Kinect
In order to avoid random gestures that the receiver might interpret as commands, WiSee would require a gesture sequence prior to a command, the equivalent of the Star Trek crew saying "computer" before they give a command to their all-knowing system. Additionally, a specific gesture could be programmed to refer to a specific device, such as an up-and-down arm motion indicating that the volume on the main sound system should be lowered or raised.
Ross Rubin, principal analyst for Reticle Research, said the system "sounds complementary to systems like Kinect." He noted there are two trends in gestural interaction -- less expensive systems and more precise ones. Leap Motion, for instance, promises highly accurate movement detection, and the new, high resolution Kinect on the just-unveiled Xbox One supposedly can detect and measure a user's heart rate.
The WiSee system, Rubin said, appears to fall into the "less expensive" category. This system could "enable gestural control where it is not feasible today," he noted, and other systems could then provide higher levels of precision when needed for gaming and other specific applications.
Posted: 2013-06-14 @ 12:48pm PT
I love the idea of being able to do this purely for the fact that that means we are still advancing to what can be out there. As a scientist myself I look forward to seeing how far we can advance.