(Note: This is a fictionalized account of how a journalist may have covered the paper, “Capturing the Human Figure Through a Wall” by Fadel Adib, Chen-Yu Hsu, Hongzi Mao, Dina Katabi and Fredo Durand. None of the quotations below are true and must be interpreted as fiction only.)
Massachusetts Institute of Technology
Imagine being able to interact with your Kinect placed in your living room from the comfort of your bed, or imagine dodging enemies in a virtual reality game as you physically move around your house. While these ideas might seem slightly out of reach given current technology, recent research at MIT may turn this into reality very soon.
The team has developed RF-Capture, a radio-based prototype that is able to capture the coarse skeleton of a human walking towards it on the other side of a wall. “This is extremely exciting from the perspective of applications such as virtual reality gaming,” says the lead author of the study. “Current such systems, like Kinect, are based on optical and infrared cameras. This physically constrains the user to an area directly in sight of the system in order to interact with it using hand gestures. However, our prototype breaks that barrier by using radio signals to form coarse images of the interacting user. So, with our system, a user in a different room at the opposite end of their home may still be able to interact with their Kinect.”
Their research builds upon their previous work on device-free through-the-wall localization. RF-Capture, like their previous prototypes, is also based on multiple antennas that transmit and receive specialized radio signals. These signals reflect off the environment before being received back at the receive antennas. The special structure of these signals enable the researchers to calculate the depth of the target from the reflected signals, and the multiple received signals allow the researchers to pinpoint the exact 3D locations of the target. Similar to their prior work, user motion is used to separate human targets from the static environment.
“To enable the signals to pass through walls and other occlusions, we use relatively low system frequencies between 5-8 GHz,” says the lead author. “These low frequencies allow us to reduce the cost of our system compared to those with, say, mm-wave or terahertz components. However, at such low frequencies, the human body acts as a reflector rather than as a scatterer. Therefore, only a few signals that fall near the surface normals to the body get reflected back to the system at any given time. To ensure full coverage, we use the fact that as the person moves towards our system, different parts of the body reflect the signals back. Thus, we form multiple images as the person walks towards the system and stitch them all together to form the final skeletal structure.”
Similar to their previous work, RF-Capture is also capable of tracking user movements, such as hand gestures. “Currently, we are only able to track hand gestures, but we would like to extend our system to track a variety of other complicated limb movements as well. What’s exciting though are the possibilities our work opens up in enabling faster, more seamless human-computer interaction,” says the senior researcher associated with the project. “We look forward to pushing the limits of what we can achieve with low-cost technologies like ours to bring our vision to reality.”
###
