It looks like a small white canvas hanging on your wall. There are no bright screens, no flashing lights. And yet, using radio waves with 1/1000th the power of your wi-fi signal, it can peer through your walls and track your movements–all without literally seeing you, as an invasive camera might.
Called WiGait, it’s the result of research led by MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), a combination of hardware and software algorithms that have been in iterative development since 2015.
“The vision, really, is that the future of homes will be health aware,” says professor Dina Katabi. “Today you have a smoke detector in your home, but you don’t have a health emergency detector. You really want that. That’s more important.”
Katabi’s team is starting by measuring a metric most of us might not think about, but that doctors have called “the sixth vital sign:” Gait velocity. The simple measurement of how fast you walk, coupled with your stride length, gives doctors a tangible measure of elderly mobility, along with complications stemming from diseases like Parkinson’s and Alzheimer’s. Gait velocity can be the canary in the coal mine for decreasing cardiac and cognitive health, but the problem is, it’s measured by a doctor watching the patient walk with a stopwatch in hand–and only every few months or each year a patient might visit the office.
There have been attempts to decouple elderly mobility from the doctor’s office, of course. But they all come with trade-offs in accuracy or domestic UX. As PhD student Zachary Kabelac explains, the Fitbit measures footfalls, but it can’t see stride length–which is guesstimated by someone’s age and height (essentially negating the Fitbit’s ability to track gait velocity). Meanwhile, Microsoft and others have experimented with the Kinect depth-sensing camera as a means to track the movement of someone around their home. But the Kinect is a camera–one that can take deep and intimate images of your body–which has historically led to a lot of privacy concerns and did with MIT’s polled subjects as well.
“We don’t create an image,” says Katabi of WiGait. “One of the main things we’re trying to avoid is being a camera or Kinect, and having a minimal amount of information. It’s not an image. You don’t get to see the person or discover something that could be personal. You can just track the movement–and even track it through walls.”
The WiGait essentially gets a deep but fuzzy portrait of your activity through hard measured numbers–not a portrait made of pixels–interpreted by an algorithm that uses this data to identify and quantify movements. Now, that algorithm can still conclude a lot, with the ability to distinguish between how you put away clothing and how you walk around a room, but it doesn’t actually have the fidelity to create clear pictures, and so it abstracts your body as a single locational dot during its own data collection process.
“Of course you can imagine doing [all sorts of] things with cameras and Kinect, but when we started talking to people about putting a Kinect in their home, hardly anyone accepted that,” says Katabi–adding later that Amazon’s new home camera was “alarming.”
But even without full vision, WiGait can track gait velocity with an 88.4% to 99.3% accuracy, following up to four people at a time. And it can even recognize when someone falls–giving it great potential as an emergency response system for a senior living alone.
The researchers plan to keep studying the technology with doctors and patients. However, in the future, there’s very little standing in the way of this domestic AI that tracks personal health non-invasively–and without trying to sell you something. As Katabi put it to me, “This is basically a radio. Radios are typically cheap. And if it’s mass produced, of course it would be the same cost as a consumer device.”