MIT researchers create AI that lets smart homes identify different people.
The past two decades of technology have dramatically changed the way people live. As homes become “smart”—approximately 27 percent of U.S. homes have at least three smart home devices, according to a GfK study—researchers at MIT are taking the technology to the next level.
In a recent published paper, the research team shared its new system, which is capable of differentiating between occupants, even when a person isn’t carrying a mobile device. Dubbed Duet, the system locates individuals through reflected wireless signals and uses algorithms to ping nearby devices. It then uses those pings to determine who used the system last and then predicts their movements.
“Smart homes are still based on explicit input from apps or telling Alexa to do something. Ideally, we want homes to be more reactive to what we do, to adapt to us,” said Deepak Vasisht, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) doctoral student. “If you enable location awareness and identification awareness for smart homes, you could do this automatically. Your home knows it’s you walking, and where you’re walking, and it can update itself.”
Although systems are available that can locate people when they are carrying a mobile device, the MIT team realized that when at home, people aren’t always carrying a device. Building upon their own device-based localization system, which is cable of tracking individuals within 10cm, the researchers added a device-free tracking system, WiTrack, with algorithms that use logical reasoning. The system incorporates a floor map and can recognize entrances and exits in each room of a home. It then determines the likely profile of who is within the boundaries by measuring the reflections of wireless signals through a process of elimination, which also accounts for blocked signals and when people are in a similar trajectory.
The team’s two-week experiment involved four people in a two-bedroom apartment and nine people in an office. The Duet wireless sensor was installed on a wall that was about a foot and a half squared. The results were 96 percent and 94 percent accurate, respectively, even when a person wasn’t carrying a mobile device.
The team’s next goal is to expand Duet for long-term use and in more spaces. They believe the system’s potential benefits could include recognizing intruders, monitoring health-care issues, and providing responsive smart home features, such as instantly setting a preferred temperature when someone enters a room or turning on the TV to a person’s favorite channel when they sit on the couch. The researchers also have discussed developing an app that would either grant or revoke Duet’s access.
Interested in more “smart” innovations? Check out Smart Home Sensor Sees Through Walls and No More “Nuisance Trips” with Smart Power Outlets for Smart Homes.