HomeTechnologyFutureMIT Develops AI That "Sees" People through Walls

MIT Develops AI That "Sees" People through Walls

Seeing through walls, as we know it nowadays, is typically associated with the use of special electromagnetic waves. This is the reason why in sci-fi we usually call the wide range of abilities that can achieve similar feats as X-ray vision.

But while today this requires the use of very specialized equipment, researchers at MIT have found another novel way to achieve this, with a project that uses a nifty combination of neural network AI and wireless signals from your own phone.

Stick figures seen through walls

The RF-Pose, as it is officially named, is an integrated system capable of generating live images of people by analyzing wireless RF (radio frequency) signals as they bounce off of people. This sounds relatively simple and even unimpressive at first, but when you realize that this project can actually generate the data live, you immediately understand how astonishing, or even scary, this new feature is.

According to the official blog post that introduced the project, the neural network was conditioned with a series of video feeds, with a corresponding RF map of the same area. As people get in and move about within the camera’s view range, the neural network AI was eventually able to collate adequate data that associated which movement corresponded to which change in the RF map. This then was used as the framework for RF-Pose.

When the researchers finally tested a more finalized version of the system, they were surprised that not only was RF-Pose able to see and detect people, it could even sense and calculate movement. A virtual stick figure was able to more or less match its movement to the human that it represents. Even more impressive, the system also works when several people are within the viewable area. RF-Pose will generate a figure for each of those people that walks and moves in the exact same approximation.

The advent of tin foil walls?

While dystopian thoughts of privacy invasion and government spying are abound with RF-Pose’s very concept, its researchers have actually built the system with healthcare in mind. The blog post has stated that RF-Pose could be an essential tool in monitoring diseases such as Parkinson’s, multiple sclerosis (MS), and muscular dystrophy. Progression of these diseases could be monitored by the second, allowing doctors to gather more data about its progress, as well as adjusting medications in more calculated timed intervals.

RF Pose different scenes - credit MIT CSAIL

Also, since it doesn’t require attaching tracking devices to monitor people, it could be a vital tool in search and rescue operations. People trapped under and/or inside debris can be detected early, and locating survivors can be made easier during times of natural calamities.

As of now, the developers of RF-Pose are currently working towards improving the system beyond simple stick and bubble features. In the future, the team expects an enhanced version using perhaps a complete 3D model, with even more accurate micromovements being represented by the figure.

YouTube: AI Senses People Through Walls

Photo credit: All media in this article are done by and owned by MIT CSAIL. The feature image photo was taken by Jason Dorfman.
Source: Adam Conner-Simons, Rachel Gordon (MIT News)

Was this post helpful?

Christian Crisostomo
Christian Crisostomo is an avid researcher of anything new in tech. Whether it's a spanking new game, a cool new gadget, or even some radical breakthrough research, he's always there to share.