AI Can See What’s Other Side Of The Wall

X-ray vision has long been an extraordinary sci-fi fantasy, but in the last decade, a team led by Professor Dina Katabi, Computer Science, and the Artificial Intelligence Laboratory (CSAIL) MIT has certainly brought us nearby to the concept of “see through the walls.”

Their trending project, RF Pose, deploys artificial intelligence to train wireless devices to control the poses and movements of people, even on the other side of the wall.

Researchers use a neural network to analyze the radio signals that are reflected from people’s bodies, and then they can create a dynamic figure that depicts actions such as stop, walk, sit and move their limbs of the person who performs activities.

The team says that RF-Pose can be used to control diseases such as Parkinson’s disease, multiple sclerosis (MS) and muscular dystrophy, which helps to better understand the progression of the disease and allow doctors to adjust medications appropriately. It can also help older people to live more independently while providing additional protection for injuries and changes in activity structures. Currently, the team works with doctors to study RF-Pose application in health care.

All data collected by the team has the consent of the subjects and have been anonymized and encrypted to protect the privacy of users. To implement future applications in the real world, they plan an “agreement mechanism” when the person installing the device must perform a specific set of actions to monitor the environment.

In addition to health, the team states that RF Pose can be applied for new kinds of video games in which players were in motion in the house or even search and rescue missions to help find survivors.

One of the problems faced by researchers is that most neural networks are trained using data with manual dialing. Meanwhile, people cannot easily identify and label by radio signals.

Earlier, the team lead by Katabi had also revealed a mechanism that detects our walking speed with the help of wireless signals.