JUN 22, 2018

AI to Monitor Patients, Track Movements Through Walls

WRITTEN BY: Julia Travers

A team of researchers from MIT’s Computer Science and Artificial Intelligence Laboratory, or CSAIL, is refining a system that can track people’s movements through walls and partitions. They say it could one day be used in medical fields for patient monitoring. The system named “RF-Pose” – “RF” referring to radio frequencies -- uses wireless signals to monitor human movements and then represents them in a visual interface as stick figures.

“We’ve seen that monitoring patients’ walking speed and ability to do basic activities on their own gives health care providers a window into their lives that they [doctors] didn’t have before, which could be meaningful for a whole range of diseases,” CSAIL Professor Dina Katabi, who is the project leader and co-author on the related paper titled, “Through-Wall Human Pose Estimation Using Radio Signals,” told MIT News in June 2018.

The entire project is based on the fact that “wireless signals in the Wi-Fi frequencies traverse walls and reflect off the human body.” When the team sent radio frequencies at the wall, they passed through the barrier or occlusion and bounced back from the person’s form. The AI extracts a skeleton image from the returning waves and then represents the person’s pose as a stick figure. The person might be shown to be sitting, standing or walking, for example.

“By using this combination of visual data and AI to see through walls, we can enable better scene understanding and smarter environments to live safer, more productive lives,” Ph.D. student Hang Zhao said.

In order to train the AI system’s neural network, or artificial brain, the team first had to gather data with which to teach it. Using a camera and wireless device, the researchers documented examples of people in numerous positions and activities, such as opening doors, talking and waiting for elevators. Camera images were used to create stick figures, which were matched up with corresponding radio signals. By studying this input, RF-Pose learned to read wireless reflections, estimate a person's motion and create a corresponding stick-person-image showing a 2D pose. The material collected for this endeavor was gathered with the subjects’ consent, anonymized and encrypted.

“A key advantage of our approach is that patients do not have to wear sensors or remember to charge their devices,” Katabi said.

Along with helping medical professionals watch patients, the tech might help to find people during rescue missions.

“Estimating the human pose is an important task in computer vision with applications in surveillance, activity recognition, gaming, etc.” the paper states. Whenever surveillance tech takes a leap forward, questions of personal privacy and freedoms are bound to arise as well.

CSAIL is an MIT lab that focuses on innovative computing projects. It has taken part in developing the World Wide Web, Ethernet, RSA encryption and parallel computing, among other significant breakthroughs.

View RF-Pose at work below.

Sources:

CSAIL

MIT News