In June 2018, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory published a system that uses ordinary wifi signals to generate a moving skeletal outline of a person on the other side of a wall, tracking head, elbow, knee, and other joint positions in real time, without a camera, a wearable sensor, or the subject’s knowledge. The project, called RF-Pose, was led by CSAIL professor Dina Katabi and built on nearly a decade of published work on radio-frequency sensing at MIT.
How RF-Pose Actually Works
The system transmits low-power wireless signals that pass through walls and reflect off a human body. A neural network, trained by matching those reflected signals against camera footage of people performing the same movements, learned to translate the radio reflections into a dynamic stick figure: a real-time skeletal model of a person walking, sitting, or moving their limbs, generated entirely from wifi bounce-back. Katabi’s team had demonstrated earlier, cruder versions of through-wall sensing as far back as 2013; RF-Pose was the point at which the output became precise enough to track individual joints rather than a blob-like silhouette.
The Stated Uses
MIT’s own published applications for RF-Pose are healthcare-focused: monitoring the gait and movement of patients with Parkinson’s disease, multiple sclerosis, or muscular dystrophy, without requiring a clinician or a wearable device. The team has also floated search-and-rescue use, locating survivors through rubble or walls, and gesture-based applications like gaming. The research was partially funded by the National Science Foundation and was demonstrated to President Obama at the White House in 2015.
What the Technology Doesn’t Require
What makes RF-Pose different from a security camera is precisely what makes it notable from a surveillance standpoint: it requires no visible hardware inside the room being monitored, no camera, no microphone, and no cooperation or awareness from the person being tracked. A transmitter on one side of a wall is sufficient. MIT’s published materials describe the team building in encryption and anonymization safeguards and a mechanism to let users block the signal, which is itself an acknowledgment that the underlying capability, tracking a specific person’s precise movements through a solid wall without their knowledge, was understood internally as something that needed a technical safeguard.
What Remains Undisclosed
What MIT has published is the capability itself, peer-reviewed and demonstrated publicly, including to a sitting president. What has not been published is any account of whether the underlying technique, or a derivative of it, has been licensed, adapted, or independently developed for use outside the healthcare and consumer applications MIT describes. A technology that can resolve a named individual’s joint-by-joint movement through a wall, using nothing but ambient wifi, is dual-use by its nature, and the gap between a published academic capability and any private or governmental use of the same underlying method is not something this research, on its own, answers either way.
Leave a Reply