A team at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) announces The RF Pose, the technology that teaches wireless devices to sense movements and postures of people on the other side of the wall.
The team of researchers led by Professor Dina Katabi has developed a system that uses a neural network to analyze radio signals that bounce off people’s bodies creating a stick figure of that individual, and this graphical representation lively depicts the individual’s posture or actions.
RF-Pose technology can be of great benefit to physicians providing medical care to patients suffering from diseases like Parkinson ’s, Multiple Sclerosis and similar ailments. RF-Pose helps practitioners gain access to real-time conditions of these patients and thereby assist the sick by providing better treatment. It could even serve the elderly with added security while they live independently.
One of the key advantages of this project is “freedom”. Individuals aren’t forced to put on any wearables or charge their devices. For future applications, Katabi speaks about their plans to include “consent mechanism” which would need the device owners to do a specific set of actions, so as to record and feed the machine-learning algorithms to better understand the “normal” user.
Neural networks are usually trained using data labeled by hand. And so, to develop a neural network that mapped radio signals to stick figures was challenging. Besides benefitting the healthcare, the technology was 83% likely to identify 1 from a line of 100 individuals, and this is a search-and-rescue implication.