Researchers at The Massachusetts Institute of Technology have succeeded. They developed artificial intelligence software that can be looked through walls. How is that even possible?
The system uses radio waves that reflect people and also pass through walls. A neural network is trained to make simple dynamic skeletons of seated, walking and talking people. By teaching the system how these skeletons are connected to radio signals, the posture and movement of a person can be generated. Even if the person is behind a wall.
How did they learn this from the system? At first thousands of images of human activities have been recorded, such as sitting, walking, talking and opening doors. The researchers used a wireless device to rebound radio waves during these activities. The system learned which skeleton belonged to which radio wave. Thus, a radio wave could be connected to each type of movement and the system could gauge movements and body posture. This is so accurate that it works even when a wall is in between and if there are several people in a room.
The system can also identify and distinguish individuals. If there are 100 people in a row, the system will be able to identify who is who with 83% accuracy.
Application in health care
This technique can for example be used well in health care. Movements and motor skills of people with Parkinson’s disease can be well visualized in this way. The doctors can also monitor patients from outside, so that they can lead a more independent life. The system can also be used for games, or to track people in disaster areas.
A not very surprising issue is the privacy of people. Spying on this AI is not very difficult. According to the researchers, the system can only be used with permission. How exactly this is worked out is not yet known. Risks will always be there, but there are also a lot of opportunities for this.