Developers from the United States have created a neural network that can recognize people's movements using radio wave data through a wall. She is able to identify, for example, a handshake.
An article about the development was published on the website of the Massachusetts Institute of Technology (MIT). It describes an algorithm that can analyze both video recordings and radio wave signals. First, it reads data of both types from a camera or radio transceiver, and then the neural network creates a skeletal body model from them. Then another algorithm analyzes the models on the frame, trying to determine their actions.
Radio wave scanning is carried out using a device with two antennas that emit the actual radio waves, and then receive them reflected from objects. First, two-dimensional images are formed from these signals, and only then a special neural network makes skeletal models.
A development by researchers at MIT has demonstrated an accuracy of up to 83% when scanning through a wall.
