Artificial intelligence recognizes the person through the wall
Scientists have created the technology of RF-Pose, which can determine the posture and human movement behind the wall.
Technology can be useful in rescue work, and also for monitoring patients suffering from Parkinson’s disease or multiple sclerosis. Briefly about the development of reports TechXplore, full version will be presented at the conference on computer vision and device recognition (CVPR 2018), which will begin June 18 in salt lake city.
Today to find people trapped under the rubble or trapped in an avalanche, use special devices, such as radars (the radars). By studying the data on the reflected signal, rescuers can determine the location of the person and help him. However, the radar did not always work for sure: their work could be improved thanks to the technology of machine learning.
Researchers from the Massachusetts Institute of technology (MIT) under the guidance of Professor Diana Kataby (Dina Katabi) Lab computer science and artificial intelligence (CSAIL) have developed technology of RF-Pose that combines artificial intelligence and a wireless radio. It is designed to determine the position of people in space. First, engineers with the camera watched as participated in the experiment, the volunteers say, walk, sit, open doors, or waiting for the Elevator. Scientists have collected several thousand images. Based on images they have built for each individual skeletal model, and then demonstrated her computer along with the corresponding radio signal. Thanks to the combination of data, the AI system learned how to define the relationship between a posture and movement model and data radios.
Initially, the researchers taught the software to determine the position of participants in the zone of visibility, as the camera can’t “see” the man behind the obstacle. However, the system has compiled the data and were able to track the poses of the people who were behind the wall. “If you imagine that computer vision is a teacher, you stunning example of how well the student surpassed the teacher,” says MIT Professor Antonio Torralba (Antonio Torralba).