Wireless smart home in the hottest computer scienc

  • Detail

The wireless smart home system of computer science and artificial intelligence laboratory can monitor diseases and help the elderly in place

the wireless smart home system of computer science and artificial intelligence laboratory can monitor diseases and help the elderly in place

17:30:30 source:

X-ray vision seems to be a far fetched science fiction fantasy for a long time, but in the past decade, The team led by Professor Dina katabi of the computer science and Artificial Intelligence Laboratory (CSAIL) at the Massachusetts Institute of technology keeps us closer to seeing through the wall

their latest project RF pose uses artificial intelligence (AI) to teach wireless devices to sense people's gestures and movements, even from the other side of the wall

researchers use neural networks to analyze radio signals bouncing back from the human body, and then can create a dynamic rod shape. When people perform these actions, it will walk, stop, sit down and move limbs

the team said that RF pose can be used to monitor diseases such as Parkinson's disease, multiple sclerosis (MS) and muscular atrophy, so as to better understand the progress of the disease and allow doctors to adjust drugs accordingly. It can also help older people live more independently, while providing additional safety monitoring for falls, injuries and changes in activity patterns. The team is currently working with doctors to explore the application of RF pose in health care

all data collected by the team are agreed by the subject and are anonymous and encrypted to protect user privacy. For future practical applications, they plan to implement a consent mechanism, in which the person installing the device needs to make a specific set of moves in order to start monitoring the environment

we have seen that monitoring the walking speed of patients and their ability to do basic activities provides health care providers with a life window they did not have before, which is meaningful for all kinds of diseases. Katabi co wrote a new paper on the project. A key advantage of our approach is that patients do not have to wear sensors or remember to charge their devices

in addition to health care, the team also said that RF pose can also be used in new video games, where players can move around their homes and even help find survivors in search and rescue missions

katabi wrote a new paper with Mingmin Zhao, a doctoral student and the main author, Antonio Torr ~ 0, a professor at MIT, testing 50MPa optional alba, Mohammad Abu alsheikh, a postdoctoral student, Tianhong Li, a graduate student, and Yonglong Tian and Zhang Zhao, doctoral students. They will deliver a speech at the computer vision and pattern recognition Conference (CVPR) held in Salt Lake City, Utah later this month when the prices of the cathode material ternary material and lithium cobalt oxide have risen sharply

a challenge that researchers must address is that most neural networks are trained using manually labeled data. For example, training to recognize the neural network of a cat requires people to view a large image data set and mark each image as cat or non cat. At the same time, radio signals are not easy to be marked by humans

to solve this problem, the researchers collected some examples using their wireless devices and cameras. They collected what thousands of people did, such as walking, talking, sitting, opening doors, waiting for elevators

then, they used these images from the camera to extract bar graphs, which were displayed to the neural network together with the corresponding radio signals. This example combination enables the system to learn the association between the radio signal and the simple strokes of the characters in the scene

after training, RF pose can only use the wireless reflection reflected from the human body to estimate a person's posture and movement without a camera

since the camera cannot be seen through the wall, the network has never explicitly trained the data on the other side of the wall - which surprised the MIT team in particular, the network can summarize its knowledge as being able to deal with wall movement in the following ways

if you think of a computer vision system as a teacher, this is a very interesting example of how students perform better than teachers, toralba said

in addition to sensing movement, the authors also showed that they could accurately identify 83% of the time in the lineup of 100 people using wireless signals. This capability is particularly useful for the application of search and rescue operations, as it may help to understand the identity of specific persons

for this article, the model outputs a two-dimensional bar graph, but the team is also committed to creating a three-dimensional representation that can reflect smaller fretting. For example, it may be able to see whether the old Dell launched the new xps13 on the eve of CES exhibition. The main types of shaking are cylindrical force sensor, spoke force sensor, s double hole sensor, 10 beam sensor and so on, so that they may want to check

by using the combination of visual data and artificial intelligence to perspective the wall, we can achieve better scene understanding and a more intelligent environment, so that they can live a safer and more productive life, Zhao said

Copyright © 2011 JIN SHI