- Tech that converts human figures into stick figures.
- Analyses radio signals that come from human body.
- Researchers in MIT developed their own training data using wireless devices and cameras and put together a database of thousands of images.
We always knew superman could see bad guys through walls and catch a hold of them. That might not be fiction after all!
Researchers from MIT’s Computer Vision and Artificial Intelligence lab have have built a neural network that estimates the pose and movement of people behind a wall. This project is called RF-pose.
How it works:
The neural network has been built in a way that it can sense and analyse radio signals that come from a person’s body. It then creates a digital stick figure to show where the person is and what pose they are currently in. (standing, sitting or moving around)
Most neural networks work on supervised learning methods and require a huge amount of data to be trained properly. Hence, a significant challenge was finding labelled training data, which was hard to find. So the researchers developed their own training data using wireless devices and cameras. They put together a database which has thousands of images of people doing various activities like standing, sitting, moving etc.
The next step involves extracting stick figures from the camera images which were shown to the neural network, along with radio signals corresponding with them. After the training, the network was able to estimate the pose and movement without the use of the camera.
The network was able to identify people in a line-up with an accuracy of upto 83%.
Listed below are a couple of links below that provide an in depth understanding of the subject.
So let’s summarize:
The applications of this technology are numerous and are not limited to just a particular field. It can be used in healthcare to maybe detect where a cancer or a tumor is located, or even in tactical missions by the army to have a count of how many terrorists are located within a building (just like superman). There is huge scope of development in this field and this is just the beginning!