Judging by the video they achieve 83% success rate with a pool of 8 people in what looks like perfect conditions (only one person, no perturbation of any kind etc...). The video also states, rather suspiciously IMO, that "This scene on the other side is shown solely for presentation and was not used for identification". Why not show the actual footage of the experiment?
>The lab has tested their new technology on 1,488 WiFi-video pairs, drawn from a pool of eight people, and in three different behind-wall areas, and achieved an overall accuracy of 84% in correctly identifying the person behind the wall.
What does it mean exactly? They only need one short video and one short wifi capture to get 84% success rate? That's what the video seems to imply but I find that very hard to believe. Or maybe it's just because it's fairly easy to distinguish among 8 people (especially if they have significantly different body types) and it won't work quite as well at large. I can identify my girlfriend's footsteps in the staircase with remarkable accuracy but I can guarantee you that it won't scale to the general population.
Maybe it works better than I give it credit for but they need to bring up better evidence IMO.
>What does it mean exactly? They only need one short video and one short wifi capture to get 84% success rate?
Key aspects upon this are that it works by matching up their style of walking (gate), so the person needs to be moving not only in the wifi data capture, but also the video they are matching with.
They use the video to extrapolate a 3D mesh model of that person walking. Then model a wifi signal upon that model and produce a signature for that person that they can then use to match up with what the wifi signal is picking up.
Whilst measuring some bodies way of walking has been done many times before, this add's another way of doing that - but without the need for a camera feed.
So with this approach - no they can't produce a video image of the scene behind the wall from the wifi signal(s).
Remember - individuals have many unique traits, many of which we havn't tapped into yet, though as far as walking styles go - that has been done for years, this just creates another way of measuring those and matching up with other forms of data, video in this case. Other things that could be explioted this way would be for example sound - a contact mic array would allow you to (thru a wall) measure a person's heartbeat to some degree and match that up to an individual. Privacy wise, this may well traction well compared to other forms of surveillance.
> "This scene on the other side is shown solely for presentation and was not used for identification". Why not show the actual footage of the experiment?
My understanding is that the video on the left (the one showing the other side of the wall) is just here to illustrate what's the wifi setup senses and was not used for identification as the whole point is to identify without (optically) seeing what's on the other side.
>The lab has tested their new technology on 1,488 WiFi-video pairs, drawn from a pool of eight people, and in three different behind-wall areas, and achieved an overall accuracy of 84% in correctly identifying the person behind the wall.
What does it mean exactly? They only need one short video and one short wifi capture to get 84% success rate? That's what the video seems to imply but I find that very hard to believe. Or maybe it's just because it's fairly easy to distinguish among 8 people (especially if they have significantly different body types) and it won't work quite as well at large. I can identify my girlfriend's footsteps in the staircase with remarkable accuracy but I can guarantee you that it won't scale to the general population.
Maybe it works better than I give it credit for but they need to bring up better evidence IMO.