|
Saw that on Hackaday a while back. Blew my mind when I watched the video!
Had to read the research paper to see how the heck it worked. I think I understand it well enough I could code a replication, at least a primitive one.
It only works with things that oscillate, and the approximate frequency must be known in advance.
Consider a tuned circuit. You apply the resonant frequency, it builds up and is amplified. Here they use an equivalent digital filter, with a separate one attached to every pixel in the video.
If that baby's heart were to abruptly stop (horror!), the video would actually show it continuing to beat, with the strength gradually dropping off. And if it were to restart, it would take time to register that too.
All the same rules that apply to tuned circuits and filters apply here. The more amplification is required and/or the lower the frequency, the slower the filter responds. Multiple filters can be attached to each pixel to detect a wider range of frequencies, with each filter set to a different center frequency. Better frequency resolution can be achieved by using more filters, and increasing the selectivity of each.
Another caveat - it only works on fairly stationary objects. If moving too much, none of the filters attached to any pixel will have time to respond.
A lot of folks were wondering if it could be used as a lie detector for people on TV, in particular for presidential debates. I don't think it would work, all TV is digital and compressed now, and compression tends to remove the fine temporal details needed. A shame analog TV isn't around anymore!
Can't think of any way to use it to detect CO2 levels. Even if a camera could somehow pick up a hint of CO2 concentrations, this technique would only amplify oscillations, and the day-to-day rise/fall of CO2 wouldn't be particularly useful.
|
Login
Register
FAQ
Search
