Recovering sound from video ... Freaky!

[mamalala]

Hey all,

dunno if that is the right section of the forum (it says "technical stuff", after all), but check out his:



That is ... uhh ... well ... On one hand it is fantastic, that this can even be achieved ... On the other hand this is somewhat frightening. And goes to show why you better use a CCD cam instead of a CMOS, in case of the low-framerate-thingy

Greetings,

Chris

[Dongulus]

I'm impressed that they actually got it to work. I think it's a neat exercise in video processing, but I doubt it actually brings any implications to be concerned about. The camera outside of the room capturing sound appears to be most potentially threatening because it could be done remotely. However, I question whether the detail required to extract sound from any vibrating object could be captured from any range far enough away to be inconspicuous.

[corrado33]

That is very interesting, but I don't see a huge application for it (not that it needs one.) Old silent video is too slow to have this technique work, and now-a-day it's much easier to hide a microphone in something if you really want to spy.

Then again, it could be used for applications where the room is swept for "bugs" or if somehow it's in a faraday cage.

Cool programming nonetheless.

[JoeO]

The Zapruder Film?

[Tinkerer]

That is impressive.

The Zapruder Film?

And you know what, this would be an excellent technique for trying to recover sound from old videos, assuming they arnt super low frame rate.

[max_torque]

Whilst the image data processing and manipulations/filtering is pretty interesting, "remotely" intercepting sounds is hardly a new thing, with "laser" microphones having been around for ages!

[T3sl4co1l]

The nice thing about remote acoustics (like the laser microphone) is, while possible, they're actually really hard to use properly, and/or conspicuous or expensive (not that cost is any problem to a well financed adversary).

A high speed camera with a telephoto lens (rather conspicuous if not well hidden) could prove much more useful than the old laser method though.

Tim

[Chipset]

Keep in mind that the impracticalities of this method that we're currently discussing are only as part of the test setup. Of course hiding a high speed camera is much more impractical than just planting a microphone if you just wanna listen in, the point is this could be applied to surveillance camera footage, cell phone footage and home video footage. It opens up a whole new avenue of data harvesting that has some quite serious privacy implications.

[peter.mitchell]

When i saw the title, i immediately thought of this; http://www.windytan.com/2014/02/mystery-signal-from-helicopter.html
Completely different in all ways, but at the same time, still interesting.

[ivaylo]

I listened on the radio when they announced this (NPR's Science Friday I believe) and the host made a joke something to the effect of "... and one day we could listen to what Jackie and John are talking about behind the curtain on those old reels...", and the guest (one of the inventors) said, absolutely... Not sure he was joking or not. One thing is using a high speed camera, another crappy 25fps old footage, even with this second technique they mentioned going above the frequency of the camera frames. But let's wait and see I guess...

[HighVoltage]

A while back I saw a report on German TV about this subject and they used it to recover the voice of Hitler on so many silent films that where taken of him.
They used a combination of lip analysis and frequency analysis of nearby objects.
And they proved that the technology was working, by testing it on film footage that they had the sound for.
It was very impressive.

[Mechatrommer]

the test setup was using loudspeaker or significant volume to subject matter. do you think terrorists will pass vocal info in loudspeaker? not to mention if surrounding is busy traffic or other dominant sound sources. good luck, nothing is impossible. just damn hard and expensive.