In one set of experiments, they were able to recover intelligible speech from the vibrations of a potato-chip bag photographed from 15 feet away through soundproof glass.
In other experiments, they extracted useful audio signals from videos of aluminum foil, the surface of a glass of water, and even the leaves of a potted plant. The researchers will present their findings in a paper at this year's Siggraph, the premier computer graphics conference.
"When sound hits an object, it causes the object to vibrate," says Abe Davis, a graduate student in electrical engineering and computer science at MIT and first author on the new paper. "The motion of this vibration creates a very subtle visual signal that's usually invisible to the naked eye. People didn't realize that this information was there."
Abstract (full text available as PDF): http://dl.acm.org/citation.cfm?doid=2601097.2601119
(Score: 5, Insightful) by bob_super on Monday August 04 2014, @10:35PM
Researchers: Point your super-high-speed camera at the right angle, suspend the sheet such that it may vibrate nicely, and you get some garbled output that you may recognize under ideal conditions...
CSI writer: So the convenience store camera across the street captured the vibration of the [insert ad] poster in his office on the 10th floor, and our vocal recognition database says the it's clearly Ricky, under a tremendous amount of stress, drinking [ad] and telling Nancy how his software stole 12.758 millions, and therefore he implies Samantha is a problem, and there's that [other ad] beach party next week...
(Score: 3, Interesting) by jcross on Tuesday August 05 2014, @12:34AM
You make a good point, but I'd like to point out that in section 6 of the paper, they do show how you can use the rolling shutter properties of a normal video camera to recover sound. Not great sound, with an upper limit of 2 kHz or so, but pretty neat that it's possible at all.
(Score: 2) by JoeMerchant on Tuesday August 05 2014, @02:42AM
Doesn't take too much math or tech knowledge to figure out that speech needs at least 150, preferably 4000 Hz data to be intelligible, and your average camera phone (or even broadcast quality studio camera) maxes out at 60Hz frame rates, soooo...
Just get an unusually fast camera with unusually good resolution and/or zoom capabilities, and "whoop: there it is."
What I find much more interesting is some of the work that has been done on wide dynamic range video differential signal processing, looking for faint changes in hue, saturation and value - you can see a person's pulse in their skin (especially the face) just from minute changes in color... 12 bits of dynamic range or less are usually sufficient to see a pulse, while your average LCD screen only displays about 6... This kind of camera can do a lot of polygraph type data collection without being intrusive or even known to the subject under study.
🌻🌻 [google.com]
(Score: 2, Informative) by black_trout on Monday August 04 2014, @10:37PM
This is an interesting discovery, but hardly ground breaking. It seems to be the next logical step to using a laser microphone http://en.wikipedia.org/wiki/Laser_microphone [wikipedia.org] for eavesdropping.
(Score: 0) by Anonymous Coward on Monday August 04 2014, @11:23PM
What is it they say about hindsight?
(Score: 2, Funny) by black_trout on Monday August 04 2014, @11:35PM
I dunno, I forgot to record the conversation with my spy laser...
(Score: 2, Redundant) by kaszz on Tuesday August 05 2014, @12:58AM
What they are doing is the same as with the laser pointer method. The difference is that ambient light is used. That means a simple star telescope can be used to listen in.
Most normal cameras have a frame rate of at most 60 half pictures per second. And nominal voice is at 85 - 255 Hz so as soon as you see the word "high speed camera" the game is up ;) They seem to mention something about extracting from sound from ordinary videos. Which perhaps can be accomplished if the sound is considered from the perspective of the sample range to be static in frequency. Perhaps adjacent pixels can complete the data needed?
In the bigger picture it's just causality. If you change anything, there will be waves all around the place. Just because you can't measure them doesn't mean they are absent. In other words secrecy is only as good as your opponents lack of good enough equipment and methods to make use of them. Minus some laws of nature like entropy, noise floor and Shannon limit.
How to turn those pixels into sound
(Score: 0) by Anonymous Coward on Tuesday August 05 2014, @01:38AM
http://www.schlockmercenary.com/2014-05-05 [schlockmercenary.com]
From Schlock Mercenary. This is how he hears.
(Score: 2) by Geotti on Tuesday August 05 2014, @02:46AM
For anyone interested in electronic (dance, more than electro-acoustic) music world, It's All Gone Pete Tong [imdb.com] is a highly recommended movie and touches upon this topic.