It’s the first I’ve heard of any technology like this and it certainly looks very promising. Displayed by the university researchers from MIT in the YouTube video linked, we see sound extracted by looking at the minute vibrations on objects that we, with just our eyes, observe to be still. These vibrations can be equated to be reverberating at frequencies which directly correlate to the sound from the room that is bouncing off of or being absorbed by the object.
This works best with high speed cameras but as shown in the video, also works well enough with a DSLR which is slightly more accessible to the general public, and are more likely to belong to people, aside from their price.
The sound extracted from the silent video footage of the object closely matches the sound that is playing in the room, in this case a rendition of “Mary had a little lamb”, as well as voices in the room speaking the words of the song.
I can see this, once improved becoming a very useful technology. For example, cctv that has no sound may one day allow us to extract sound from the objects in the video and let us find out what is really going on in crime scenes or even footage from the past that doesn’t have any sound. This, paired with the advancements in imaging from CCTV that I also wrote about here, could help to provide a very futuristic and currently science fiction level of policing, but also increases the potential for the invasion of our privacy which has recently being very prominently talked about with regards to the nsa and even Facebook, who have allegedly been quietly allowing themselves access to our phone microphones without our true consent, in apps such as Facebook Messenger.
Let me know what you think of this in the comments here or on my Facebook page, and thanks for reading.
Project page: http://people.csail.mit.edu/mrub/VisualMic/