h a l f b a k e r y
Fewer ducks than estimates indicate.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
or get an account
Take a video of a visual image and its reflection in the
surface of a liquid, as a series of sounds are being played,
that the disturbance in the reflected image is measurable.
Then play only the visual part of the video and infer
backwards what the sound was. Then compare the inferred
sound and the actual sound and do this a bunch of times
changing the conditions incrementally and develop
algorithms for inferring audio from visual data.
[FlyingToaster, Jan 01 2013]
Baked, only a year later
[notexactly, Oct 01 2019]
||umm... you do realize that video is at 60 samples per second, while CD audio is at over 44,000.
||This technology has already been conceived, developed,
refined, and had all of the useless bits picked out and has
yielded the laser microphone, which discerns sounds at
long range, by measuring the vibration frequency of nearby
objects, especially glass.
||//all of the useless bits picked out// [marked-for-tagline]
||Yep, baked. I've seen a video of people recovering reasonable
audio of a conversation from a video of a crisp-packet.