Google Researchers Create TV Audio Analysis System

segphault writes "Ars Technica reports on a paper (PDF) about ambient audio analysis authored by Google researchers. The system described in the paper can effectively determine what television show a user is watching just by capturing a short audio clip. The paper explains how a regular computer microphone can be used to record an audio clip that is then converted into a statistical data summary and transmitted to a remote server which matches the clip against archived data in order to ascertain which TV show it is associated with. Apparently, the system is fully viable, and other kinds of ambient noise don't negatively impact its accuracy. The paper also describes how web services can provide contextually relevant information based on a consumer's television viewing activities."

Re:This already exists?

[Nimloth]

Not quite, the service you mentionned recognizes a sound clip against its more or less exact replica in the (large) database.
This here matches a sound clip to a pattern to find the TV show, meaning it doesn't have all the current episodes of the program in its database, it just has statistical data and patterns which help it match the audio. The latter could successfully match new (live) episodes without having the database updated. Your tune system wouldn't.

Re:This already exists?

[asuffield]

I always wanted to have the ability to "hash" songs, and come up with an algorithm that would be robust enough to work across multiple codecs and encoding options, different (relative) normalizations, and maybe even be able to handle empty space at the beginning and/or end of the song.

It's been done. Here's a system where you can hum a tune and it tells you the song: http://www.musipedia.org/ [musipedia.org]

Current systems are mostly based on pitch changes, so they aren't perfect (especially with the recycled slush turned out by low-grade high-visibility pop acts), and largely useless for rap, but they mostly work. There are numerous variations on the system, this is just one of the more significant ones that is publically availabel on the web.

I would think by making a hash based on values relative to sound signatures within the clip this might be possible, but I don't really know how this stuff works

What google is doing may or may not be related. They might instead be using a form of speech recognition technology, or a combination of both, or something else entirely.

Re:recognizing sound samples

[McCart42]

MusicBrainz is the service you're looking for; it's been around for a few years now and they're just coming out with a new beta for Linux and Windows users alike. Someone already mentioned iEatBrainz, the Mac variant.

http://en.wikipedia.org/wiki/MusicBrainz [wikipedia.org]

http://musicbrainz.org/ [musicbrainz.org]