If you're referring to the way, e.g. Youtube, automatically detects the author of the video or artist upon upload, then this article will give you an overview about tagging "audio":<p><a href="http://en.wikipedia.org/wiki/Acoustic_fingerprint" rel="nofollow">http://en.wikipedia.org/wiki/Acoustic_fingerprint</a>
One way might be to run the audio stream through a speech-to-text engine and parse the resulting transcript.<p>A video recognition system could also be used to identify faces, landmarks and common objects.