> It added it had received the microphone data only as code rather than audio, and that it could match that code with audio data from a match.
That sounds funnily absurd to me. By that line of argument, even sound is not really audio. After all, it's being encoded as air pressure waves :-)
EDIT: Pardon my ignorance. Based on Google Translate's translation of their statement [1], it seems that they are using some kind of perceptual hashing which is quite interesting.