Verizon SongID - How is it programmed?

Question

For those new to Verizon SongID, this is a free application downloaded via the Verizon VCast network. He listens to the song for 10 seconds at any time during the song, and then sends this data to some all-knowing algorithmic beast that chews it and sends you all the ID3 tags (artist, album, song, etc.).

The first two parts and the last part are simple, but what happens during processing after sending the recorded sound?

I believe that he should take a sound file (what format?), Analyze it (how? With what?) For some key identifiers (what are these regular attributes of wave functions: phase / shift / amplitude / etc), and check it against the database.

Everything I found on the Internet about how this works is something in common, like what I typed above.

From audiotag.info

This service is based on a sophisticated sound recognition algorithm that combines advanced sound with fingerprint technology and great songs. When you upload an audio file, it analyzes the sound engine. During the analysis, the sound “fingerprint” is extracted and determined by matching it with the music database. Upon completion of this recognition process, information about the songs with their corresponding probability are displayed on the screen.

Answer 1

All these services work by taking a “fingerprint” from the sampled audio data on the client side, sending it to the server and comparing it with the fingerprint database.

One of the Shazam developers wrote an extremely informative white paper on how the technology works. This should provide you with all the necessary information.