It's called "audio fingerprinting" and it basically converts a snippet of audio from your phone is something similar to a hash of that audio. Then, using some c

It's called "audio fingerprinting" and it basically converts a snippet of audio from your phone is something similar to a hash of that audio. Then, using some clever algorithms, it can be quickly matched against a database of existing hashes. Because of the way the algorithm works the match doesn't have to be either 0% or 100%, because realistically whatever is recorded on your phone will never exactly match the original. Disclaimer: I used to work at a company that built this type of technology.