It was 2005.
I was driving in my car, listening to the radio, and had just endured an annoying song. As a commercial started, I switched to another station and heard the end of a song I loved! Urgh! Why did the universe do this to me!?
Then I thought, could I get a computer to listen to all the stations, know which songs I like, and let me know when one of them is playing on a different station?
The first step was to see if I could get the computer to "listen" to anything... In the proceeding image you see the paths marked out by Enya's Once You Had Gold (blue) superposed over Mariah Carey’s Fantasy (orange). That's the computer, "listening" to the songs. :)
This started as a project in a machine learning class. With the help of a couple of my professors it became a published paper. But I really was inspired by the experience in my car.
Essentially I took the fourier transform of the log of fourier transform of the songs as they play, a few seconds worth at a time. This is called "timbre". That gave me like 19 axises of information. I then used a minimum distance metric to measure whether a song was "similar" or not. Surprisingly, songs that "sounded" similar to me, also lined up with the similarity metric. Yay!
But's that's as far as I got with the idea, my time ran out. Nowadays, just get a Spotify account. ;)
Let's talk about what I can do for you (demo reel on request).