New Music Search Engine: Queen And Led Zeppelin Segmented
By Morelli
At the 2009 International Conference on Acoustics, Speech, and Signal Processing (ICASSP) in Taiwan, UC San Diego electrical engineering Ph.D. student Luke Barrington presented a new model for music analysis (pdf) that can capture both the sound of a song and how it changes over time. The system is being used for a new search engine, dubbed Google for Music, and users would search for descriptive words, like “mellow jazz”, instead of song titles, album names or artist names.
The algorithm labels tracks differently; it cuts them into segments such as verse, chorus and bridge. It’s more accurate because it gives labels to smaller sections of a song, rather than attempting to broadly label the entire thing. But the computer must be taught, so Barrington’s team has made a game for Facebook, called Herd It, which asks you to identify instruments, genres, artist names and emotions triggered by a song.
Herd It goes as far as asking you which activities you might do while listening to a specific song. “The Facebook games are a lot of fun and a great way to discover new music. At the same time, the games deliver the data we need to teach our computer audition system to listen to and describe music like humans do,” said Gert Lanckriet, electrical engineering professor from the Jacobs School of Engineering who is mentoring the project. Here’s a segmentation of Queen’s Bohemian Rhapsody:
Those tags are automatically generated. Check out an overview of the project and more data.
Barrington asks Facebook users to help “expose” the search engine to the world of music, so it can better identify songs in searches. “The more examples of romantic songs our search engine is exposed to, the more accurately it will be able to identify romantic songs it has never heard before,” he explains.
The system is to be released later this year, so we’ll have to wait until we can compare its results with Google. In the meantime, we get to can simultaneously play games and help investigative science. Can the system be more accurate than other search engines? Do you like the games?
Play Herd It: http://apps.facebook.com/herd-it
