The Sound of Your Emotions

DTU Compute postdoc Jens Madsen is researching algorithms to pick from a music library music to match an emotion described by a user.


Music services like Spotify, Apple Music and Wimp provide access to millions of songs, but picking out just a few suitable for a particular mood is a daunting task for `wetware’, the conscious mind easily distracted by many pieces of irrelevant information. Hence the need for a clever Big Data search based on a mathematical description of mood.

“An algorithm can of course never describe the magical element in music”, says Jens Madsen, “I only seek to quantify what aspects of music make many people happy, or ignite other well defined emotions. I had to go out into literature on the psychology of perception. I realised that I had to focus on the emotions expressed in the music by the artist, rather that on the emotions evoked in the listener. The latter is simply too complicated and unpredictable.

In my initial model I had hundreds of parameters. Now, in this postdoc project, things are falling into place and the program can begin to take on really large sets of data. I am slowly approaching a good algorithm which can match the gradings and annotations by test volunteers.”

“Ultimately I hope to create a tool which can improve life for some people — a sort of psychological tool, but without any chemical side effects. An algorithm could at worst only suggest music you happen not to like, it never ruins anything in your body”, says Jens Madsen.
Contact Jens Madsen: