Mathematical Fingerprints of Songs

Katie Kinnaird

Dartmouth Mathematics Dept Graduate Program


Programs like iTunes Genius and Pandora Internet Radio attempt to create mixes of songs that are similar to each other based on various features. These programs rely on individual song analysis by music experts. In this work, we attempt to compare songs by finding structure or significant features using numerical linear algebra and statistics (i.e. without listening to the songs). In this talk, we will also discuss previous work by M. Casey, C. Rhodes, and M. Slaney.

Back to ACMS schedule