Music Research: Drum and Bass beat detection at McGill

Who says music researchers are stodgy?  Some interesting new research going on at McGill that could someday work to improve beat detection in all kinds of music software:

An essential first step in understanding how various producers uniquely use percussion, melody, and harmony in their tracks is downbeat detection (to find the first beat of every measure). We’ve developed a style-specific method of downbeat detection catered to Hardcore, Jungle, and Drum and Bass (HJDB) by combining multiple forms of metrical information: low-frequency onset detection; beat tracking; and a regression model (SVR) trained on the timbre and sequence order of breakbeats. In a recent evaluation using 206 HJDB tracks, we demonstrate superior accuracy of our style-specific method over four general downbeat detection algorithms (including two commercial algorithms).

Read the rest at Breakscience.

Author: Will Kuhn

I teach music technology to high schoolers. I do some other stuff too. @willkuhn on Twitter. View all posts by Will Kuhn