|Home||Papers||Reports||Projects||Code Fragments||Dissertations||Presentations||Posters||Proposals||Lectures given||Course notes|
Werner Van Belle1 - firstname.lastname@example.org, email@example.com
Abstract : This talk demonstrates BpmDj, its beatgraphs, how it finds similar music and how transitions can be edited. The talk goes into the details of tempo measurement and how autodifferencing can be improved to provide results that contain much less 3/5/7 harmonics. After explaining tempo measurement, we describe how feature vectors respresenting a musical rhythm can be extracted. These feature vectors are located in a multi dimensional space and of allow for nearest neighbor detection. However, in order to make that work one must have an appropriate weighing scheme present. At the end of the talk we describe the development of a fast converging weighing scheme to learn distance metrics.
tempo decection, bpm detection, bpmdj, bpm counter, musical nearest neighbors, beatgraphs, distance metric learning
Reference: Werner Van Belle; Measuring Tempo, Extracting Rhythm, Finding similar Music; Presented at Observe, Hack, Make 2013: Oudkarpsel; Netherlands; August 2013