Towards Extraction of Ground Truth Data from DJ Mixes
Résumé
DJ techniques are an important part of popular music culture but that are so far not very well researched because of the lack of annotated databases of DJ mixes. We first offer an overview of the necessary components to annotate recorded mixes automatically, which are: fingerprinting to obtain the tracklist, alignment to determine where in the mix each track starts and what tempo changes are applied to acheive beat-synchronicity, unmixing to estimate the fade curves for volume, bass and treble, and content and metadata analysis to derive the genre and social tags attached to the music to inform about the choices a DJ makes when creating a mix. Most of these components have been addressed by recent MIR research, except the alignment part for which we will give a first attempt using multi-scale correlation and dynamic time warping.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...