Planned intervention: On Wednesday April 3rd 05:30 UTC Zenodo will be unavailable for up to 2-10 minutes to perform a storage cluster upgrade.
Published January 30, 2018 | Version v1
Thesis Open

Genre Classification based on Predominant Melodic Pitch Contours

  • 1. University of Coimbra

Contributors

  • 1. Pompeu Fabra University
  • 2. New York University

Description

We present an automatic genre classification system based on melodic features. First a ground truth genre dataset composed of polyphonic music excerpts is compiled. Predominant melodic pitch contours are then estimated, from which a series of descriptors is extracted. These features are related to melody pitch, variation and expressiveness (e.g. vibrato characteristics, pitch distributions, contour shape classes). We compare different standard classification algorithms to automatically classify genre using the extracted features. Finally, the model is evaluated and refined, and a working prototype is implemented. The results show that the set of melody descriptors developed is robust and reliable. They also reveal that complementing low level timbre features with high level melody features is a promising direction for genre classification.

Files

Rocha-Bruno-Master-thesis-2011_0.1.pdf

Files (2.2 MB)

Name Size Download all
md5:47f8af650660eda863e8d005248aea95
2.2 MB Preview Download