Structural segmentation of music with fuzzy clustering

Auteurs-es

  • Daniel Graves Dept. of Electrical and Comp. Eng., University of Alberta, 9107-116 St., Edmonton, AB T6G 2V4, Canada
  • Witold Pedrycz Dept. of Electrical and Comp. Eng., University of Alberta, 9107-116 St., Edmonton, AB T6G 2V4, Canada

Mots-clés :

Agglomeration, Flow of solids, Motion Picture Experts Group standards, Signal encoding, Chroma features, Computationally efficient, Constrained clustering, Digital music files, Mel Frequency Cepstral coefficients, MP3 files, Musical segmentations, Structural components

Résumé

A study was conducted to segment a digital music file, such as an MP3 file. The study demonstrated that musical segmentation was performed by using MPEG-7 features and constrained clustering based on Means. A team of researchers developed a method, called RefraiD that defects the chorus sections of music and can detect key changes in choruses, using the 12-dimensional chroma feature vector. The researchers investigated musical segmentation of structural components, using Mel frequency Cepstral Coefficient (MFCC) and compared the sequence approach of structural segmentation with the state approach (HMM). The researchers showed that the state approach is more robust and computationally efficient. A method was also proposed for musical segmentation by detecting boundaries and aggregation.

Fichiers supplémentaires

Publié-e

2008-09-01

Comment citer

1.
Graves D, Pedrycz W. Structural segmentation of music with fuzzy clustering. Canadian Acoustics [Internet]. 1 sept. 2008 [cité 7 juill. 2026];36(3):84-5. Disponible à: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/2049

Numéro

Rubrique

Actes du congrès de la Semaine canadienne d'acoustique