Symbolic Textural Features and Melody/Accompaniment Detection in String Quartets - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

Symbolic Textural Features and Melody/Accompaniment Detection in String Quartets

Résumé

Music is often described as melody and accompaniment, and several MIR studies try to identify melodies. But the organization of voices is not limited to such a distinction between melody and accompaniment: Textural effects – such as repeated notes, syncopes, homorhythmy, parallel moves or imitation – underline the melody/accompaniment layout, and changes in texture usually mark structural transitions in music. We investigate how textural and other characteristics can help to identify melodic voices in polyphonic music. We select measure-level features to analyze symbolic scores of string quartets, including new textural features, and propose models to predict, on each measure, melodic and accompaniement layers in such scores, each layer possibly including several instruments. We evaluate these sets of features and the models on 12 movements in Haydn and Mozart string quartets. The best models have an average accuracy of more than 85%, taking into account both statistical and textural features.
Fichier principal
Vignette du fichier
2021-cmmr-melody-texture.pdf (264.13 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03322543 , version 1 (26-11-2021)

Identifiants

  • HAL Id : hal-03322543 , version 1

Citer

Louis Soum-Fontez, Mathieu Giraud, Nicolas Guiomard-Kagan, Florence Levé. Symbolic Textural Features and Melody/Accompaniment Detection in String Quartets. International Symposium on Computer Music Multidisciplinary Research (CMMR 2021), Nov 2021, Online, Japan. pp.175-184. ⟨hal-03322543⟩
191 Consultations
142 Téléchargements

Partager

Gmail Facebook X LinkedIn More