Characterizing Texture for Symbolic Piano Music

Louis Couturier (1) , Louis Bigo (2) , Florence Levé (1, 2)

1 MIS - Modélisation, Information et Systèmes - UR UPJV 4290
2 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189

Musical texture describes how sounding components are organized, individually or with each other. In this work, we propose a new syntax to characterize symbolic texture in classical piano music and thus allow its automated analysis. We annotated the score of 9 movements of W.A. Mozart’s sonatas (totalling 1164 bars), on which tested binary classification models show promising results for automatic texture retrieval.

Mots clés: Texture; Symbolic Music; Piano; Modeling; Machine Learning

Domaines: Informatique [cs]; Musique, musicologie et arts de la scène; Apprentissage [cs.LG]

Louis Couturier, Louis Bigo, Florence Levé. Characterizing Texture for Symbolic Piano Music. Digital Music Research Network (DMRN+16), Dec 2021, Londres, United Kingdom. . ⟨hal-03526689⟩