Un système interactif pour l'analyse des musiques électroacoustiques

Abstract : Electro-acoustic music is still hardly studied in the field of Music Information Retrieval. Most research on this type of music focuses on composition tools, pedagogy and music analysis. In this thesis, we focus on scientific issues related to the analysis of electro-acoustic music. After placing this music into historical context, a study of the practices of three professional musicologist allows us to obtain guidelines for building an analysis system. Thus, we propose an interactive system for helping the analysis of electro-acoustic music that allows one to find the various instances of the sound objects of a polyphonic piece. The proposed system first performs a segmentation to identify the initial instances of the main sound objects. Then, the user can select the target sound objects before entering an interactive loop that uses active learning and relevance feedback provided by the user. The feedback of the user is then used by the system to perform a multilabel classification of sound segments based on the selected ones sound objects. An evaluation of the system is performed by user simulation using a synthetic corpus. The evaluation shows that our approach achieves satisfying results in a reasonable number of interactions.
Complete list of metadatas

Cited literature [112 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00676691
Contributor : Sébastien Gulluni <>
Submitted on : Tuesday, March 6, 2012 - 9:14:20 AM
Last modification on : Monday, February 25, 2019 - 11:08:10 AM
Long-term archiving on : Wednesday, December 14, 2016 - 10:27:17 AM

Identifiers

  • HAL Id : pastel-00676691, version 1

Citation

Sébastien Gulluni. Un système interactif pour l'analyse des musiques électroacoustiques. Traitement du signal et de l'image [eess.SP]. Télécom ParisTech, 2011. Français. ⟨pastel-00676691⟩

Share

Metrics

Record views

377

Files downloads

938