Classification automatique des signaux audio-fréquences : reconnaissance des instruments de musique

Abstract : The aim of this work is to contribute efficient solutions to machine recognition of musical instruments both for the mono-instrumental and multi-instrumental cases. We tackle the problem following an automatic classification approach. Much effort is dedicated to obtain adequate instances of the different blocs constituting the system which we propose for instrument recognition. We adopt a hierarchical classification scheme based on computer-generated taxonomies of instruments and musical ensembles. The system utilises a novel feature selection technique producing an efficient description of audio signals which, in association with support vector machines, entails high recognition accuracy on sound excerpts translating the diversity of the musical performance and recording conditions found in the real world. Our proposal is thus able to recognise up to four instruments played concurrently in jazz music possibly involving percussion.
Complete list of metadatas

Cited literature [30 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00002738
Contributor : Ecole Télécom Paristech <>
Submitted on : Friday, November 19, 2010 - 4:52:48 PM
Last modification on : Wednesday, March 27, 2019 - 1:28:10 AM
Long-term archiving on : Monday, July 26, 2010 - 8:11:09 PM

Identifiers

  • HAL Id : pastel-00002738, version 1

Citation

Slim Essid. Classification automatique des signaux audio-fréquences : reconnaissance des instruments de musique. Traitement du signal et de l'image [eess.SP]. Université Pierre et Marie Curie - Paris VI, 2005. Français. ⟨pastel-00002738⟩

Share

Metrics

Record views

1697

Files downloads

4813