Skip to Main content Skip to Navigation
Theses

Modélisation sinusoïdale et applications à l'indexation sonore

Abstract : The aim of the thesis is the analysis of digital audio signals based on sinusoidal modelling. The first part of the thesis is dedicated to the estimation of sinusoidal parameters, and in particular to the methods based on the Fourier transform. The advantages of this family of methods are low algorithmic complexity and ease of use. A complete state of the art of the sinusoidal estimation methods based on the Fourier transform is presented. Then, the new estimators that were developed during the thesis are detailed, in particular two new methods dedicated to the estimation of all the parameters of a sine wave modulated in both magnitude and frequency, and whose performances have proven better than the only equivalent method of the state of the art, the quadratically interpolated Fourier transform (QIFFT). Sound indexation is a fairly large research field which issue is to meet the needs of access by the content for audio material. In the second part of the thesis we have applied the sinusoidal modelling on two audio indexing tasks for which this model is particularly appropriate: pitch estimation and detection of audio objects. The two developed algorithms involve similar principles: a matching between sinusoidal peaks estimated in the audio stream and those of the referring sound object, and a measure of likelihood of the match.
Document type :
Theses
Complete list of metadatas

Cited literature [159 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00004089
Contributor : Ecole Télécom Paristech <>
Submitted on : Friday, January 9, 2009 - 8:00:00 AM
Last modification on : Friday, July 31, 2020 - 10:44:05 AM
Long-term archiving on: : Friday, September 10, 2010 - 12:52:36 PM

Identifiers

  • HAL Id : pastel-00004089, version 1

Citation

Michaël A. Betser. Modélisation sinusoïdale et applications à l'indexation sonore. Mathematics [math]. Télécom ParisTech, 2008. English. ⟨pastel-00004089⟩

Share

Metrics

Record views

300

Files downloads

5520