Research interests / Thématiques de recherche

Page en cours de construction

THIS PAGE IS NOT MAINTAINED: PLEASE REFER TO THE PUBLICATION PAGE INSTEAD.

Source separation (METISS - IRISA - INRIA, from 2008)

My current works deal with audio source separation, in association with Emmanuel Vincent and Rémi Gribonval at IRISA with the METISS team/project.
I am particularly focusing on:

• Oracle performance bounds for audio source separation.
• Discriminant approaches for single-channel audio source separation.
• Objective and subjective measures for the evaluation of audio source separation results.

Automatic transcription of piano music (PhD, TELECOM ParisTech, 2004-2008)

Automatic transcription of music consists in analyzing a recording ( e.g. a .wav file) in order to extract and recognize musical events ( e.g. notes, rhythm, tempo, etc). It can be seen as the conversion process from raw data ( i.e. the physical signal) to symbolic representation ( i.e. the score), or as the inverse of synthesis or musical interpretation. Automatic transcription is often seen as a complex issue composed of specitic tasks, the main ones being pitch estimation of simultaneous notes and onset/offset detection, and takes part of a more general research field called Music Information Retrieval (MIR).
My PhD research project focuses on automatic transcription of piano music , under the supervision of Bertrand David and Roland Badeau, within the audio team of TSI Departement, TELECOM ParisTech (ENST).
My PhD defense took place on October 8th, 2008 at TELECOM ParisTech. The final version of the thesis will be freely available soon.

Time-frequency analysis (2004-2005)

My works on time-frequency analysis deal with the estimation of sinusoidal and noise components of signals.