Skip to Main content Skip to Navigation
Theses

Développement de méthodes de traitement de signaux spectroscopiques : estimation de la ligne de base et du spectre de raies

Abstract : This thesis is part of a collaboration between the CRAN (UMR 7039) and the LCPME (UMR 7564) aiming at developing analysis methods for spectroscopic signals.

Firstly, a deterministic method is proposed. It allows to estimate the baseline as the polynomial minimizing a non-quadratic cost function (Huber function, truncated quadratic). In particular, asymmetrical cost functions are well adapted for spectra whose peaks are positive. The minimization is achieved by the half-quadratic minimisation algorithm LEGEND.

Secondly, we consider the problem of the spectral peaks estimation: the Bayesian approach coupled with MCMC methods provides a very powerful theoretical framework. In a first approach, the problem is considered as an unsupervised blind positive sparse spike deconvolution. The sparse spike train is modelled as a Bernoulli-Gaussian process with positive support; an original mixed accept-reject algorithm allows the simulation of positive normal variables. An interesting alternative method consists in considering the problem as a decomposition into elementary signals. An original model is introduced, allowing to keep the model dimension fixed. The label switching problem is also studied and a relabelling algorithm is proposed.

The methods are applied to both simulated and experimental spectra (infrared and Raman).
Complete list of metadata

Cited literature [119 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00011477
Contributor : Vincent Mazet <>
Submitted on : Friday, January 27, 2006 - 9:39:26 AM
Last modification on : Friday, October 23, 2020 - 8:38:04 AM
Long-term archiving on: : Saturday, April 3, 2010 - 9:53:31 PM

Identifiers

  • HAL Id : tel-00011477, version 1

Collections

Citation

Vincent Mazet. Développement de méthodes de traitement de signaux spectroscopiques : estimation de la ligne de base et du spectre de raies. Traitement du signal et de l'image [eess.SP]. Université Henri Poincaré - Nancy I, 2005. Français. ⟨tel-00011477⟩

Share

Metrics

Record views

760

Files downloads

8373