Skip to Main content Skip to Navigation
Theses

Modélisation de signaux fortement non stationnaires à phase et à amplitude locales polynomiales.

Abstract : This work concentrates on the estimation and reconstruction of highly non-stationary signals having both
non-linear amplitude and frequency modulations. We propose a new method for estimating the above mentioned
signals with high additive noises. Most of the previously published work in this domain is plagued
by shortcomings and at times even fails to address the problem of estimation.
We have shown that by using a local approach, which provides flexibility to fit to the local variations of
the instantaneous amplitude and frequency, the estimation results are improved significantly. The novelty
of the presented work consists in the use of parametric models well-adapted and defined on short time
segments. We have also proposed a new technique for obtaining short time segments from the entire signal.
Finally, we propose a merging strategy to reconstruct the entire signal along with its modulations. The
proposed estimation approach is great interest as it does not require a higher order model to estimate the
entire signal.
Initially, we validate the estimation effeciency of the local model considering a short time segment.
This local model uses polynomial approximations of both local instantaneous amplitude and frequency,
decomposed using a discrete orthonormal base which we derived. We also affirm that this base reduces
the coupling of model parameters. Hence, the estimation accuracy is enhanced. Furthermore, we have
compared two two techniques to estimate the model parameters. The first appraoch is based upon the
maximization of the likelihood function by means of an optimization stochastique technique named Simulated
Annealing. The second approach uses the MCMC methods with Metroplois-Hastings algorithm
We demonstrate using simulations and real applications that the prposed approach improves both
performance and robustness of signal estimation as compared to the existing Higher Ambiguity function
based technique.
Complete list of metadatas

Cited literature [98 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00176079
Contributor : Meryem Jabloun <>
Submitted on : Tuesday, October 2, 2007 - 1:06:02 PM
Last modification on : Thursday, November 19, 2020 - 12:59:40 PM
Long-term archiving on: : Thursday, September 27, 2012 - 12:31:06 PM

Identifiers

  • HAL Id : tel-00176079, version 1

Citation

Meryem Jabloun. Modélisation de signaux fortement non stationnaires à phase et à amplitude locales polynomiales.. Traitement du signal et de l'image [eess.SP]. Institut National Polytechnique de Grenoble - INPG, 2007. Français. ⟨tel-00176079⟩

Share

Metrics

Record views

516

Files downloads

2084