Skip to Main content Skip to Navigation
Theses

Algorithmes adaptatifs d'identification et de reconstruction de processus AR à échantillons manquants

Abstract : We are concerned in online reconstruction of signals subject to missing samples using a parametric approach. We propose adaptive algorithms for identification and reconstruction of AR processes with missing samples. Firstly, we consider the extension of gradient algorithms to the case of signals with missing samples. We propose two alternatives to an existent algorithm based on two other predictors. The proposed algorithms converge toward unbiased estimation of the parameters. However, gradient algorithms suffer from slow convergence. Therefore, we consider the RLS algorithm extension to the case of signals subject to missing samples. We use jointly, the pseudo-linear RLS algorithm for the identification and a Kalman filter for optimal reconstruction of the signal in the least mean square sense. The estimated parameters, using the proposed algorithm, are unbiased. In addition, it is fast and well adapted to the identification of non stationary signals. Nevertheless, looking for the control of the identified filter stability, we propose to identify the signal using the lattice structure of the filter. We propose an extension of the Burg adaptive algorithm to the case of signals subject to missing observations, using a Kalman filter for the prediction. The estimated parameters guarantee the stability of the corresponding filter. In addition, it offers a faster tracking parameter. Finally, we use the proposed algorithms in non uniform transmission systems. We then get the improvement of both the SNR and the average transmission rate.
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00273585
Contributor : Karine El Rassi <>
Submitted on : Tuesday, April 15, 2008 - 4:29:28 PM
Last modification on : Monday, December 14, 2020 - 12:28:40 PM
Long-term archiving on: : Friday, May 21, 2010 - 1:46:09 AM

Identifiers

  • HAL Id : tel-00273585, version 1

Collections

Citation

Rawad Zgheib. Algorithmes adaptatifs d'identification et de reconstruction de processus AR à échantillons manquants. Mathématiques [math]. Université Paris Sud - Paris XI, 2007. Français. ⟨tel-00273585⟩

Share

Metrics

Record views

730

Files downloads

209