New trends in parametric models from 1 to the 3-D case [1/10]

Опубликовано 6 сентября 2016, 5:37

In the framework of model based multidimensional signal processing this lecture deals with a derivation of new parametric models. In the first part we will provide with the limitations of the AR models in the case of signals with periodicities and propose the so called sine-AR models exercised in various real signals [1]. In the second part we provide the limitations of the AR models in the 2-D case, an extension of the Schur-Cohn stability test to the 2-D case [7][8][9] and propose new models for texture characterization based on the 2-D Wold decomposition [2]. In a third part we will provide with the very current results on 3-D models to characterize textured volumes [3], [4], [5], [6]. References : 1- D. Labarre, E. Grivel, Y. Berthoumieu and M. Najim: “A stochastic sinusoidal model with application to speech and EEG-sleep spindles signals” Proc. EUSIPCO 2002. 2- Cl. Ramanjarasoa, O. Alata, M. Najim: “2-D Wold decomposition: a new parameter estimation approach to evanescent field EUSIPCO, 2000, extended paper submitted to IEEE Transactions on Image Processing, 2004. 3-Y. Stitou, F. Turcu et M. Najim: «Nouveaux modèles paramétriques 3-D »  Proc. GRETSI 2003, Paris sept. 9-13. B. Aksasse, Y. Stitou, Y. Berthoumieu, M. Najim: «Order estimation for the 3-D AR model » under revision for the IEEE Trans. on Signal Processing, 2004. F. Turcu, Y. Stitou and M. Najim: Multidimensional