Outils statistiques pour le positionnement optimal de capteurs dans le contexte de la localisation de sources

摘要: This thesis deals with the array geometry optimization problem in context of sources localization. We have considered two approaches for optimization: performance estimation terms mean square error approach and statistical resolution limit (SRL) approach. In first approach, lower bounds on which are usually used processing to evaluate independently estimator been considered. investigated kinds bounds: well-known Cramer-Rao bound (CRB) deterministic model parameters assumed be deterministic, Weiss-Weinstein (WWB) is less studied, Bayesian model, which, random some prior distributions. proposed closed-form expressions these bounds, can as a tool design. Compared CRB, WWB predict threshold effect MSE non-asymptotic area. Moreover, general Gaussian parameterized or covariance matrix also useful other problems. Based expressions, 3D classical planar under (i) conditional observation source signal modeled sequence (ii) unconditional process. Conditions concerning isotropic uncoupling properties were then derived.In second we characterizes minimal separation between closed spaced still allows determine correctly number sources. this thesis, interested SRL studied literature. linearized minimum probability error, introduced based detection information theories could lead interesting tools system