Title:On a Class of Parameters Estimators in Linear Models Dominating the Least Squares one, Based on Compressed Sensing Techniques

Abstract:The estimation of parameters in a linear model is considered under the hypothesis that the noise, with finite second order statistics, can be represented in a given deterministic basis by random coefficients. An extended underdetermined design matrix is then considered and an estimator of the extended parameters is proposed with minimum $l_1$ norm. It is proved that if the noise variance is larger than a threshold, which depends on the unknown parameters and on the extended design matrix, then the proposed estimator of the original parameters dominates the least-squares estimator in the sense of the mean square error. A small simulation illustrates the behavior of the proposed estimator. Moreover it is shown experimentally that the proposed estimator can be convenient even if the design matrix is not known but only an estimate can be used. Furthermore the noise basis can eventually be used to introduce some prior information in the estimation process. These points are illustrated by simulation by using the proposed estimator for solving a difficult inverse ill-posed problem related to the complex moments of an atomic complex measure.