No abstract

In the big data era, one has often to conduct the problem of parcimonious data representation. In this paper, the data under study are curves and the sparse representation stands on a semi-parametric model. Indeed, we propose an original registration model for noisy curves. The model is built transforming an unknown function by plane similarities. We develop a statistical method that allows to estimate the parameters characterizing the plane similarities. The properties of the statistical procedure is studied. We show the convergence and the asymptotic normality of the estimators. Numerical simulations and a real life aeronautic example illustrate and demonstrate the strength of our methodology.