Kernel-based deadbeat parametric estimation of bias-affected damped sinusoidal signals

This paper deals with a novel non-asymptotic algorithm to estimate four characteristic parameters of biased damped sinusoidal signals. The proposed scheme is based on the linear integral technique introduced by [1], which relies on processing the measured signal by Volterra operators with suitably designed kernel functions. The main feature of the proposed kernels consists in the possibility to annihilate the effect of the unknown initial conditions of the hidden internal states of the system. Therefore, in the ideal case, finite-time convergence of the estimation error can be obtained. Extensive numerical simulations are presented confirming the effectiveness and the robustness of the proposed methodology.