System identification from nonuniformly spaced signal measurements

A couple of techniques for parameter estimation of a signal consisting of complex exponentials are presented. The system model employs the higher-order derivatives, or zero-initial conditioned integrals of the signal, together with the signal values. When the signal is sampled at nonuniformly distributed points, the orthogonal polynomial approximation and minimum error-variance criterion are used to compute all the values needed in the system models. The developed system models are demonstrated to give results better in accuracy than what can be obtained by employing Prony's method in a specific problem.