This article presents an algebraic on-line parameters estimation method for Linear Time Invariant (LTI) systems subject to polynomial perturbations. Particular attention is given to practical implementation. The estimator may be expressed as simple FIR filter, which is advantageous for online application, since the coefficients may be computed offline. The robustification against possible singularities is shown and the approach has the advantage of eliminating singularities that may occur in the experimental identification. This algorithm is illustrated experimentally on a Permanent Magnet Stepper Motor (PMSM) and a magnetic bearing for which parametric estimation is known to be a difficult task. The parameters to be estimated are sometimes measurable with difficulties and may slightly vary over time.