Search

The force and position data used to construct models of limb impedance are often obtained from closed-loop experiments. If the system is tested in a stiff environment, it is possible to treat the data as if they were obtained in open loop. However, when limb impedance is studied in a compliant environment, the presence of feedback cannot be ignored. While unbiased estimates of a system can be obtained directly using the prediction error method, the same cannot be said when linear regression or correlation analysis is used to fit nonparametric time- or frequency-domain models. We develop a prediction error minimization-based identification method for a nonparametric time-domain model augmented with a parametric noise model. The identification algorithm is tested on a dynamic mass–spring–damper system and returns consistent estimates of the system's properties under both stiff and compliant feedback control. The algorithm is then used to estimate the impedance of a human elbow joint in both stiff and compliant environments. [ABSTRACT FROM PUBLISHER]

Abstract: Repetitive peripheral magnetic stimulation (RPMS) is an innovative approach in treatment of central paresis, e.g. after stroke. In this article we present a neuromuscular model for the RPMS-induced isometric muscle contraction. This model is the basis for our two recent goals in research: improvement and assessment of the RPMS therapy by means of position controlled induction of functional movements and with automated system identification based therapy evaluation. In order to adapt the model parameters to the individual a nonlinear on-line system identification method is proposed. Physiological systems may be complex and detailed so that in many cases macroscopic models of the dominant characteristics are built. Therefore, mathematical descriptions of these models and respective parameter identification methods have to cope with uncertainties. This paper presents a separable nonlinear regression model of Hammerstein structures that maximizes the possibility of incorporation of a priori knowledge and is still flexible to structural uncertainties. A robust on-line identification method based on the Levenberg–Marquardt algorithm is presented that works in a reduced parameter space, due to the separability of the model equation. [Copyright &y& Elsevier]