Parametric modeling of the reflex contribution to dynamic ankle stiffness in normal and spinal cord injured spastic subjects

Dynamic ankle stiffness was estimated at different levels of tonic voluntary contraction in normal and spastic, spinal cord injured (SCI) subjects. A new parallel-cascade identification method was used to identify the reflex contribution to overall stiffness in terms of a static nonlinearity in series with a linear system. Linear dynamics were estimated nonparametrically as impulse response functions (IRFs). Parametric models were fit to these IRFs using non-linear, least-squares methods. A simple, second-order, low-pass model was adequate for some cases, but did not work well with data from normal subjects at high levels of activation or with from spastic subjects. Much better fits were obtained using a third order model.