Creating non-linear models of human gait

Abstract

s / Gait & Posture 39S (2014) S1–S141 S141 number (called the ERR – Error Reduction Ratio) associated with each term. This is used during the model estimation process in order to eliminate irrelevant terms but it is also useful at a later stage (during system characterisation) in order to identify themost important contributors to the output. In the example given above the termwith thehighestERRwas+0.0044635495×u1(n)whereu1 was the ipsilateral ankle flexionmoment. Themodel appears therefore to have captured the importance of plantarflexor/knee extensor couple. The fact that the model has identified a known mechanism provides reassurance that the approach has significant potential to identify other such features of human gait. This is an avenuewe are currently exploring further as part of future work. In addition, and perhapsmore importantly, we also hope to use themodels’ predictive capacity to forecast treatment outcome, for example to predict kinematic changes resulting from therapies targeted at influencing joint moments (muscle strengthening, spasticity management etc.).