Continuous angular position estimation of human ankle during unconstrained locomotion

Abstract

ObjectiveActive lower limb prostheses required an efficient user interface and robust control structure for its seamless operation. A nonlinear autoregressive model has been proposed and evaluated to estimate the ankle joint angle continuously. It utilizes surface electromyogram (SEMG) and knee joint angle (KA) signals inputs. MethodThe performance of the proposed model has been assessed based on the accuracy and consistency of angle estimation. A dataset of ten subjects acquired for five daily life locomotor activities has been used for model performance evaluation. Also, the contribution of KA signal towards ankle joint angle estimation has been examined. ResultsThe average angle estimation error over the subjects has been found in the range of 2.38 ± 0.78° to 5.45 ± 1.98° for various dynamic activities. The contribution of KA signal has been found significant (One-way ANOVA, p-value<0.05) as it improves the model accuracy and consistency by 52% and 36%, respectively. SignificanceThe proposed model provides an opportunity for direct control of ankle-foot prostheses by continuously predicting the ankle joint angle using SEMG and KA signal. The model’s performance proves its applicability for ankle joint angular orientation estimation for active prostheses, orthoses, and lower limb rehabilitation.