Elbow Joint Angle Estimation with Surface Electromyography Using Autoregressive Models

This paper presents a method to estimate the elbow joint angle from surface electromyography (sEMG) measurements of biceps, triceps and brachioradialis. This estimation is of major importance for the design of human robot interfaces based on sEMG. It is also relevant to model the muscular system and to design biomimetic mechanisms. However, the processing and interpretation of electromyographic signals is challenging due to nonlinearities, unmodeled muscle dynamics, noise and interferences. In order to determine an estimation model and a calibration procedure for the model parameters, a set of experiments were carried out with six subjects. The experiments consisted of series of continuous (cyclical) and discrete elbow flexo-extensions with three different loads (i.e. 0 kg, 1.5kg and 3 kg). The sEMG data from the biceps brachii, triceps brachii and brachioradialis and the joint angle were recorded. Four different modeling techniques were evaluated: State Space (SS), Autoregressive with Exogenous Input (ARX), Autoregressive Moving-Average with Exogenous Input (ARMAX), Autoregressive Integrated Moving-Average with Exogenous Input (ARIMAX). After the model was selected, a second experiment was performed in order to validate the estimation procedure. The results show a procedure to estimate the EMG-to-angle relation with high correlation and low mean-square-root errors with respect to the measured angle data.