Electromyographic (EMG) signal to joint torque processing and effect of various factors on EMG to torque model

Abstract

This study present electromyographic (EMG) signal to torque model and investigates the effects of various factors on EMG signal and EMG to torque model. Pre-processing techniques are applied on EMG signal in order to remove the DC offset, 60 Hz noise and to estimate the EMG amplitude. The estimated EMG amplitude is then mapped to joint torque using a new non-linear equation. This equation uses some parameters, whose values are obtained using nonlinear regression. Ten subjects took part in the experiments and performed variable force maximal voluntary contractions (MVC) and sub-maximal voluntary contractions (SMVC). The resulting elbow joint torque and EMG signals were pre-processed and entered to the model to find value of the parameters using nonlinear regression. Once these values were obtained they were put into the model and thus joint torque was estimated. Also EMG is analysed for effect of various factors like muscle fatigue, cross talk and different joint velocity. The results obtained from this model are highly correlated with the true values of the torque and the average correlation and mean square error for different experiments are 0.9997 and 0.047 Nm respectively. This new mathematical equation can be used to design a control system for rehabilitation and wearable robots.   Key words: Electromyographic (EMG), maximal voluntary contractions (MVC), power spectral density, levenberg-marquardt algorithm, non-linear regression.