Generation of Lower Limb Exoskeleton Joint Angles from sEMG Inputs using Machine Learning

The field of biomechanics has shown tremendous progress in prosthetics with the emergence of artificial intelligence. Among the forefront technologies developed for a social cause, lower limb exoskeletons (LLE) combined with robotics, machine and deep learning techniques is the pioneer research area. The conversion of noisy, random surface electromyography (sEMG) signals to smooth continuous joint angles is important in lower limb prosthetics for the safety as well as comfortability of the wearer. Some recent technologies developed have been concentrated only on knee and hip angles with sEMG taken from different leg muscles. An LLE with 3 degrees of freedom on each leg is considered in this work. The sEMG signals are considered from only four lower limb muscles. These sEMG signals are trained using different machine learning algorithms to predict joint angles. The comparison of the models is carried out using regression plots and various performance indices. It is found that random forest algorithm performed better in predicting the joint angles.