Using Bilateral Lower Limb Kinematic and Myoelectric Signals to Predict Locomotor Activities: A Pilot Study

Abstract

Active lower limb exoskeletons can provide assistance to the lower extremities and may drastically improve the walking abilities of millions of individuals with gait impairments. However, most currently available control systems for these devices cannot predict the user's intended movements and have yet to enable walking with seamless transitions. Recent developments in intent recognition for active lower limb prostheses have demonstrated that using kinematic and kinetic signals from the device and myoelectric signals from the user can provide an intuitive control interface for seamlessly transitioning between different locomotor activities. In this work, we determined the baseline performance of intent recognition systems using neuromechanical signals presumably accessible for controlling active lower limb exoskeletons. We collected bilateral lower limb joint kinematics and muscle activity from three able-bodied subjects while they walked on level ground, ramps, and stairs in order to train an intent recognition system. We found that both combining kinematic and myoelectric signals and including signals from the contralateral leg significantly improved intent recognition performance. We achieved an average offline prediction error rate of 1.4 ± 0.90% using bilateral kinematic and myoelectric signals, demonstrating the promising potential of translating prosthesis-based intent recognition as an alternative control strategy for active lower limb exoskeletons.