Lower Limb Muscle Coactivation Analysis Using Symbolic Transfer Entropy of Simultaneous Surface EMG Measurements

Surface electromyography (sEMG) signals from the coactivation of agonist and antagonist muscles can provide precise and powerful control of a lower limb prosthesis along with proprioceptive sensory feedback. However, the analysis of coactivation is challenging due to the inherent nonlinearity of the signals and the nonlinear interactions within the muscular systems during dynamic contractions. In this study, a novel nonlinear approach based on symbolic transfer entropy (STE) is proposed to characterize the coactivation of muscles at different speeds of gait. For this purpose, the sEMG is recorded from the rectus femoris (RF) and vastus lateralis (VL) of the quadriceps, as well as the biceps femoris (BF) and semitendinosus (SEM) of the hamstring muscles. The signals are collected from 20 healthy subjects walking on a treadmill at gait speeds of 2.5, 3.5, and 4.5 kilometres per hour (km/h). In addition, the knee joint angles are also obtained from the inertial measurement units. The sEMG signals are pre-processed, and eight distinct phases of gait are segmented using joint angles. A suitable symbolic scale is selected after a detailed analysis, and STE is extracted to characterize the coactivation of agonist and antagonist muscle pairs: RF-BF, RF-SEM, VL-BF and VL-SEM. The results show that STE increases with gait speed irrespective of muscle combinations, which indicates the stronger coactivation during faster locomotion. The variation of STE with respect to each phase exhibits a complex dynamic pattern in muscle coactivation. The information transfer is bidirectional and the distribution of STE is found to have significant differences across directions, phases and speeds (p¡0.001). Furthermore, the proposed STE is superior to traditional transfer entropy in terms of capturing nonlinear interactions. The study facilitates researchers in developing gait phase-based features that account for coactivation, enabling them to achieve significantly more natural and efficient gait patterns in prosthetic lower limbs.