FES-Induced Torque Prediction With Evoked EMG Sensing for Muscle Fatigue Tracking

Abstract

This study investigates a torque estimation method for muscle fatigue tracking, using stimulus evoked electromyography (eEMG) in the context of a functional electrical stimulation (FES) rehabilitation system. Although FES is able to effectively restore motor function in spinal cord injured (SCI) individuals, its application is inevitably restricted by muscle fatigue. In addition, the sensory feedback indicating fatigue is missing in such patients. Therefore, torque estimation is essential to provide feedback or feedforward signal for adaptive FES control. In this work, a fatigue-inducing protocol is conducted on five SCI subjects via transcutaneous electrodes under isometric condition, and eEMG signals are collected by surface electrodes. A myoelectrical mechanical muscle model based on the Hammerstein structure with eEMG as model input is employed to capture muscle contraction dynamics. It is demonstrated that the correlation between eEMG and torque is time-varying during muscle fatigue. Compared to conventional fixed-parameter models, the adaptedparameter model shows better torque prediction performance in fatiguing muscles. It motivates us to use a Kalman filter with forgetting factor for estimating the time-varying parameters and for tracking muscle fatigue. The assessment with experimental data reveals that the identified eEMG-to-torque model properly predicts fatiguing muscle behavior. Furthermore, the performance of the time-varying parameter estimation is efficient, suggesting that real-time tracking is feasible with a Kalman filter and driven by eEMG sensing in the application of FES.

FES-induced muscular torque prediction with evoked EMG synthesized by NARX-type recurrent neural network

“…NARX-RNN muscle model structure stimulation artifacts; 2) applying low-pass filter for measured torque; 3) calculating the mean absolute value (MAV) of eEMG induced by FES. For more details on the stimulation experiment session and data processing procedure, please refer to [13]. The MAV of EMG u(t) ∈ R and measured muscle torque T (t) ∈ R are normalized with their maximum values respectively, before equipped into the model as the input and output targets for learning respectively, i.e., after normalization, T (t) and u(t) are within range [0, 1].…”

“…The patient configuration is shown by Tab. I [13]. T6 means that 6th thoracic vertebra is damaged for their injury.…”

“…To remedy current estimation/prediction methods for stimulated muscular force, a Kalman filter with fixed forgetting factor was proposed to estimate the muscular torque with FES induction and eEMG [13], by choosing Polynomial Hammerstein Model (PHM) with eEMG and introducing muscle torque [14] as the muscle-model input and output respectively. Such work shows better prediction results with the latest model estimation if the forgetting factor is well chosen.…”

“…In this paper, in order to improve performance of predicting potential muscle fatigue, a NARX-type recurrent neural network (NARX-RNN) model is proposed to establish the modeling between FES-induced torque/force and eEMG with real-time identification capability. Different from NARX model used in [13], the proposed NARX-RNN is established based on a NARX form supplemented with eEMG and torque coupled terms. To enhance convergence of the estimated parameters [16] of such NARX-RNN model, directly-learning pattern [17] is employed [18].…”

A New Method towards Achieving FES-Induced Movement

“…Control systems for lower limb FES-assisted movements do exist but were a predominantly open loop. The expectation, however, was that shifting to closed loop could enhance both safety and performance for such systems [3][4][5]. Achieving these could facilitate the emergence of a more FES-aided orthosis which is scarce; imagine even in the United States of America (USA) currently the health department approves only a single device for that purpose [6].…”