Electromyography (EMG)-based interfaces are trending toward continuous, simultaneous control with multiple degrees of freedom. Emerging methods range from data-driven approaches to biomechanical model-based methods. However, there has been no direct comparison between these two types of continuous EMG-based interfaces. The aim of this study was to compare a musculoskeletal model (MM) with two data-driven approaches, linear regression (LR) and artificial neural network (ANN), for predicting continuous wrist and hand motions for EMG-based interfaces. Six able-bodied subjects and one transradial amputee subject performed (missing) metacarpophalangeal (MCP) and wrist flexion/extension, simultaneously or independently, while four EMG signals were recorded from forearm muscles. To add variation to the EMG signals, the subjects repeated the MCP and wrist motions at various upper extremity postures. For each subject, the EMG signals collected from the neutral posture were used to build the EMG interfaces; the EMG signals collected from all postures were used to evaluate the interfaces. The performance of the interface was quantified by Pearson’s correlation coefficient (r) and the normalized root mean square error (NRMSE) between measured and estimated joint angles. The results demonstrated that the MM predicted movements more accurately, with higher r values and lower NRMSE, than either LR or ANN. Similar results were observed in the transradial amputee. Additionally, the variation in r across postures, an indicator of reliability against posture changes, was significantly lower (better) for the MM than for either LR or ANN. Our findings suggest that incorporating musculoskeletal knowledge into EMG-based human-machine interfaces could improve the estimation of continuous, coordinated motion.

Objective . We investigated the feasibility of a novel, customizable, simplified EMG-driven musculoskeletal model for estimating coordinated hand and wrist motions during a real-time path tracing task. Approach . A two-degree-of-freedom computational musculoskeletal model was implemented for real-time EMG-driven control of a stick figure hand displayed on a computer screen. After 5–10 minutes of undirected practice, subjects were given three attempts to trace 10 straight paths, one at a time, with the fingertip of the virtual hand. Able-bodied subjects completed the task on two separate test days. Main results . Across subjects and test days, there was a significant linear relationship between log-transformed measures of accuracy and speed (Pearson’s r = 0.25, p < 0.0001). The amputee subject could coordinate movement between the wrist and MCP joints, but favored metacarpophalangeal joint motion more highly than able-bodied subjects in 8 of 10 trials. For able-bodied subjects, tracing accuracy was lower at the extremes of the model’s range of motion, though there was no apparent relationship between tracing accuracy and fingertip location for the amputee. Our result suggests that, unlike able-bodied subjects, the amputee’s motor control patterns were not accustomed to the multi-joint dynamics of the wrist and hand, possibly as a result of post-amputation cortical plasticity, disuse, or sensory deficits. Significance . To our knowledge, our study is one of very few that have demonstrated the real-time simultaneous control of multi-joint movements, especially wrist and finger movements, using an EMG-driven musculoskeletal model, which differs from the many data-driven algorithms that dominate the literature on EMG-driven prosthesis control. Real-time control was achieved with very little training and simple, quick (~15 s) calibration. Thus, our model is potentially a practical and effective control platform for multifunctional myoelectric prostheses that could restore more life-like hand function for individuals with upper limb amputation.

Simultaneous and proportional control (SPC) of neural-machine interfaces uses magnitudes of smoothed electromyograms (EMG) as control inputs. Though surface EMG (sEMG) electrodes are common for clinical neural-machine interfaces, intramuscular EMG (iEMG) electrodes may be indicated in some circumstances (e.g., for controlling many degrees of freedom). However, differences in signal characteristics between sEMG and iEMG may influence SPC performance. We conducted a pilot study to determine the effect of electrode type (sEMG and iEMG) on real-time task performance with SPC based on a novel 2-degree-of-freedom EMG-driven musculoskeletal model of the wrist and hand. Four able-bodied subjects and one transradial amputee performed a virtual posture matching task with either sEMG or iEMG. There was a trend of better task performance with sEMG than iEMG for both able-bodied and amputee subjects, though the difference was not statistically significant. Thus, while iEMG may permit targeted recording of EMG, its signal characteristics may not be as ideal for SPC as those of sEMG. The tradeoff between recording specificity and signal characteristics is an important consideration for development and clinical implementation of SPC for neural-machine interfaces.