Estimating speed-accuracy trade-offs to evaluate and understand closed-loop prosthesis interfaces

Measuring and monitoring skill learning in closed-loop myoelectric hand prostheses using speed-accuracy tradeoffs

Objective.Closed-loop myoelectric prostheses, which combine supplementary sensory feedback and electromyography (EMG) based control, hold the potential to narrow the divide between natural and bionic hands. The use of these devices, however, requires dedicated training. Therefore, it is crucial to develop methods that quantify how users acquire skilled control over their prostheses to effectively monitor skill progression and inform the development of interfaces that optimize this process.Approach.Building on theories of skill learning in human motor control, we measured speed-accuracy tradeoff functions (SAFs) to comprehensively characterize learning-induced changes in skill-as opposed to merely tracking changes in task success across training-facilitated by a closed-loop interface that combined proportional control and EMG feedback. Sixteen healthy participants and one individual with a transradial limb loss participated in a three-day experiment where they were instructed to perform the box-and-blocks task using a timed force-matching paradigm at four specified speeds to reach two target force levels, such that the SAF could be determined.Main results.We found that the participants' accuracy increased in a similar way across all speeds we tested. Consequently, the shape of the SAF remained similar across days, at both force levels. Further, we observed that EMG feedback enabled participants to improve their motor execution in terms of reduced trial-by-trial variability, a hallmark of skilled behavior. We then fit a power law model of the SAF, and demonstrated how the model parameters could be used to identify and monitor changes in skill.Significance.We comprehensively characterized how an EMG feedback interface enabled skill acquisition, both at the level of task performance and movement execution. More generally, we believe that the proposed methods are effective for measuring and monitoring user skill progression in closed-loop prosthesis control.

One-shot random forest model calibration for hand gesture decoding

Objective.Most existing machine learning models for myoelectric control require a large amount of data to learn user-specific characteristics of the electromyographic (EMG) signals, which is burdensome. Our objective is to develop an approach to enable the calibration of a pre-trained model with minimal data from a new myoelectric user.Approach.We trained a random forest (RF) model with EMG data from 20 people collected during the performance of multiple hand grips. To adapt the decision rules for a new user, first, the branches of the pre-trained decision trees were pruned using the validation data from the new user. Then new decision trees trained merely with data from the new user were appended to the pruned pre-trained model.Results.Real-time myoelectric experiments with 18 participants over two days demonstrated the improved accuracy of the proposed approach when compared to benchmark user-specific RF and the linear discriminant analysis models. Furthermore, the RF model that was calibrated on day one for a new participant yielded significantly higher accuracy on day two, when compared to the benchmark approaches, which reflects the robustness of the proposed approach.Significance.The proposed model calibration procedure is completely source-free, that is, once the base model is pre-trained, no access to the source data from the original 20 people is required. Our work promotes the use of efficient, explainable, and simple models for myoelectric control.

EMG feedback improves grasping of compliant objects using a myoelectric prosthesis

BACKGROUND

Closing the control loop in myoelectric prostheses by providing artificial somatosensory feedback is recognized as an important goal. However, designing a feedback interface that is effective in realistic conditions is still a challenge. Namely, in some situations, feedback can be redundant, as the information it provides can be readily obtained through hearing or vision (e.g., grasping force estimated from the deformation of a compliant object). EMG feedback is a non-invasive method wherein the tactile stimulation conveys to the user the level of their own myoelectric signal, hence a measurement intrinsic to the interface, which cannot be accessed incidentally.

METHODS

The present study investigated the efficacy of EMG feedback in prosthesis force control when 10 able-bodied participants and a person with transradial amputation used a myoelectric prosthesis to grasp compliant objects of different stiffness values. The performance with feedback was compared to that achieved when the participants relied solely on incidental cues.

RESULTS

The main outcome measures were the task success rate and completion time. EMG feedback resulted in significantly higher success rates regardless of pin stiffness, indicating that the feedback enhanced the accuracy of force application despite the abundance of incidental cues. Contrary to expectations, there was no difference in the completion time between the two feedback conditions. Additionally, the data revealed that the participants could produce smoother control signals when they received EMG feedback as well as more consistent commands across trials, signifying better control of the system by the participants.

CONCLUSIONS

The results presented in this study further support the efficacy of EMG feedback when closing the prosthesis control loop by demonstrating its benefits in particularly challenging conditions which maximized the utility of intrinsic feedback sources.