Reva E. Johnson

Does EMG control lead to distinct motor adaptation

Powered prostheses are controlled using electromyographic (EMG) signals, which may introduce high levels of uncertainty even for simple tasks. According to Bayesian theories, higher uncertainty should influence how the brain adapts motor commands in response to perceived errors. Such adaptation may critically influence how patients interact with their prosthetic devices; however, we do not yet understand adaptation behavior with EMG control. Models of adaptation can offer insights on movement planning and feedback correction, but we first need to establish their validity for EMG control interfaces. Here we created a simplified comparison of prosthesis and able-bodied control by studying adaptation with three control interfaces: joint angle, joint torque, and EMG. Subjects used each of the control interfaces to perform a target-directed task with random visual perturbations. We investigated how control interface and visual uncertainty affected trial-by-trial adaptation. As predicted by Bayesian models, increased errors and decreased visual uncertainty led to faster adaptation. The control interface had no significant effect beyond influencing error sizes. This result suggests that Bayesian models are useful for describing prosthesis control and could facilitate further investigation to characterize the uncertainty faced by prosthesis users. A better understanding of factors affecting movement uncertainty will guide sensory feedback strategies for powered prostheses and clarify what feedback information best improves control.

Adaptation to random and systematic errors: Comparison of amputee and non-amputee control interfaces with varying levels of process noise

The objective of this study was to understand how people adapt to errors when using a myoelectric control interface. We compared adaptation across 1) non-amputee subjects using joint angle, joint torque, and myoelectric control interfaces, and 2) amputee subjects using myoelectric control interfaces with residual and intact limbs (five total control interface conditions). We measured trial-by-trial adaptation to self-generated errors and random perturbations during a virtual, single degree-of-freedom task with two levels of feedback uncertainty, and evaluated adaptation by fitting a hierarchical Kalman filter model. We have two main results. First, adaptation to random perturbations was similar across all control interfaces, whereas adaptation to self-generated errors differed. These patterns matched predictions of our model, which was fit to each control interface by changing the process noise parameter that represented system variability. Second, in amputee subjects, we found similar adaptation rates and error levels between residual and intact limbs. These results link prosthesis control to broader areas of motor learning and adaptation and provide a useful model of adaptation with myoelectric control. The model of adaptation will help us understand and solve prosthesis control challenges, such as providing additional sensory feedback.

EMG Versus Torque Control of Human–Machine Systems: Equalizing Control Signal Variability Does not Equalize Error or Uncertainty

In this paper we asked the question: if we artificially raise the variability of torque control signals to match that of EMG, do subjects make similar errors and have similar uncertainty about their movements? We answered this question using two experiments in which subjects used three different control signals: torque, torque+noise, and EMG. First, we measured error on a simple target-hitting task in which subjects received visual feedback only at the end of their movements. We found that even when the signal-to-noise ratio was equal across EMG and torque+noise control signals, EMG resulted in larger errors. Second, we quantified uncertainty by measuring the just-noticeable difference of a visual perturbation. We found that for equal errors, EMG resulted in higher movement uncertainty than both torque and torque+noise. The differences suggest that performance and confidence are influenced by more than just the noisiness of the control signal, and suggest that other factors, such as the user’s ability to incorporate feedback and develop accurate internal models, also have significant impacts on the performance and confidence of a person’s actions. We theorize that users have difficulty distinguishing between random and systematic errors for EMG control, and future work should examine in more detail the types of errors made with EMG control.

Comparison of Body-Powered Voluntary Opening and Voluntary Closing Prehensor for Activities of Daily Life

Persons with an upper-limb amputation who use a body-powered prosthesis typically control the prehensor through contralateral shoulder movement, which is transmitted through a Bowden cable. Increased cable tension either opens or closes the prehensor; when tension is released, some passive element, such as a spring, returns the prehensor to the default state (closed or open). In this study, we used the Southampton Hand Assessment Procedure to examine functional differences between these two types of prehensors in 29 nondisabled subjects (who used a body-powered bypass prosthesis) and 2 persons with unilateral transradial amputations (who used a conventional body-powered device). We also administered a survey to determine whether subjects preferred one prehensor or the other for specific tasks, with a long-term goal of assessing whether a prehensor that could switch between both modes would be advantageous. We found that using the voluntary closing prehensor was 1.3 s faster (p = 0.02) than using the voluntary opening prehensor, across tasks, and that there was consensus among subjects on which types of tasks they preferred to do with each prehensor type. Twenty-five subjects wanted a device that could switch between the two modes in order to perform particular tasks.

Comparing Functional EMG Characteristics Between Zero-Order and First-Order Interface Dynamics

The optimal control scheme for powered prostheses can be determined using simulation experiments, for which an accurate model of prosthesis control is essential. This paper focuses on electromyographic (EMG) control signal characteristics across two different control schemes. We constructed a functional EMG model comprising three EMG signal characteristics-standard deviation, kurtosis, and median power frequency-using data collected under realistic conditions for prosthesis control (closed-loop, dynamic, anisometric contractions). We examined how the model changed when subjects used zero-order or first-order control. Control order had a statistically significant effect on EMG characteristics, but the effect size was small and generally did not exceed inter-subject variability. Therefore, we suggest that this functional EMG model remains valid across different control schemes.

Similar trial-by-trial adaptation behavior across transhumeral amputees and able-bodied subjects

EMG control of powered upper limb prostheses is difficult and imprecise. One approach for improving control is to help amputees develop more accurate internal models of their prosthetic device. This may be facilitated by an intuitive mapping of neural signals to device movement, a way of providing sensory feedback, or training methods. A first step, arguably, is to understand how an amputation affects adaptation. Here we studied trial-by-trial adaptation in a simple target-directed task with transhumeral amputees and healthy controls. We found that adaptation behavior was indistinguishable between amputees using the residual limb, amputees using the intact limb, and able-bodied subjects. Transhumeral amputees completed the task with larger errors than able-bodied subjects, but there was, perhaps surprisingly, no difference between the residual and intact limb.

Joint Speed Discrimination and Augmentation For Prosthesis Feedback

Sensory feedback is critical in fine motor control, learning, and adaptation. However, robotic prosthetic limbs currently lack the feedback segment of the communication loop between user and device. Artificial sensory feedback can close this gap, but sometimes this improvement only persists when users cannot see their prosthesis. suggesting the provided feedback is redundant with vision. Thus, given the choice, users rely on vision over artificial feedback. To effectively augment vision, sensory feedback must provide information that vision cannot provide or provides poorly. Although vision is known to be less precise at estimating speed than position, no work has compared speed precision of biomimetic arm movements. In this study, we investigated the uncertainty of visual speed estimates as defined by different virtual arm movements. We found that uncertainty was greatest for visual estimates of joint speeds, compared to absolute or linear endpoint speeds. Furthermore, this uncertainty increased when the joint reference frame speed varied over time, potentially caused by an overestimation of joint speed. Finally, we demonstrate a joint-based sensory feedback paradigm capable of significantly reducing joint speed uncertainty when paired with vision. Ultimately, this work may lead to improved prosthesis control and capacity for motor learning.

Joint-based velocity feedback to virtual limb dynamic perturbations

Despite significant research developing myoelectric prosthesis controllers, many amputees have difficulty controlling their devices due in part to reduced sensory feedback. Many attempts at providing supplemental sensory feedback have not significantly aided control. We hypothesize this is because the feedback provided contains redundant information already provided by vision. However, whereas vision provides egocentric, position-based feedback, sensory feedback tied to joint coordinates may provide information complementary to vision. In this study, we tested if providing audio feedback of joint velocities can improve performance and adaptation to dynamic perturbations while controlling a virtual limb. While subjects performed time-controlled center-out reaches, we perturbed the dynamics of the system and measured the rate subjects adapted to this change. Our results suggest that initial errors were reduced in the presence of audio feedback, and we theorize this is due to subjects identifying the perturbed limb dynamics sooner. We also noted other possible benefits including improved muscle activation detection.

Effect of internal model development on effort and error during EMG control of three functional tracking tasks

Powered upper limb prostheses typically use EMG to control movement. EMG control is often variable and inefficient, and it is unclear if persons benefit from use of internal models, which have been shown to improve performance with traditional human-machine interfaces. We investigated how internal model use affected errors and effort across a group of 20 subjects using EMG control to perform a tracking task. To vary the ability of subjects to form an internal model, we altered the amount of available information using three visual displays: compensatory, pursuit, and preview. Subjects were more accurate and exerted less effort with visual displays that provided more information and enabled stronger internal model formation (p<;0.01).

The effect of powered prosthesis control signals on trial-by-trial adaptation to visual perturbations.

Powered prostheses have the potential to restore abilities lost to amputation; however, many users report dissatisfaction with the control of their devices. The high variability of the EMG signals used to control powered devices likely burdens amputees with high movement uncertainty. In able-bodied subjects uncertainty affects adaptation, control, and feedback processing, which are often modeled using Bayesian statistics. Understanding the role of uncertainty for amputees might thus be important for the design and control of prosthetic devices. Here we quantified the role of uncertainty using a visual trial-by-trial adaptation approach. We compared adaptation behavior with two control interfaces meant to mimic able-bodied and prosthesis control: torque control and EMG control. In both control interfaces, adaptation rate decreased with high feedback uncertainty and increased with high mean error. However, we did observe different patterns of learning as the experiment progressed. For torque control, subjects improved and consequently adapted slower as the experiment progressed, while no such improvements were made for EMG control. Thus, EMG control resulted in overall adaptation behavior that supports Bayesian models, but with altered learning patterns and higher errors. These findings encourage further studies of adaptation with powered prostheses. A better understanding of the factors that alter learning patterns and errors will help design prosthesis control systems that optimize learning and performance for the prosthesis user.