A Framework and Call to Action for the Future Development of EMG-Based Input in HCI

Abstract

Electromyography (EMG) has been explored as an HCI input modality following a long history of success for prosthesis control. While EMG has the potential to address a range of hands-free interaction needs, it has yet to be widely accepted outside of prosthetics due to a perceived lack of robustness and intuitiveness. To understand how EMG input systems can be better designed, we sampled the ACM digital library to identify limitations in the approaches taken. Leveraging these works in combination with our research group's extensive interdisciplinary experience in this field, four themes emerged (1) interaction design, (2) model design, (3) system evaluation, and (4) reproducibility. Using these themes, we provide a step-by-step framework for designing EMG-based input systems to strengthen the foundation on which EMG-based interactions are built. Additionally, we provide a call-to-action for researchers to unlock the hidden potential of EMG as a widely applicable and highly usable input modality. CCS CONCEPTS• Human-centered computing → Interaction techniques; Gestural input; Interaction design.

Using Eye Gaze to Train an Adaptive Myoelectric Interface

“…1b). Previous literature defined this cursor-tracking as a continuous task [15,70], which has the potential to be translated to other continuous computer or human-device tasks, such as handwriting [32,39], prostheses control [21,75], or wheelchair navigation [49]. Our results demonstrate the practical use of our gaze-trained paradigm, which worked as effectively as the previous task-trained method when natural eye movements were present.…”

“…To measure the online usability of the adaptive interface, we applied the online evaluation method discussed in prior literature [15] and quantified the continuous task performance as the time-domain tracking ability [70]. Specifically, we quantified task performance over time interval [𝑡 0 , 𝑡 1 ] using the mean square error between the 2D target 𝑟 and cursor 𝑦 signals,…”

“…For instance, prior to operating an interface, users' muscle activity is measured to capture specific movements or contractions, and this recorded activity is used to program a fixed mapping. Although this scheme has promising offline accuracy, the dynamic nature of human users in the loop and the large signal variability in sEMG [67] often results in a mismatch between offline algorithm performance and real-time interface usage [15,27,66]. Another limitation of myoelectric interfaces in current studies is the pre-assigning gestures for specific tasks.…”

“…Adaptive interfaces are not new in the HCI community, as they have been studied in many machine learning use cases [15,17]. However, adoption of adaptive algorithms is still limited in myoelectric applications.…”

“…For instance, offline methods can be used to synthesize an interface to optimize performance with respect to the distribution of subjects represented in a set of labeled training data [39,50,60,75,76]. However, the performance observed online with this approach consistently falls short of that obtained during offline training [9,15,27,38]. This finding is unsurprising, since users are not stationary data sources -they adapt to the interfaces they use and tasks they are assigned.…”