Learning to modulate the partial powers of a single sEMG power spectrum through a novel human–computer interface

Abstract

New human-computer interfaces that use bioelectrical signals as input are allowing study of the flexibility of the human neuromuscular system. We have developed a myoelectric human-computer interface which enables users to navigate a cursor to targets through manipulations of partial powers within a single surface electromyography (sEMG) signal. Users obtain two-dimensional control through simultaneous adjustments of powers in two frequency bands within the sEMG spectrum, creating power profiles corresponding to cursor positions. It is unlikely that these types of bioelectrical manipulations are required during routine muscle contractions. Here, we formally establish the neuromuscular ability to voluntarily modulate single-site sEMG power profiles in a group of naïve subjects under restricted and controlled conditions using a wrist muscle. All subjects used the same pre-selected frequency bands for control and underwent the same training, allowing a description of the average learning progress throughout eight sessions. We show that subjects steadily increased target hit rates from 48% to 71% and exhibited greater control of the cursor's trajectories following practice. Our results point towards an adaptable neuromuscular skill, which may allow humans to utilize single muscle sites as limited general-purpose signal generators. Ultimately, the goal is to translate this neuromuscular ability to practical interfaces for the disabled by using a spared muscle to control external machines.

A Novel Feature Optimization for Wearable Human-Computer Interfaces Using Surface Electromyography Sensors

“…Human-computer interfaces (HCI) for those with motor deficits based on bioelectrical signals have received increasing attention in the last decade. HCI provides communication and control channels between human subjects and the surrounding environment with the purpose of replacement or augmentation of muscle activity [ 1 ]. Common classes of bio-signals used to control assistive devices include electromyography (EMG) [ 2 , 3 ], electroencephalography (EEG) [ 4 , 5 ], electrooculography (EOG) [ 6 , 7 ], and fusions of these signals [ 8 , 9 , 10 , 11 ].…”

Assistive grasping with an augmented reality user interface

“…A series of works (Joshi et al, 2011; Perez-Maldonado et al, 2010; Skavhaug et al, 2012, 2016) have shown that the input device can record two simultaneous channels from a single recording site. This is achieved by training the subject to modulate the activation of the muscles near the recording site in order to control the power voluntarily in two separate frequency bands.…”

Effects of muscle fatigue on the usability of a myoelectric human-computer interface

“…We employ a previously described novel myoelectric HCI, which requires subjects to manipulate the power spectrum in a single surface EMG (sEMG) signal through low-level dynamic muscle contractions (Perez-Maldonado et al, 2010;Vernon & Joshi, 2011, Skavhaug, Bobell, Vernon, & Joshi, 2012Lyons & Joshi, 2013;Skavhaug, Lyons, Nemchuk, Muroff, & Joshi, 2016). The ultimate purpose of the sEMGbased device is the translation to a practical user interface for people with paralysis.…”

“…Here, we add to the limited body of work by examining the effects of fatigue on a novel single-site myoelectric human-computer interface (HCI) controlled via contractions of a head muscle (the auricularis posterior; AP) and a wrist muscle (the Extensor Pollicis Longus; EPL). We monitored interface use during a canonical cursor-to-target task (Perez-Maldonado, Wexler, & Joshi, 2010;Skavhaug, Lyons, Nemchuk, Muroff, &Joshi, 2016, Wolpaw andMcFarland 2004) across two sessions. In the first session, we asked subjects to complete 300 trials without breaks (2.5-3 hours of continuous use).…”