Electromyographic Classification to Control the SPAR Glove⁎This work was supported by NASA Space Technology Research Fellowship (NSTRF, NNX13AM70H); Rice University IDEA Grant; Mission Connect, a project of the TIRR Foundation (015-103, 017-102); and the Rice Global Engineering and Construction Forum. Issue 20 (2021)

• Record Type: Journal Article
• Authors: Britt, John E.; O'Malley, Marcia K.; Rose, Chad G.
• Is Part Of: IFAC-PapersOnLine. Volume 54:Issue 20(2021)
• Page Start: 244
• Page End: 250
• Publication Date: 2021
• Subjects: Assistive -- Rehabilitation Robotics -- Robotics -- Machine Learning in modeling -- estimation -- control; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2021.11.182
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: The SeptaPose Assistive and Rehabilitative (SPAR) Glove has been developed to assist individuals with upper extremity impairment arising from neuromuscular injury. The glove detects user intent via the MYO wearable electromyography (EMG) device. In this manuscript, pattern recognition tools infer the desired hand pose from EMG activity. The ability of the measurement and classification methods to distinguish between hand poses was evaluated with nine able-bodied participants and three participants with spinal cord injury (SCI) in an offline experiment. The strong performance of the proposed intent detection method is shown in the steady-state classification accuracy, presented as confusion matrices, as well as the average confidence for each classification. Building upon the strong performance in detecting pose, a pilot study with two participants with SCI presents the initial results of the real-time implementation of the system, which suggests directions for future work in improving the steady-state classification accuracy through expanded measurement and a refined taxonomy to enable intuitive control.