AUTHOR=Asogbon Mojisola Grace , Samuel Oluwarotimi Williams , Nsugbe Ejay , Li Yongcheng , Kulwa Frank , Mzurikwao Deogratias , Chen Shixiong , Li Guanglin 

TITLE=Ascertaining the optimal myoelectric signal recording duration for pattern recognition based prostheses control

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 17 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2023.1018037

DOI=10.3389/fnins.2023.1018037

ISSN=1662-453X

ABSTRACT=Electromyogram-based pattern recognition (EMG-PR) has been widely considered an essentially intuitive control method for multifunctional upper limb prostheses. A crucial aspect of the scheme is the EMG signal recording duration (SRD) from which requisite motor tasks are characterized per time, impacting the system's overall performance. For instance, lengthy SRD inevitable introduces fatigue (that alter the muscle contraction patterns of specific limb motions) and incur high computational costs in building the motion intent decoder, resulting to inadequate prosthetic control and controller delay in practical usage. Conversely, relatively shorter SRD may lead to insufficient exemplars that would preclude adequate motion intent decoding required for dexterous control of the prostheses. Therefore, determining the optimal SRD required to characterize limb motion intents adequately and aid intuitive PR-based control remains an open research question. Thus, for the first time, this study systematically investigated the impact of varied lengths of myoelectric SRD on the characterization of inherent motor intents associated with multiple classes of finger gestures. The investigation involved characterizing fifteen classes of finger gestures performed from eight normally limb subjects under various groups of EMG SRD denoted as short, moderate, and long. Experimental results showed that short EMG SRD yielded the least classification error and highest Matthew Coefficient Correlation values compared to other SRD groups. In addition, qualitative visualization via the power spectrum analysis showed that short signal length exhibits an intense frequency spectrum, which may be helpful in decoding inherent motor tasks more accurately. In inclusion, the study's findings suggest that considering EMG SRD per time would aid adequate characterization of multiple classes of limb motion tasks, reduced computational cost, and classifier training time in PR-based control schemes for multifunctional prostheses.