Sung-Gwi Cho;Muhammad Akmal Bin Mohammed Zaffir;Masahiro Yoshikawa;Jun Takamatsu;Takahiro Wada
{"title":"Influence of Forearm Postures on Hand-Wrist Gesture Recognition With Forearm Deformation Measurements","authors":"Sung-Gwi Cho;Muhammad Akmal Bin Mohammed Zaffir;Masahiro Yoshikawa;Jun Takamatsu;Takahiro Wada","doi":"10.1109/TMRB.2024.3377364","DOIUrl":null,"url":null,"abstract":"In hand-wrist gesture recognition based on biosignal, the negative influence of forearm posture variation on recognition accuracy is a common problem. Although the elbow/forearm-rotation angle influence has been investigated in several previous studies, the combined influence of these angles is still unclear. Therefore, we investigated the influence of forearm postures (both elbow and forearm rotation angles) by comparing the accuracies under various data configurations in which the posture combinations used for training the recognition model were different. We collected forearm deformation as biosignal for seven hand-wrist gestures under nine different forearm postures (combinations of three elbow and forearm rotation angles). The accuracy comparison results showed that the forearm rotation angle strongly affected recognition compared with the elbow angle, and the complex combination of elbow and forearm rotation angles had a stronger influence. The results of this study suggest that data collection can be made efficient by considering variations in the forearm postures. If time is available for data collection, it is effective to focus on the interpolation of forearm deformation to the untrained forearm postures based on those of the trained posture. If the time for data collection is limited, it is preferable to focus on variations in forearm rotation angle.","PeriodicalId":73318,"journal":{"name":"IEEE transactions on medical robotics and bionics","volume":null,"pages":null},"PeriodicalIF":3.4000,"publicationDate":"2024-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE transactions on medical robotics and bionics","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10472601/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, BIOMEDICAL","Score":null,"Total":0}
引用次数: 0
Abstract
In hand-wrist gesture recognition based on biosignal, the negative influence of forearm posture variation on recognition accuracy is a common problem. Although the elbow/forearm-rotation angle influence has been investigated in several previous studies, the combined influence of these angles is still unclear. Therefore, we investigated the influence of forearm postures (both elbow and forearm rotation angles) by comparing the accuracies under various data configurations in which the posture combinations used for training the recognition model were different. We collected forearm deformation as biosignal for seven hand-wrist gestures under nine different forearm postures (combinations of three elbow and forearm rotation angles). The accuracy comparison results showed that the forearm rotation angle strongly affected recognition compared with the elbow angle, and the complex combination of elbow and forearm rotation angles had a stronger influence. The results of this study suggest that data collection can be made efficient by considering variations in the forearm postures. If time is available for data collection, it is effective to focus on the interpolation of forearm deformation to the untrained forearm postures based on those of the trained posture. If the time for data collection is limited, it is preferable to focus on variations in forearm rotation angle.