@article{oai:takushoku-u.repo.nii.ac.jp:00000479,
 author = {西川, 佳男 and 舒, 羽 and 香川, 美仁 and Nishikawa , Yoshio and Jyo, U and Kagawa, Yoshihito},
 journal = {拓殖大学理工学研究報告, Bulletin of science and engineering, Takushoku University},
 month = {Mar},
 note = {This paper describes an identification method of forearm motions from Surface Electromyography (sEMG). When sEMG is used to control an artificial arm or other device, great care must be taken to avoid improper operations. For this reason, the motions should be estimated from the sEMG without making any mistakes. It is also necessary to estimate the motion in the shortest time after the sEMG begins occurring. In this paper, we developed a motion discrimination system that combined a wavelet transform and a convolutional neural network (CNN) and determined forearm motions by CNN output. The required parameters for the system were experimentally determined. We carried out identification experiments of three motions such as "grasping", "flexing" and "turning" by sEMG, and demonstrated the accurate identification of these motions without mistakes. After that, the effects of electrode replacement and time lapse after attachment on the identification accuracy were investigated. It was exemplified that the proposed motion identification system had robustness to sEMG changes due to the above factors.},
 pages = {21--26},
 title = {電極貼り直しに対してロバストな表面筋電位による動作識別システムの研究},
 volume = {18},
 year = {2021},
 yomi = {ニシカワ, ヨシオ and ジョ, ウ and カガワ, ヨシヒト}
}