In this study, we focus on scenarios where a user performs the sit-to-stand (STS) motion with a power-assisted system—such as a lower-limb exoskeleton or an automatic assistive chair. If the user changes their intention from standing up to stopping the motion, subsequently sitting down again, due to physical discomfort, the assistive force may become inconsistent with the user’s actual intention, potentially leading to loss of balance or even falls.
To address this issue, we propose a dual-DNN architecture that recognizes intention changes using surface EMG. Experiments with ten participants show that the system can detect intention changes with an average accuracy of 70%, approximately 146.65 ms before the actual hip-rising motion stops. Such early detection enables assistive devices to respond to behavior changes in advance, thereby improving usability and safety.
In the field of power assist in the context of STS, such a function improves the adaptability of the assistive system and provides users with a safer and more intuitive experience.
| 氏名 | コース | 研究室 | 役職/学年 |
|---|---|---|---|
| YANG YUANDA | 学術情報メディアセンター | 中村研究室 | 博士2回生 |
| Julian ILHAM | 学術情報メディアセンター | 中村研究室 | 研究員 |
| 井藤 隆秀 | その他の専攻・大学 | 理化学研究所 | 技術職員 |
| 下西 慶 | 学術情報メディアセンター | 中村研究室 | 助教 |
| 近藤 一晃 | 学術情報メディアセンター | 中村研究室 | 准教授 |
| 中村 裕一 | 学術情報メディアセンター | 中村研究室 | 教授 |