請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/7541
標題: | 基於深度學習實現運動預測控制上肢復健外骨骼機器人 Deep Learning based Motion Prediction for Exoskeleton Robot Control in Upper Limb Rehabilitation |
作者: | Jia-Liang Ren 任嘉梁 |
指導教授: | 傅立成 |
關鍵字: | 復健機器人,手臂姿態,肌電訊號偵測,機器學習,主動控制,引導控制,NTUH-II, rehabilitation robotics,arm dynamics,EMG sensing,machine learning,active control,guide control,NTUH-II, |
出版年 : | 2018 |
學位: | 碩士 |
摘要: | 因中風導致神經麻痹是種很常見的疾病,臨床症狀常見為肌力不足、肌肉痙攣及無法自主控制關節活動等病徵。通過長期反覆性的復健治療後,能夠有效幫助患者恢復原有運動機能,並且可以防止發生二次併發症。以機器人輔助提供上肢復健,可以為患者提供更好的復健療程同時減少治療師的負擔。本研究利用慣性測量單元還原手臂運動的姿態,並結合肌電訊號來訓練深度學習模型,預測使用者手臂想要動作的位置(方向和速度),達到機器人主動控制與引導控制。
於相關的文獻中,使用力/力矩感測器或肌電訊號的控制方式建立人機互動模型在多軸主動控制上有較高的難度。本研究提出的深度學習模型架構與傳統模型以及其他深度模型架構相比,擁有更高的準確率且對不同受試者的影響較小。此模型更是可以透過少量的數據對特殊的病人進行微調,以實現更好的結果。 本研究提出的方法經過三位健康受試者在線測試,並實現在上肢復健外骨骼機器人NTUH-II上。實驗結果顯示本研究提出的方法於復健任務中的表現優於相關的研究。此外,本方法能夠簡單的擴展為各種不同復健療程。 Neural paralysis due to stroke is a common disease, and clinical symptoms are often characterized by insufficient muscle strength, muscle spasms and inability to control joint activity. Long-term repetitive rehabilitation treatment can effectively help patients to restore their original motor function and can prevent the secondary complications. Robot-assisted upper limb rehabilitation can provide patients better rehabilitative treatment while reducing the burden on the therapist. In this study, the inertial measurement unit is used to estimate arm dynamics and is combined with muscle electromyography to train deep learning model for human arm joint angles prediction. This model can be applied to the active control and guide control of the robot arm. In the relevant literature, the use of force/torque sensors or myoelectric signals based control has a higher difficulty in establishing a human-robot interaction model for active rehabilitation. In this thesis, a learning model is proposed. Compared with the traditional model and other architecture of deep learning model, the proposed model in this study has a higher accuracy rate and has less impact on different subjects. This model can be fine-tuned to adapt special patients through a small amount of data to achieve better results. The method proposed in this study was online tested by three healthy subjects and implemented on the upper limb rehabilitation exoskeleton robot NTUH-II. The experimental results show that it outperforms than relevant research works. In addition, the method can be simply extended to various rehabilitation therapies. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/7541 |
DOI: | 10.6342/NTU201802164 |
全文授權: | 同意授權(全球公開) |
電子全文公開日期: | 2023-08-02 |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-107-1.pdf | 8.39 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。