基於表面肌電訊號量測運動意圖程度智慧主動和輔助控制復健外骨骼機器人

Abstract

上肢運動功能失常是許多中風患者常見的症狀，因為神經的受損容易引起肌肉無力，活動範圍受限等問題。經臨床研究顯示，經由頻繁的復健治療可以有效恢復運動功能。此外，針對不同嚴重程度的患者，如果能透過適合的復健模式，更能使療程具備效果。復健機器人加入療程，治療師的負擔不僅能減少，還能藉由感測器獲得運動的客觀數據，建立更適合病患的復健策略。因此人機互動控制以及具備多種復健模式的機器人有其必要性。 本研究提出一種基於偵測患者運動意圖程度協助調整復健模式的控制策略，並用於上肢復健的外骨骼機器人。首先，透過特徵模型將表面肌電訊號(sEMG)提取出有效的肌肉啟動訊號，並結合所提出的模型得到低噪的運動意圖程度訊號。接著，基於速度場理論建立理想復健速度軌跡，在輔助模式下，評估當前位置誤差與運動意圖給予適當的輔助速度；另外，在主動模式下，則會針對交互扭矩觀測器得到的主動速度來調整復健速度軌跡讓病患能自主進行復健。最後，提出適當的整合方式，能透過動作意圖並能切換適合的復健模式控制方法，使病人提高復健參與度，以及正確完成復健動作。 本研究提出的控制策略，透過四位健康受測者與兩位中風患者做測試，來驗證模式切換的合理性性以及不同模式的控制表現。並透過運動意圖回饋讓使用者能更加願意積極地完成復健動作。

Upper limb impairment is a common symptom in stroke patients, as nerve damage can cause muscle weakness, limited range of motion, and other issues. According to the result of Clinical studies, long-term rehabilitation therapy can effectively restore motor function. In addition, for patients with different severity levels, if suitable rehabilitation modes can be used, the treatment course can be more effective. The introduction of robots into the treatment course can not only reduce the burden of the therapist, but also obtain objective motion data through the sensors on the machine, so as to establish the more suitable rehabilitation strategy for patients. Therefore, human-computer interaction control and robots with multiple rehabilitation modes are necessary. In this research, we propose a control strategy based on the detection of the patient’s motion intention level to assist in adjusting the rehabilitation mode, which is applied to the upper limb rehabilitation exoskeleton robot. Firstly, effective muscle activation signals are extracted from surface electromyography (sEMG) signals through feature models, and use the proposed model to obtain low noise motor intention level signals. Second, after an ideal rehabilitation velocity trajectory based on the theory of velocity field, our system proposes an autonomous switching mechanism, which can switch the exoskeleton robot from assistive mode to active mode based on the subject's motion level. In assistive mode, the system evaluates the current position error and motion intention to provide appropriate assistive velocity. In active mode, the rehabilitation velocity trajectory will be adjusted based on the active velocity obtained by the interactive torque observer, which allows the patients to take the initiative for completing rehabilitation. By proposing such hybrid control, our system can improve patient engagement in rehabilitation and more appropriately complete rehabilitation actions. The control strategy proposed in this study is verified through experiments with four healthy subjects and two stroke patients. The results prove the rationality of mode switching and validate the performance of autonomous mode switching. In addition, users can be more willing to participate in the rehabilitation exercise regardless of their muscle condition at that time.