以交互作用扭矩觀測器輔助表面肌電圖動作辨識控制應用於上肢復健外骨骼機器人

Abstract

中風是一種突發性腦血管疾病，造成患者功能失調。臨床研究顯示，若中風患者進行長期的復健治療，能夠有效幫助患者恢復原有的運動功能。復健機器人不僅可以提供患者更穩定的復健療程、減輕治療師的負擔，也可以透過機器上的感測器獲得客觀資料作為評分的依據。此外，在復健療程的過程中，患者可能會發生失能性痙攣的症狀，此時機器人若持續移動，反而會對患者造成二次傷害。因此，將偵測患者痙攣狀態的方法納入基於外骨骼的復健控制策略有其必要性。 在相關的文獻中，對於肌力不足的患者，使用交互作用扭矩觀測器或肌電訊號的控制策略來進行復健療程是有較高的難度。本研究針對上肢復健機器人，提出一個結合肌電訊號以及交互作用扭矩觀測器的控制策略。首先，本研究提出的運動意圖辨識模型會透過表面肌電訊號預測使用者上肢的運動方向，並結合先前提出的交互作用扭矩觀測器獲得使用者運動意圖。接著，根據主動肌與拮抗肌的肌電訊號來評估患者運動過程中的痙攣狀態是否異常。最後，本研究提出一個整合運動意圖以及痙攣狀態異常評估的控制策略。 本研究提出的方法在四位健康受試者上做測試，其結果顯示，結合兩種運動意圖的控制方法可以提升主動控制的表現。此外，根據主動肌與拮抗肌之肌電訊號所計算出來的指標也可以有效區分痙攣狀態的異常。

Stroke is a sudden cerebrovascular disease resulting in dysfunction. Clinical research shows that if stroke patients undergo long-term rehabilitation treatment, it can effectively help patients restore their original motor function. Rehabilitation robots can provide patients with more stable rehabilitation therapy, reduce the burden of therapists, and obtain objective data through the sensor on the machine as the basis of scoring. In addition, during the rehabilitation therapy, the patient may have disabling spasticity. If the robot continues to move, it will cause secondary injury to the patient. Therefore, it is necessary to incorporate detecting state of spasticity into the rehabilitation control strategy based on the exoskeleton. In the related literature, it is difficult for patients with muscle weakness to carry out the rehabilitation therapy with the interactive torque observer or the electromyography signal. In this research, a control strategy combining electromyography signal and interactive torque observer is developed for the upper limb rehabilitation robot. Firstly, the motion intention recognition model proposed in this research will predict the motion direction of the user’s upper limb through the surface electromyography signal and obtain the user’s motion intention combined with the previously proposed interactive torque observer. Then, a mechanism will evaluate the electromyography signals from antagonistic muscle pairs to prevent the disabling spasticity. Finally, this research proposes a control strategy integrating exercise intention and muscle co-contraction evaluation. The method proposed in this research is verified through experiments with four healthy subjects. The results show that the combination of two motion intentions can improve the performance of the active control. In addition, the indexes calculated from the antagonistic muscle pairs’ electromyographic signals can also effectively identify the abnormalities of the state of spasticity.