Development of EEG Brain-Computer Interface System for Control of Shoulder-Elbow Rehabilitation Robot

腦波腦機介面控制肩肘復健機器人之發展

朱銘祥(Ming-Shaung Ju) ； 林宙晴(Chou-Ching K. Lin) ； 林裕晴(Yu Ching Lin) ； 宋信毅(Shin-Yi Sung) ； 黃繪禎(Hui-Zhen Huang)

《中國機械工程學刊》 41卷1期 (2020/02) Pp. 39-47

摘要

In our previous work a shoulder-elbow rehabilitation robot was developed and applied for the rehabilitation of chronic stroke patients. The goal of this study is to integrate an EEG-based brain-computer-interface (BCI) and the rehabilitation robot to build a system by which patients can use imagery movement of arm to control the robot to assist forward reach movement of the arm. Two personal computers were employed, one for EEG processing and the other for controlling the robot. An optimal filter was realized to reduce noise in EEG and the algorithms to translate mu waves of C3 and C4 of human brain into robot command were proposed. Eight healthy and five chronic stroke subjects were recruited to test functions of the system. Two indices, namely accuracy and trigger time were utilized to evaluate performance of all subjects. The training lasted for eight weeks with two days per week. The results show the healthy subjects had no side difference on weekly accuracies. However, for the affected arm of stroke patients accuracy of the fourth week is significantly higher than that of the first week. Trigger time of the intact arm of the stroke group at eighth week is smaller than that of third week. The separation of EEG processing and robot command generation does improve quality of EEG and the EEG controlled rehabilitation robot might be used in future neuro-rehabilitation of patients.

關鍵字

stroke ； rehabilitation robot ； electroencephalogram ； brain-computer-interface

並列摘要

本團隊曾發展肩肘復健機器人並用於慢性期中風病人復健，本研究目的為整合腦波腦機介面系統和此復健機器人，發展以想像動作控制機器人以帶動病人上臂進行復健。本研究利用兩部個人電腦和最佳化濾波器降低腦波雜訊並提出由大腦C3，C4的μ波將受測者想像轉為驅動機器人的演算法。以8名常人和5名慢性中風病患測試系統功能。為評估受測者的表現定義了兩種性能指標：準確率和啟動時間。每位受測者共接受每周兩天共8周的訓練，實驗結果顯示：常人訓練期間周間準確率差異不大，而中風病人患側第4周平均準確率明顯比第1周高，雖然第5周起有學習高原現象，但第8周平均準確率仍高於第1周。常人非慣用側和病人健側第4周啟動時間均明顯低於第1周，且病人患側第8周啟動時間小於第3周。本研究提出的方法的確能改善腦波的品質且系統具有神經復健的功能。