開發一利用慣性感測器與肌電訊號分辨日常生活與跌倒的偵測系統

Abstract

本研究以實驗評估肌電訊號於跌倒辨識之可行性，主要目標為開發一結合慣性感測器與肌電訊號(electromyography)量測系統之跌倒辨識系統，期能提高現有穿戴式跌倒辨識技術的速度與準確率，未來用以結合跌倒防護系統降低跌倒傷害的發生率或嚴重性。 於實驗室環境中同步利用光學式動作擷取系統擷取運動學資訊，以及肌電訊號量測系統取得肌肉活化程度，紀錄受測者進行日常生活動作（例如坐到站、彎腰檢物與行走等）與行走中不預期遭絆倒之資料，用以建立跌倒判別之數學模式。僅使用運動學資訊（速度與加速度）之跌倒辨識率約為85%，而單獨使用肌電訊號之跌倒辨識率則約為70%，然而使用肌電訊號之辨識速度快於使用運動學資訊。除此之外，合併使用運動學與肌電訊號辨識跌倒，可將跌倒辨識率提升至95%，同時辨識速度也有相當的提升，相較於單獨使用運動學或肌電訊號資訊，有較佳之跌倒辨識率與較快之辨識速度。此一成果證實了肌電訊號於跌倒辨識之潛力，以及結合慣性系統取得運動學資訊以辨識跌倒的應用方式。

Falls are leading cause of unintentional injuries and deaths, especially in the elderly. To detect falls early and accurately, and activate fall protecting devices in time, is important and could reduce fall-related socioeconomic cost. Kinematic variables obtaining by inertia sensors, such as accelerometers and gyroscopes, have been used to distinguish falls from activities of daily living (ADLs). Using inertial sensors could detect a fall in 300-400 ms after the fall initiation. However, by using electromyography (EMG) of selective muscles, it is possible to reduce this time to ~200 ms. The purpose of this study is to examine the possibility of combing inertial and EMG sensors to develop a fast and accurate fall detecting device. Laboratory simulated trip falls were corrected identified with 85% accuracy by kinematic data alone, with 70% accuracy by EMG alone, and 95% accuracy by using both kinematics and EMG. Among the corrected identified falls, the fall detect time was significantly shorter when using EMG as compared with using kinematic variables alone. Therefore, EMG sensors could be combined with inertial sensors for developing a faster fall detecting device.