下肢截肢者與退化性關節炎患者在日常行走及復健過程中，需長期進行步態 訓練與功能評估。然而臨床上對於復健成效的判斷，多仰賴醫師經驗與肉眼觀察， 缺乏客觀且量化數據支持。為提升步態評估的客觀性，提出一套能夠擷取步態的 裝置與系統，並開發資料修正與補償演算法，以準確擷取下肢關節運動資訊。 本研究建立一套由多顆無線慣性感測單元組成之量測系統，透過Wi-Fi即時 傳輸慣性資料至電腦端，並分別利用LabVIEW®與MATLAB®建立訊號接收及資 料分析之人機操作介面。透過 Madgwick 演算法結合零速更新法(Zero Velocity Update， ZUPT)進行姿態估測，進而計算步態參數，並針對臨床上可能發生的感測 器安裝錯誤，提出(1)座標錯置修正演算法(2)安裝傾斜補償演算法，以修正感測 器因座標錯置或安裝傾斜所造成的角度估測誤差。 為驗證所提出演算法之有效性，分別利用高精度三軸電動旋轉平台與工業機 械手臂模擬人體關節運動。實驗結果顯示，在模擬腰椎、大腿、小腿與腳背的情 境中，未經演算法修正前之角度誤差最大高達49°以上，經演算法修正後之角度 誤差皆可有效降低至1°以下，部分情況更可達 0.5° 以下，顯示本研究方法具備 良好之穩定性與精確度。 最後，將本系統實際應用於義肢使用者與退化性關節炎患者之直線步行測試 實驗，比較經安裝傾斜補償演算法前後，腰椎、髖、膝與踝關節屈曲角度的差異 情形，對於膝下截肢者，健側膝關節屈曲角度於傾斜補償後大於患側，而膝上截 肢者，健側膝關節屈曲角度於傾斜補償後小於患側，退化性關節炎患者的髖關節 角度於補償前、後最大差5°，膝關節為18°，腰椎角度皆為3°以內，腰椎屈曲角 度於傾斜補償前有較多屈曲次數且屈曲角度較小，傾斜補償後屈曲次數減少且屈 曲角度變大，使得腰椎屈曲更明顯。;Lower-limb amputees and patients with degenerative joint diseases require long-term gait training and functional assessment during daily walking and rehabilitation. However， clinical evaluation of rehabilitation outcomes still relies heavily on physicians’ experience and visual observation， lacking objective and quantitative data support. To enhance the objectivity and accuracy of gait assessment， this study proposes a wearable gait acquisition device and analysis system， along with data correction and compensation algorithms， to accurately capture lower-limb joint motion information. In this study， a measurement system composed of multiple wireless inertial measurement units (IMUs) was developed to transmit inertial data to a computer in real time via Wi-Fi. Human–machine interfaces for signal acquisition and data analysis were implemented using LabVIEW® and MATLAB®， respectively. The system employs the Madgwick orientation estimation algorithm combined with a Zero Velocity Update (ZUPT) method to estimate segment orientations and subsequently calculate gait parameters. In addition， to address sensor installation errors that may occur in clinical applications， two correction algorithms were proposed: (1) a coordinate misalignment correction algorithm and (2) a mounting inclination compensation algorithm， which effectively compensate for angular estimation errors caused by coordinate misalignment or sensor inclination. To verify the effectiveness of the proposed algorithms， human joint motions were simulated using a high-precision three-axis motorized rotation platform and an industrial robotic arm. Experimental results demonstrate that， in simulated scenarios involving the lumbar spine， thigh， shank， and foot， the maximum angular error exceeded 49° before algorithmic correction. After applying the proposed algorithm， the corrected angular errors were successfully reduced to below 1°， and in some cases even below 0.5°， indicating that the proposed method achieves high stability and accuracy. Finally， the proposed system was applied to straight-line walking experiments involving prosthesis users and patients with degenerative joint diseases. Differences in the flexion angles of the lumbar spine， hip， knee， and ankle joints before and after applying the mounting inclination compensation algorithm were analyzed， demonstrating the feasibility of the proposed system for clinical gait analysis applications. For individuals with below-knee amputation， the knee flexion angle of the intact limb was greater than that of the amputated limb after inclination compensation. For individuals with above-knee amputation， the knee flexion angle of the intact limb was smaller than that of the amputated limb after inclination compensation. In patients with degenerative osteoarthritis， the maximum difference in hip joint angles before and after compensation was 5°， whereas the knee joint showed a difference of 18°. Lumbar spine angle differences were within 3° in all conditions. Before inclination compensation， the lumbar flexion angle had a large number of flexion cycles and a small flexion angle. After inclination compensation， the number of flexion cycles decreased and the flexion angle increased， making the lumbar flexion more obvious.
