針對衰弱老人或是中風患者的鍛鍊,主要靠著一些物理復健機制,例如與復健器材的互動來達到刺激患部肌肉的功能,但復健成效的評估仍大部份仰賴人力進行,如何應用目前逐漸成熟的影像處理軟硬體讓動作的評估能夠自動化地進行是一項重要的課題。 本研究擬建置一套動作評估系統,自動擷取肢體動作之特徵點後,整合人體骨架模型及深度影像資訊,並應用影像處理技術的模版比對功能自動標記關節位置。在動作正確性評估方面,將採用動態時間校準演算法進行測試動作與標準動作資料庫之比對,並分別針對動作完成度及正確性進行實驗。 以5項衰弱體適能運動為研究對象,透過本研究所開發之關節點3D位置的自動標記程式,以及深度影像資訊校正程序所進行的實驗結果顯示,自動標記系統正確率可達到約99%,而在動作判別正確性則可達100%,顯示本研究結果具備衰弱動作判斷的可行性及應用價值。
Currently, physical exercises for the frail elderly or stroke patients relying on physical rehabilitation, such as interaction with equipment to stimulate their muscle function. The effectiveness evaluation of rehabilitation depends mainly on human judgment. How to apply the mature techniques of image processing to automatically assess the effect of rehabilitation movement is an important issue. This study builds a system for movement assessment, it is able to detect and retrieve the features of human body joints. Integrated with human skeleton model and depth image information, image processing techniques is applied to tag the 3-dimentional joint positions. The retrieved position coordinates are used to build the movement trajectory database. In the assessment of the correctness of the action, the dynamic time warping algorithm was used to compare the performed action with the standard ones in the metrics of correctness and completeness. The experiments were conducted for five types of frail movements with the developed auto-tagging and depth-adjustment techniques. The results showed that our system achieved a correct rate of 99% for joint position tagging. The correctness of movement type judgment was 100%. The result encourages us that the proposed approach is suitable for frail movement judgment with potential application value.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。