- 學位論文
基於影像感測資訊量化復健動作之研究
Quantitative Investigation of Rehabilitation Movements from Images Data
摘要
中風是目前高齡者常見的疾病之一，患者的治療及復健均需花費大量的社會資源。中風發生內三個月，神經及運動能力恢復最快，早期介入復健評估與治療，可提高復健之療效。目前在施行中風復健治療計畫前，需針對中風患者上、下肢動作恢復與行動能力施予評估，其過程費時費力且缺乏量化客觀的評估方式。 本研究透過建置感測環境並擷取數位動作影像，針對影像資訊求取特徵參數，以描述動作軌跡變化；運用K-means分析，將參數化資料進一步分群並制定碼本；模型化復健動作並找出復健動作之相關運動學資訊，藉以量化復健動作；最後透過動態規劃演算法評估量化碼本相似性。 根據實驗的相似度比對正確率結果得知，本研究所提的方法在Top5的正確率可達87%以上，表示在Top5以內大部分可找到其正確的對應動作，證實本研究所提之以動作碼本序列表示各關節點動作方法，具有其可行性。
並列摘要
Current stroke rehabilitation treatment plan needs to assess patient’s motion ability of upper and lower limbs. The assessment is time-consuming and criticized being lack of quantitative and objective standard. This study utilizes image sensor to capture the data of rehabilitation action. The track of motion is featured and transformed by frame-based analysis. K-means clustering is used to group the features and to form codebooks for the quantization and simplification of rehabilitation assessment. The similarity matching is performed using the dynamic programming algorithm. According to the experimental similarity matching results, the proposed approach achieved 87% correctness for Top5 similarity matching. The result verifies that the proposed method is feasible for describing the variation of rehabilitation actions through the combination of different codeword sequences.