標題: 使用三軸加速度計與陀螺儀之異常步態偵測方法
A Novel Method for Abnormal Gait Detection Using 3-axis Accelerometer and Gyroscope
作者: 林旻萱
荊宇泰
Lin,Min-Hsuan
Ching,Yu-Tai
生醫工程研究所
關鍵字: 穿戴式裝置;步態;身份辨識;異常分數;Wearable device;Gait;Identity Verification;Anomaly score
公開日期: 2017
摘要: 本論文設計一個放置於腳後跟的穿戴式裝置,內部包含三軸加速度計與陀螺儀,透過此裝置收集的數值來做分析,以達到異常步態偵測的功能。資料會先轉換成一個時間相關的特徵,再利用此特徵計算異常分數並建立個人的步態模型,此模型可以用來區分新的資料是否為本人,平均準確率有93%。可以應用於觀察受測者是否產生異常步態,使其得以檢測自身的步態狀況。 實驗的結果說明本研究使用的特徵具有個人的辨識性,而且異常分數的計算方式將訓練資料的分布特性也考慮進來,相對於傳統的歐式距離有了更大的彈性。在未來的應用上可以隨著時間更新本身的訓練資料,成為一個有效辨識異常步態的模型。
This thesis presents a wearable device with an accelerometer and a gyroscope for detecting anomalies in human walking. First, the data is converted into a time-related feature. Then, we use this feature to calculate anomaly score and build a personal identification model.The accuracy can reach up to 93%. The experimental results show that the feature used in our thesis have personally identifiable information. In addition, the calculation of the anomaly scores takes into account the distribution of training data,so it has more flexibility than the Euclidean distance. For the future application, their own training data can be updated over time to be a valid identification of abnormal gait.
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070460404
http://hdl.handle.net/11536/141570
Appears in Collections:Thesis