基於穿戴式感測技術之智慧健身系統

Abstract

為了避免受傷的情況發生並可以有效的健身，正確的健身動作對於健身是很重要的。然而聘請專業教練的價格昂貴，所以很多人選擇在家裡看線上的影片來學習。但是學員如果只是在家看影片的話，無法得到教練的建議，也不會知道自己的動作到底有沒有正確。 本研究開發了一個基於穿戴式感測技術之智慧健身系統，可以幫助使用者在任意時間地點運動並可以得到建議與評估。我們開發動作計次演算法和動作切割演算法來截取使用者動作特徵。其中計次演算法的失誤率約為1.9 %。截取到的動作特徵可以用來幫助教練建立教練模型。最後，我們還開發了一套教練建議演算法，可以幫助使用者評估運動的品質並給予建議。

To avoid possible physical injury and achieve the desired results, it is essential to exercise with the correct movements and postures. However, the cost is high to frequently consult professional fitness coaches. Instead, online tutorial videos are one of the most popular alternatives. But learners may not aware of having wrong actions by only watching the videos. In this work, a fitness e-coaching system based on wearable sensing technology that gives timely advices besides watching videos is developed. Algorithms including repetition counting and activity segmentation are developed to extract motion features. The error rate of the counting algorithm is 1.9%. The extracted information can be used to build movement models, especially to build reference model from coach’s dataset. Last, a coaching method that incorporates the coach model is developed to analyze the quality of exercise then give advices. To present the result, an App is implemented. Users first choose the exercise they want to do, then a tutorial video with guides of how to do the exercise correctly will be provide to the users. Last, users do the exercise with wearable sensors and get information and advices about the exercise.