基於人體姿勢估計及壓力感測器數值之瑜珈動作分析系統

Abstract

近年來因應新型冠狀病毒（Coronavirus disease 2019, COVID-19）的爆發，居家運動風潮隨之興起，瑜珈(Yoga)這項運動不受地點限制且不需要特殊的運動器材，對於舒緩因居家上班與上課導致的緊繃肌肉有相當好的效果，因此瑜珈成為許多人居家運動的首要選擇。瑜珈這項運動講求不光是要保持姿勢標準，保持身軀平衡也是一大重點，但對於瑜珈初學者在家自主練習時，只透過範例影片做練習往往無法抓到該姿勢的重點，導致練習成效不佳，沒有訓練到正確的部位。因此本研究透過影像與壓力感測器量化瑜珈教練的姿勢數據，藉此分析使用者的姿勢。首先透過影像分類器分辨使用者的姿勢，最後評估出姿勢標準度、身軀穩定度與重心穩定度三個要點，不但可以幫助使用者在家自主練習，也可輔助瑜珈教練的教學。

In recent years, home exercise become popular because of the Coronavirus disease 2019 outbreak. Yoga is not restricted by location and does not require special exercise equipment, and it is very effective in relieving tight muscles caused by work or class from home.However, for beginners practicing yoga on their own at home, they of can’t get the key points of the yoga posture by only watching the example video, result in the poor training result. Therefore, our study propose a system to quantify the yoga coach's posture data through video and pressure sensors to analyze the user's posture. The first step is to identify user’s pose through the image classifier, and finally to evaluate the three points of posture: Pose Accuracy, Body Stability, and Center of Pressure Stability, which not only help the user to practice yoga at home, but also assist the yoga coach's teaching.