- 學位論文
Kinect 姿勢辨識對運動訓練之應用-以深蹲為例
Kinect posture recognition to sports training application - taking squat as an example
摘要
近年來，全球人類的平均年齡逐年增加，人們更加關注自己的健康使得健身的更為普及。而眾多運動方式中最受歡迎的健身姿勢之一便是被稱為運動之王的深蹲。因其運動的全身性，錯誤的深蹲往往會導致嚴重的運動損傷。如何有效保持健康，避免錯誤的運動傷害已成為健身訓練的主題之一。今日有許多運動訓練方法，然而，這些方法多半不能以客觀和基於數據的方式訓練，例如，由健身教練進行訓練。亦有使用昂貴且位置受限的方法，如動態捕捉，高速攝像機......。這種方法無法客觀地給健身初學者提供良好的訓練。 在這項研究中，我們將使用RGBD相機Kinect設計一姿勢捕捉和分析系統。 Kinect簡單而便宜，憑藉其基本的成像能力，我們可以協助深蹲訓練並避免運動傷害。大多數運動損傷是由不穩定的質心引起的。這可能導致身體部分肌肉組織和骨骼超負荷。不同的人在執行相同的動作時會有不同的質心。因此，在本研究中，我們將首先用數學模型分析人體的質心，然後確定合格的姿勢而不造成運動損傷。因此，在本研究中，我們將首先分析人類質心的數學模型。然後定義一個不會造成運動傷害的位置。在這項研究中，我們將使用Kinect捕捉人體骨架並設計一個簡單的UI。令用戶在下蹲時獲得關於他們自己的運動和身體關節的坐標的視覺反饋。我們使用坐標信息來繪製坐標和時間之間的關係，以便在運動過程中查看姿勢以幫助用戶進行深蹲訓練。
並列摘要
In recent years, the average age of global human beings has been increasing year by year. People pay more attention to their health. This led to the popularity of fitness. One of the most popular fitness poses is the squat, which is known as the king of exercise. Because squats are a full-body movement, the wrong squats can often lead to serious sports injuries. How to keep fit effectively and avoid wrong sports injury has become one of the topics of fitness training. There are many sports training methods available today. However, some methods cannot be trained in an objective and data-based way, for example, training by a fitness instructor. Some use expensive and location-constrained methods, such as dynamic capture, high-speed cameras.... This kind of method cannot give the fitness beginner the good training objectively. In this study, we will design a posture capture and analysis system with RGBD camera Kinect. Kinect is simple and inexpensive, and with its basic imaging capabilities we can assist in squat training and avoid sports injuries. Most sports injuries are caused by unstable center of mass. This can lead to overloading of parts of the body's muscle tissue and skeleton. Different people will have different centers of mass when they perform the same set of movements. Therefore, in this study, we will first analyze the center of mass of human body with mathematical model, and then define the qualified posture without causing sports injury. Therefore, in this study, we will first analyze the mathematical model of human centroid. Then define a position that does not cause athletic injury. In this study, we will use Kinect to capture the human skeleton and design a simple UI. This allows the user to get visual feedback on their own movement and the coordinates of the body's joints when squatting. We use coordinate information to plot the relation between coordinate and time to view the postures during the movement. These can assist users in squat training.