Skeleton-Joint Angle-Based Tennis Human Pose Correction Method

In order to improve the training effect, a tennis swing correction method based on the angle characteristics of bone points is proposed. Firstly, 25 skeletal points of human body are extracted by OpenPose, and an included angle is constructed for every three skeletal points. Then, according to the connectivity of skeletal points in human body parts, 10 candidate included angles are selected from 13,800 angles. Finally, through Pearson correlation analysis, six angles, such as left shoulder, right shoulder, left elbow, right elbow, left knee and right knee, are determined as the key features of tennis swing. Secondly, according to six key angles, the K-Means clustering analysis is carried out on each frame of tennis training video, and five key frames, such as the highest point of backswing, the lowest point of backswing, the hitting point, the highest point of swing and the lowest point of follow-swing, are determined. Then, the six key angles at the moment of hitting the ball are clustered and divided into three clusters, which correspond to three different hitting heights: high, medium and low. Finally, the range of the included angle of human skeleton points in each key frame when professional players deal with three kinds of ball heights is counted, and the swing correction basis is provided for amateur players' training through distance similarity. This method can provide scientific and personalized support for players' technical improvement.