인공지능을 활용한 저비용 스포츠동작 분석의 검증

Abstract

We projected to calculate and verify the flexion angle of the lower extremity joint during sports motion using low-cost artificial intelligence modeling. The data modeled by artificial intelligence was verified to the data which was produced from the IMU(Inertial Measurement Unit). We recruited seven youth male football players as participants for the project. The participants performed kicking motion of football with IMU attached to the landmark of the lower extremity joint for 3D motion analysis. The wireless data from the IMU and the video data from the general camera were measured simultaneously. The flexion angles of the pelvis, knee, and ankle were computed from the OpenPose library and the IMU data. The OpenPose library was processed on the image data from the camera. The football kicking motion was calculated over E1 (Toe Off), E2 (TOP Of Back Swing), E3 (Ball Impact), and E4 (Follow Swing) based on the kicking foot. For statistical processing, Intraclass Correlation Coefficient(ICC2,1) was performed to verify the consistency of AI modeling data. And Correlation(both, Pearson) analysis was performed to evaluate the accuracy of the data from the AI. The statistical significance level was set to .05. In the result, it showed excellent reliability and accuracy of the knee joint angle which occurs at the moment over the kicking preparation and kicking the ball(p<.05). Furthermore, It was possible to derive the characteristic trend of knee flexion angle change during football kicking. In conclusion, if image data with high-resolution and high-speed shutter speed can be obtained, quantitative kinematic data analysis might be available from artificial intelligence modeling.