Sports Action Recognition Method Integrating Pose Feature and Dynamic Time Warping(DTW)
JIN Xiangru
ZHANG Shengli
LIU Wei
Abstract：The current intelligent mode of physical education teaching is developing rapidly.However,relying solely on manual observation by teachers to identify students'sports actions during the teaching process can easily lead to issues such as untimely feedback and subjective evaluations.This study therefore proposes a sports action recognition method that integrates pose features with the dynamic time warping(DTW)algorithm.Based on the skeletal point feature output from OpenPose,the method fuses features such as the center of gravity,limb angles and orientations of students during sports activities and uses DTW for alignment and evaluation.Experimental results show that,of the 400 action samples in the training dataset,the OpenPose+DTW model correctly identified 372 samples with a total recognition rate of 93%,which is higher than that of other models.In an experiment involving 50 actions in broadcast gymnastics teaching,the OpenPose+DTW model misjudged five samples,achieving a recognition accuracy of 90%.These results suggest that the sports action recognition model integrating pose features and the DTW algorithm exhibits excellent performance and can support online physical education teaching scenarios.
Keywords：OpenPosedynamic time warping(DTW)algorithmphysical education teachingaction recognition
Publication Date：2025-09-30
Online Publishing Date：2025-10-17(First online date of this platform, not the publication date of the document)
Pages：7( 80-86 )