Currently, traditional sports equipment and technical solutions in higher education institutions are unreasonable, and outdated management models for sports training fail to provide scientific guidance for badminton athletes, resulting in suboptimal overall training outcomes. This paper develops a comprehensive contactless motion recognition system based on deep learning algorithms. The system utilizes the Microsoft Kinect V2 smart camera to accurately capture the threedimensional spatial positions of human skeletal joint points in real time and convert them into motion data streams. Additionally, the DTW algorithm is used to calculate the joint angle differences between standard motion sequences and test motion sequences, and an action evaluation formula is defined to assess core strength training movements. After testing, the system’s accuracy in evaluating movement results ranges from 90.20% to 94.69%, enabling effective assessment of core strength training movements. After applying the system to badminton core strength training, students’ foot flexibility significantly improved compared to conventional core strength training (P < 0.05). Therefore, teachers can actively apply the motion recognition system in core strength training when conducting badminton instruction.