Stance Recognition Applied to Chinese Martial Arts Long Stick Combat Sports

The popularity of sports intelligence is increasing, and Chinese martial arts should also explore this area. This study utilized Mediapipe for stance recognition, Pycharm 2022 as the compiler, and Python programming language for program design. An LSTM network was built using TensorFlow to train the detection model of the straight stabbing action of long soldiers based on 200 pictures. The detection model was applied to actual teaching work to measure the technical movement standards of beginner baguette fighters. The recognition accuracy of the model was 92% when compared to traditional visual inspection by five coaches. The use of stance recognition in Chinese martial arts has the potential to improve the practice and understanding of this martial art.