Artificial intelligence-based Bayesian optimization and transformer model for tennis motion recognition

• Published in: Journal of Applied Science and Engineering Vol. 29; no. 1; pp. 171 - 178
• Main Authors: Shaowei Shi, Kun Huang
• Format: Journal Article
• Language: English
• Published: Tamkang University Press 2026
• Subjects: artificial intelligence; bayesian optimization; key point recognition; tennis motion recognition; transformer model
• ISSN: 2708-9967; 2708-9975
• DOI: 10.6180/jase.202601_29(1).0017

Because the traditional methods are used to analyze human motion behavior, there are large errors and serious over-fitting phenomenon, so a novel tennis motion recognition based on Bayesian optimization and transformer model is proposed in this paper. First, we use an improved generative adversarial network to optimize heat map location detection of human key points. A human key point recognition algorithm is designed based on Transformer. Then, the optimal pruning rate of each layer of the network is found by using Bayesian optimization algorithm to improve the efficiency and accuracy of subnet search. Finally, the proposed method is tested on several mainstream behavior recognition datasets. The results show that the recognition accuracy rates with proposed method on UCF101, HMDB51 and Something-SomethingV1 datasets are 97.6%, 73.8% and 66.7%, respectively, when only RGB video frames are used as input. It can be seen that the proposed method can efficiently extract the spatio-temporal features of motion.