Research on Badminton Swing Motion Recognition Based on Improved DT Algorithm

In the rapid development of computer technology, the field of motion analysis faces many challenges. Aiming at the problem that it is difficult to combine spatio-temporal feature capture and computational efficiency in badminton swing recognition, the study proposes a badminton swing recognition model based on the improved dense trajectory (Improved DT, IDT) algorithm. The dense trajectory algorithm is optimised by region-constrained feature extraction, dynamic non-constantlength trajectory management and multimodal feature enhancement coding, in order to achieve an action recognition method that combines efficiency, robustness and interpretability. Experiments show that the recognition accuracy and recall of the improved algorithm are $93.62 \%$ and $92.46 \%$, respectively, which are improved by $6.13 \%$ and $8.62 \%$ compared with other algorithms. The average processing time is 27.89 ms, which is $42.15 \%$ lower than other algorithms. The F1 scores were kept above $\mathbf{9 1 \%}$ in fine-grained actions such as dunking and lofting. The results show that the recognition model proposed in the study can accurately and quickly identify different types of swing movements, which provides powerful technical support for the technical training and movement analysis of badminton players, and has high application value and promotion prospects.