Classic action recognition of cross-country skiing sports combining ant colony and clustering algorithm

As one of the oldest sports events in world sports history, cross-country skiing involves basic techniques such as mountaineering, skiing, and gliding, covering multiple classic actions during the process.

To improve the current cross-country skiing training technology, this study designed a classic action recognition method for cross-country skiing that combines ant colony and clustering algorithms. This method mainly utilizes ant colony algorithm to classify and partition actions, and implements classical action recognition based on current data parameters. The research will improve the ant colony clustering algorithm compared to traditional K-means algorithm, hierarchical clustering algorithm, density clustering algorithm, and basic ant colony algorithm. Metrics such as the sum of squared intra-cluster errors are used to evaluate cluster tightness for validation analysis.

The results show that the proposed method achieves 92.3% accuracy in recognizing classical actions with a maximum improvement of 2.9% compared to the K-means algorithm. Meanwhile, the average percentage error value of this method is only 0.62. In the action recognition of athletes at different levels, the convergence decreased by 19%.

From this, it can be concluded that the proposed method has high action recognition ability and recognition accuracy, which can effectively improve the recognition effect of cross-country skiing.