Tracking and decomposition of throwing and jumping movements in high level figure skating based on deep learning

In order to overcome the problems of high average noise and poor decomposition accuracy of throwing jump in traditional motion tracking decomposition methods, this paper proposes a new high-level figure skating throwing jump motion tracking decomposition method based on deep learning. The average depth of the key frame of the throwing action image is calculated, and the average depth is input into the depth learning neural network for training. According to the training results, the depth image is regularised to track the throwing action. According to the tracking results, the AHP judgment matrix is given, and the target trajectory characteristics of figure skating throwing jump are obtained, and thereby the decomposition of high-level figure skating throwing jump is completed. The experimental results show that the mean noise of the designed method is 0.05 dB, and the decomposition ability is higher. Copyright © 2023 Inderscience Enterprises Ltd.