Title: An action recognition of track and field athletes based on Gaussian mixture model

Authors: Qin Yang; Zhenhua Zhou

Addresses: School of Physical Education, Hunan City University, Yi'yang, 413000, China ' School of Physical Education, Hunan Normal University, Chang'sha, 410012, China

Abstract: To solve the problem of low recognition accuracy caused by the complexity of individual actions in track and field in the past, a method of action recognition for track and field athletes based on Gaussian mixture model was proposed. First, the data is analysed by the interaction of spatiotemporal features. Secondly, a low-pass filter is used to eliminate the impact of noise on the data to reduce the calculation loss. On the basis of pre-processing data, Hilbert Huang transform (HHT) was used for feature extraction to capture and understand athletes' motion features more accurately, thus significantly improving the accuracy of movement recognition. Then, the Gaussian mixture model is used to model the characteristic parameters, determine the number of mixed components and initialise the model parameters, and complete the movement recognition of track and field athletes. The experimental results show that the traditional method has high computational loss and low recognition accuracy, while the proposed method has very low computational loss and the highest recognition accuracy can reach 98%. The comparison shows that this method has the advantages of low computational complexity, high accuracy and good recognition performance.

Keywords: Gaussian mixture model; GMM; interaction of spatiotemporal features; action data; low pass filter; athlete movement recognition.

DOI: 10.1504/IJBM.2025.145914

International Journal of Biometrics, 2025 Vol.17 No.3, pp.292 - 310

Received: 21 Sep 2023
Accepted: 29 Mar 2024
Published online: 30 Apr 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article