Title: Recognition of foul actions of football players based on self-attention mechanism

Authors: Yingdong Song

Addresses: Department of Physical Education and Military Education, Jingdezhen Ceramic University, Jingdezhen, 333403, China

Abstract: In order to solve the problems of low recognition accuracy and long recognition time in traditional methods, a recognition method of foul actions of football players based on self-attention mechanism was designed. Firstly, the image of football player is obtained by Kinect device and the image features are extracted. Secondly, the background of action image is separated by background clipping method, and the noise of motion pixel point eight neighbourhood is removed by smoothing noise reduction method. Finally, set a set of key value pairs for foul actions, encode them using the encoder in the self-attention mechanism, calculate the probability of foul actions occurring, and decode the dependency relationship of the encoded foul actions to determine whether the action is a foul action and obtain relevant recognition results. The test results show that the proposed method can improve the accuracy of foul action recognition and has a good effect.

Keywords: self-attention mechanism; football players; foul actions; recognition; background clipping method; encoder.

DOI: 10.1504/IJRIS.2025.147456

International Journal of Reasoning-based Intelligent Systems, 2025 Vol.17 No.3, pp.209 - 216

Received: 11 Apr 2023
Accepted: 23 May 2023
Published online: 16 Jul 2025 *

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