Title: Sports athletes' action identification based on dynamic Bayesian network

Authors: Jinrong Li; Hui Wang; Luojing Wang

Addresses: Department of Mathematics and Information Engineering, Puyang Vocational and Technical College, Puyang, 457000, China ' Xinxiang Vocational and Technical College, No. 6, JingSan Road, JingKai District, Xinxiang 453000, China ' Shangqiu Polytechnic, Shangqiu, 476000, China

Abstract: In order to overcome the problems of poor accuracy, recall rate and poor real-time performance of traditional methods, the sports athletes' action identification method based on dynamic Bayesian network was proposed. Firstly, the shape vector is used to quantify the motion features of athletes, and the motion shape feature vector is projected, so that all vectors are in the same coordinate system, and then the feature data is preprocessed. Secondly, on the basis of feature data preprocessing, a phase space is constructed to extract athletes' action features. Finally, dynamic Bayesian network is used to identify athletes' action based on feature extraction results. Experimental results show that the accuracy of the proposed method is above 95%, and the recall rate is stable at about 9%, and all action identification can be completed in five minutes.

Keywords: dynamic Bayesian network; DBN; athlete; action identification; data preprocessing.

DOI: 10.1504/IJRIS.2022.126660

International Journal of Reasoning-based Intelligent Systems, 2022 Vol.14 No.4, pp.161 - 168

Received: 28 Dec 2021
Accepted: 14 Jul 2022
Published online: 31 Oct 2022 *

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