A skeleton-based method and benchmark for real-time action classification of functional movement screen. (September 2022)

• Record Type: Journal Article
• Authors: Wenbo, Wang; Chongwen, Wang
• Is Part Of: Computers & electrical engineering. Volume 102(2022)
• Publication Date: 2022-09
• Subjects: FMS -- Functional movement screen -- Action classification -- Pose estimation -- FMS action evaluation; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.108151
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: FMS stands for functional movement screen, which is a simple and effective method to evaluate athletes' basic sports ability. This paper proposed a real-time FMS action classification method. Correspondingly, a video set was constructed with two different perspectives including 8 testers, 13 independent testing processes and 360574 images. Moreover, a normalization algorithm and a result correction algorithm is proposed to improve the performance of the models and make the result sequence more continuously. Furthermore, it has the vital significance to FMS action evaluation. Finally, this paper analyzed the effectiveness of different models and compared their performance from the aspects of accuracy, continuity, running speed, etc. The experimental results show that our method can achieve 96.7% precision score and 94.7% recall score on the test set. And the average running speed of the system is over 30 FPS. All related data, benchmarks and codes will be uploaded on https://github.com/bobogo/FMS-evaluation-system . Graphical abstract: Highlights: The basic architecture and test environment of FMS motion evaluation are designed. An appropriate and efficient method to classify 14 FMS actions is presented. A new FMS dataset with two orthogonal perspectives and wide-ranging cases is built. Benchmark for FMS imagery are given in terms of accuracy, continuity and speed.