Welcome to Francis Academic Press

Academic Journal of Computing & Information Science, 2023, 6(13); doi: 10.25236/AJCIS.2023.061311.

Automatic Description Method of Floor Exercise Video Based on Three-dimensional Convolutional Network and Multi-label Classification

Author(s)

Feng He1, Wenda Gao2, Yue You3

Corresponding Author:
Feng He
Affiliation(s)

1College of Continuing Education, Civil Aviation Flight University of China, Guanghan, China

2Department of Atmospheric Sciences, Civil Aviation Flight University of China, Guanghan, China

3Department of Atmospheric Sciences, Civil Aviation Flight University of China, Guanghan, China

Abstract

As people's attention to health and sports increases, the amount of data and audience of sports video is also growing rapidly. Based on this, the automatic description method of floor exercise video has attracted the attention of scientific researchers and industry. The research focus of this paper is on the automatic description of floors video, that is, to generate professional nouns by observing the movements of athletes in the video. This research has a wide range of application value, involving sports analysis, automatic interpretation and sports guidance. This paper specifically studies the automatic understanding of human movements in floor exercise videos, and combines the knowledge of computer vision and deep learning to realize the intelligent labeling and representation of specific human movements in video sequences.An automatic description method of floor exercise video based on three-dimensional convolutional network and multi-label classification is proposed. The floor exercise action is composed of multiple decomposition actions. In the work of this paper, a classifier of single decomposition action is constructed, and the automatic description problem of floor exercise action is transformed into a multi-label classification problem. Since the two-dimensional convolutional neural network loses time information when extracting features, this paper uses a three-dimensional convolutional network to extract spatio-temporal features of the video. Through multiple binary classifications, the goal of multi-label classification is achieved. In order to verify the effectiveness of the method, the classification results are randomly combined into a sentence, which is compared with the results of the automatic description method.

Keywords

Floor exercise; Three-dimensional convolutional network; Multi-label classification; Automatic description

Cite This Paper

Feng He, Wenda Gao, Yue You. Automatic Description Method of Floor Exercise Video Based on Three-dimensional Convolutional Network and Multi-label Classification. Academic Journal of Computing & Information Science (2023), Vol. 6, Issue 13: 67-75. https://doi.org/10.25236/AJCIS.2023.061311.

References

[1] Qin Dan.Research on video content automatic classification algorithm based on multi-feature combination and SVM [D].Shanghai Jiaotong University, 2009.

[2] Li Haixia. Design and implementation of video automatic description system based on deep learning [D].University of Electronic Science and Technology of China, 2018.

[3] Zhang Lu.Comparative study on the current situation of amateur gymnastics training between China and the United States [D].Chengdu Institute of Physical Education, 2014. 

[4] Xin Lijuan. Architecture and Application of Smart Sports [J]. Electronic Technology and Software Engineering, 2018 (15): 128-128.

[5] Kojima A, Tamura T, Fukunaga K. Natural language description of human activities from video images based on concept hierarchy of actions[J]. International Journal of Computer Vision, 2002, 50(2): 171-184.

[6] Wang Jianhong.Research on key technologies of video-based human action recognition [D]. Southeast University, 2017.

[7] Liu Duanyang, Qiu Weijie. Multi-label classification based on weighted SVM active learning [J]. Computer Engineering, 2011, 37 (8): 3.DOI: 10.3969 / j.issn.1000-3428.2011.08.062.

[8] Zhao Pengpeng, Jiao Yang, Xian Xuefeng, etc. A multi-label active learning classification method and system based on SVM: CN201410184086.8 [P].CN201410184086.8 [2023-11-30].

[9] Qian Wenzhuo. Human action recognition technology based on MA-C3D neural network [J].Modern computer, 2021, 27 (35): 6. 

[10] Jiao Chunpeng. Comparative study of multi-classification methods based on binary SVM [D].Xi 'an University of Electronic Science and Technology [2023-11-30].DOI: CNKI: CDMD: 2.1013.111129. 

[11] Callison-Burch C , Osborne M , Koehn P .Re-evaluation the Role of Bleu in Machine Translation Research[C]//EACL 2006, 11st Conference of the European Chapter of the Association for Computational Linguistics, Proceedings of the Conference, April 3-7, 2006, Trento, Italy.2006. DOI:http://dx.doi.org/.

[12] Zhang Yu. Research on the optimization design of individual complete sets of movements of Chinese women's competitive gymnastics under the international gymnastics rules in 2013-2016 [D].Hunan Normal University, 2014. 

[13] Tong Zhanbei, Zhong Jianwei, Li Zhenwei, et al. Power quality disturbance classification based on VGG16 image feature extraction and SVM [J].Electricity, 2023 (7): 7-13.

[14] Mogan J N, Lee C P, Anbananthen K S M ,et al. Gait-DenseNet: A Hybrid Convolutional Neural Network for Gait Recognition[J].IAENG Internaitonal journal of computer science, 2022(2 Pt.2): 49.