Open Access

Research on Sports Dance Training and Teaching in Modern Colleges and Universities Combined with Deep Learning

Jie Jiao

Jie Jiao

Luoyang Institute of Science and TechnologyLuoyang, China
Search for this author on
Sciendo|Google Scholar
Jiao, Jie
  
Mar 21, 2025
Research on Sports Dance Training and Teaching in Modern Colleges and Universities Combined with Deep Learning's Cover Image
About this article
Previous Article
Next Article
Cite
Download Cover
Published Online: Mar 21, 2025
Received: Oct 13, 2024
Accepted: Feb 10, 2025
DOI: https://doi.org/10.2478/amns-2025-0602
Keywords
© 2025 Jie Jiao, published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1.

Two-branch multi-stage convolutional neural network
Two-branch multi-stage convolutional neural network

Figure 2.

Schematic diagram of affinity vector field of key points of a single limb
Schematic diagram of affinity vector field of key points of a single limb

Figure 3.

Network structure of the TAR-DL
Network structure of the TAR-DL

Figure 4.

Mapping between sequential points before and after dynamic time planning
Mapping between sequential points before and after dynamic time planning

Figure 5.

Confusion matrix of ballroom dancing basic movement recognition
Confusion matrix of ballroom dancing basic movement recognition

Figure 6.

Comparison of foot angle change of athletes with standard sequence
Comparison of foot angle change of athletes with standard sequence

Comparison of scoring results

Dancer Upper-body fluidity Lower-body fluidity Musical timing Body balance Choreography Stotal
A1 4 4 5 5 4 90.0
A2 4 5 5 5 5 90.3
A3 4 3 5 3 4 70.9
A4 4 5 3 2 5 69.8
A5 4 3 4 4 4 71.5
A6 4 2 4 2 3 61.6
A7 2 3 3 2 3 55.7

Experimental comparison

Algorithms Accuracy rate/% Time consuming/ms
DTW 91.25 2.04
FastDTW 86.73 1.74
HMM 90.68 2.18
Improved-DTW 94.39 1.67

Evaluation results with and without alignment processing

Dancer The score of sports dance teachers after the normalization process Align the processed Stotal Do not use aligned Stotal
A1 93 90.0 67.4
A2 93 90.3 38.5
A3 82 70.9 60.2
A4 78 69.8 44.8
A5 82 71.5 52.9
A6 58 61.6 30.6
A7 50 55.7 41.5

Ballroom dancing movement similarity score using the method in this paper

Dancer Spos Svel Sangle Sdist Stotal
A1 91.9 93.0 90.6 84.6 90.0
A2 89.8 90.8 91.0 89.7 90.3
A3 70.1 72.2 69.2 72.1 70.9
A4 79.5 80.9 63.7 55.2 69.8
A5 77.9 81.4 65.9 60.8 71.5
A6 65.8 58.9 61.1 60.5 61.6
A7 56.0 60.3 49.7 56.8 55.7

Comparison of 3D-based recognition methods in NTU RGB-D

Methods CV/% CS/%
ST-LSTM 78.94 70.31
TSRJI 81.41 74.25
Clips+CNN+MTLN 85.67 80.46
ST-GCN 89.44 82.17
DPRL 90.75 84.38
SGN 94.38 87.52
2S-AGCN 96.24 89.43
2S-NLGCN 96.24 89.43
MS-AAGCN 97.16 90.27
Sym-GNN 97.58 90.49
MS-AAGCN+TEM 97.69 91.84
TAR-DL 99.72 97.41
Language:
English
Publication timeframe:
1 times per year
Journal Subjects:
Life Sciences, Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics, Physics, other
Journal RSS Feed