Preventive Medicine 173 (2023) 107589

Contents lists available at ScienceDirect

Preventive Medicine
journal homepage: www.elsevier.com/locate/ypmed

Data analysis of physical recovery and injury prevention in sports teaching

A R T I C L E I N F O A B S T R A C T

Keywords: Athletes can also cause damage to some parts of their body during training, so specialized preparation activities
Wearable devices should be carried out before athlete training to reduce the damage caused to the athlete’s body, allowing the
Physical education teaching stressed parts to move and distribute the load. Excessive recovery has a significant effect on improving the
Physical recovery
performance level of the athletes studied and preventing sports injuries. This article studies the data analysis of
Sports injury prevention
body recovery and injury prevention in physical education teaching based on wearable devices. Real time
collection of students’ exercise data, including indicators such as exercise volume, heart rate, steps, distance, etc.,
by wearing wearable devices. By using Internet of Things technology to transmit data to cloud servers, data
analysis and mining techniques are used to process the data and study issues related to body recovery and injury
prevention. Specifically, this article adopts methods such as time series analysis, machine learning algorithms,
and artificial neural networks to analyze the relationship between exercise data and body recovery and injury
prevention, providing scientific guidance and support for physical education teaching. This method can monitor
students’ exercise data in real-time, predict the risk of physical recovery and injury, and provide corresponding
prevention and guidance suggestions.

1. Introduction injury. However, athletes cannot avoid injuries during regular training
and competitions, which brings great inconvenience to their learning,
Since the 1960s to the present, many research fields in sports training health, and daily life. It can also affect their regular training, thereby
have proposed many research perspectives, among which the most affecting their performance. Therefore, we need to prevent injuries
important is to study how athletes’ physical abilities change under among athletes.
strong sports stimuli. Many relevant scholars have proposed various
hypotheses about this issue, and the most important of these scholars’ 2. Related work
hypotheses is the theory of excessive recovery (Maciejczyk et al., 2014).
The theory proposed by scholars fully explains that sports training can The AlexNet model proposed in the literature has made a historic
increase athletes’ competitive abilities, as well as improve the muscles breakthrough in the ImageNet competition and has become one of the
and various systems of the human body. However, as various coaches representative applications of deep learning technology (Alom et al.,
and researchers engaged in theoretical and practical discussions, they 2018; Gu et al., 2017). The AlexNet model abstracts image information
began to have some confusion and confusion about this theory of layer by layer by introducing multi-layer convolution and pooling op­
excessive recovery (Van Tuyckom and Scheerder, 2010). At present, erations, thereby achieving high recognition accuracy. The literature
most researchers believe that this theory lacks practical evidence and is has made improvements to the AlexNet model, such as adjusting the size
difficult to operate in real life. At this time, the theory of excessive re­ and step size of convolutional kernels, and achieved good performance.
covery became a hot topic in the field of sports at that time. However, in The biggest contribution of this model is to lead the neural network
recent years, competitive competitions have become increasingly structure deeper. By increasing the network depth and introducing the
intense day by day, and athletes often use more difficult movements in Inception structure, higher classification accuracy and fewer parameter
competitions and regular exercises, which can easily lead to physical quantities have been achieved. The literature further improves the

* Corresponding author.
E-mail address: wuxiangzhi_7912@126.com (X. Wu).

https://doi.org/10.1016/j.ypmed.2023.107589
Received 19 May 2023; Received in revised form 20 June 2023; Accepted 23 June 2023
Available online 24 June 2023
0091-7435/© 2023 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J. Cui et al. Preventive Medicine 173 (2023) 107589

performance and efficiency of image recognition and classification can improve the performance of classifiers. For example, the use of data
(Yang et al., 2000). These research results are not only widely used in the augmentation techniques in literature can increase the diversity of
field of image recognition, but also provide important reference for in- training data, thereby improving the generalization ability of classifiers;
depth learning research in speech recognition, natural language pro­ The literature uses transfer learning to transfer the learned knowledge to
cessing and other fields (Pan et al., 2020). The ResNet network was new tasks, thus improving the learning efficiency and accuracy of the
proposed in the literature to solve the degradation problem by intro­ classifier.
ducing residual units, and the network layers were deepened to over a
hundred layers. This method makes the training of deep neural networks 3. Construction of a sports teaching action monitoring system
easier and more effective (Oono and Suzuki, 2019). based on wearable devices
The literature considers the category with the highest number of
categories among the K samples closest to a sample to be the category of In the field of sports injury protection in physical education teaching,
that sample (Kuang and Zhao, 2009). This algorithm is simple and the system stores a large number of unstructured data, such as qualita­
intuitive, easy to understand and implement, but it requires a large tive data of sports medicine, training data of college students, and so on
amount of computation when processing large-scale data and needs to (Verhagen et al., 2010). These data need to be processed through data
be optimized during algorithm implementation (Moldagulova and mining algorithms to accurately assess the health status of college stu­
Sulaiman, 2017). The KNN algorithm has been widely applied in fields dents and intelligently recommend personalized and scientific training
such as image recognition, recommendation systems, and data mining methods. In order to process these unstructured data, we need to design
(Deng et al., 2016). In the literature, AdaBoost RandomForest algorithm specific data mining algorithms and models (Xue and Liu, 2018). These
of random forest and SVM KNN algorithm based on SVM are used to algorithms and models need to consider the characteristics of the data,
construct classifiers to recognize human posture (Bhushan et al., 2022). such as data size, data type, data quality, etc., to ensure that the analysis
This combination method can fully utilize the advantages of various results are consistent with the actual situation. To achieve this, we need
algorithms and improve the performance and accuracy of the classifier. to perform data preprocessing, including steps such as data cleaning,
In addition to combining algorithms, there are also other methods that data conversion, and data integration, to ensure the quality and

Fig. 1. Physical education teaching action monitoring system.

2
J. Cui et al. Preventive Medicine 173 (2023) 107589

accuracy of the data. teaching and exercise guidance for students based on data analysis and
As shown in Fig. 1, the wearable device based sports teaching action mining results, helping them better improve their sports skills and
monitoring system is composed of seven levels, which are interrelated health levels.
and form a complete sports teaching action monitoring system. Through
data collection, processing, analysis, and display, it helps teachers and 4. Experimental design and simulation result analysis
students better understand and master their sports status and progress,
improve sports skills and abilities, and achieve better teaching effects. 4.1. Experimental subjects
In the intelligent sports and health management system based on
wearable devices, the functional layer is the core part of the system, This experiment selected athletes from a certain club as the research
responsible for important tasks such as data processing, analysis, and subjects, and randomly selected 20 athletes with similar levels of exer­
recommendation (Nahavandi et al., 2022). This layer obtains training cise and basic physical conditions as the experimental subjects. Divide
data that needs to be calculated from the data layer, such as exercise the selected 20 athletes into 2 groups based on their level and training
trajectory, heart rate, steps, calories, etc., and calculates the data using years. One group is the experimental group, and the 10 athletes in the
appropriate operation rules set in the rule layer, thereby obtaining the experimental group were trained through excessive recovery during this
analysis results of exercise data and health assessment information. In experiment. The 10 athletes in the control group used traditional
the functional layer, the system also supports sports injury and disease training methods during this experiment. The specific basic information
prediction functions. By analyzing sports data, the risk of sports injury of the athletes in the experimental group and control group, including
and disease for college students can be predicted, reminding them to pay age, height, weight, and years of training, is shown in Table 1. The basic
attention to sports safety and health issues. In addition, the functional physical condition and sports skills of the two groups of athletes. There
layer also provides scientific training method recommendations, rec­ was no significant difference before the experiment.
ommending suitable training methods and plans for college students In order to better explore whether excessive recovery methods can
based on their physical condition, health goals, exercise habits, and effectively improve the performance level of athletes and whether they
other factors. can have sufficient preventive effects on sports injuries. Before con­
The data layer is a key component of an intelligent sports and health ducting the experiment, test the basic performance of the experimental
management system, used to store sports information of college students and control group athletes, and compare them with the basic perfor­
(Huifeng et al., 2020). The data layer includes two sub databases, mance after the experiment to explore which of the two training
namely the personal basic sports database and the personal sports status methods is more effective in improving the performance level of ath­
database, which are used to store the basic sports data and sports status letes. Record the athletes’ basic FMS test scores before and after the
data of college students, respectively. experiment, compare and analyze the basic FMS test scores of the two
Fig. 2 shows the workflow of a wearable device based sports teaching groups of athletes before and after the experiment, in order to explore
data collection system. Firstly, students wear wearable devices, and the
devices collect students’ movement data through sensors, including
Table 1
movement posture, exercise volume, heart rate, and other data. Sec­
Basic information of two groups of athletes.
ondly, the collected data is transmitted through wireless communication
technology to the central control system for storage and analysis. The Test indicators Test group(N = 10) Control group(N = 10) P value

central control system can monitor students’ exercise and health status Age 18.86 ± 1.5574 18.95 ± 1.6378 0.8106
in real-time, and provide corresponding feedback and guidance through Height (cm) 174.21 ± 6.1699 172.27 ± 5.3779 0.8105
Weight(kg) 65.70 ± 7.2514 67.76 ± 5.4502 0.7361
data analysis and mining. Finally, teachers can view students’ exercise
Training years (years) 4.87 ± 0.5074 5.06 ± 1.2559 0.8835
data through the central control system, and provide personalized

Fig. 2. Data collection process.

3
J. Cui et al. Preventive Medicine 173 (2023) 107589

which training method can have a better preventive effect on athletes’ Table 3
sports injuries. The effect of excessive recovery on the improvement of athletes’ sensitivity
level.
Test indicators Test group(N = 10) Control group(N = 10) P value
4.2. The effect of excessive recovery on improving athletes’ performance
level Hexa jump(s) 11.61 ± 0.5781 12.27 ± 0.5382 0.0665
“T” test(s) 5.55 ± 0.3329 6.13 ± 0.3177 0.0730
Nebraska test(s) 11.39 ± 0.3718 12.24 ± 0.3517 0.0822
The endurance scores of the two groups of athletes after the exper­ Pro test(s) 5.40 ± 0.1994 5.73 ± 0.3342 0.0693
iment are shown in Table 2. After the experiment, the lmin whip leg Slide back and forth (s) 6.08 ± 0.5631 6.84 ± 0.7545 0.0628
performance of the experimental group was 106.30 ± 4.7514 times, the Slide left and right (s) 6.18 ± 0.4668 6.67 ± 0.9270 0.0806
1 min punch leg performance was 111.27 ± 3.7679 times, the 30s
sliding step performance was 19.34 ± 3.7677 times, the 20s straight leg
performance was 90.79 ± 7.1032 times, and the 20s whip leg perfor­ Table 4
mance was 48.03 ± 1.8570 times. After professional training, the con­ Basic FMS test results of athletes in the pre experiment test.
trol group’s lmin whip leg performance was 101.65 ± 4.9737 times, The Test indicators Test group(N = 10) Control group(N = 10) P value
1-min punch score was 110.98 ± 4.0694 times, the 30-s turn to slide
Squat 1.68 ± 0.8676 1.76 ± 0.6043 0.0838
score was 17.30 ± 3.3827 times, the 20-s straight punch score was 85.50
Hurdle step 1.79 ± 0.7226 1.93 ± 0.4924 0.0787
± 10.6027 times, and the 20-s whip leg score was 43.26 ± 1.1263 times. Straight lunge 1.55 ± 0.5236 1.67 ± 0.6254 0.0760
After the experiment, the endurance scores of 20 athletes were measured Shoulder mobility 2.13 ± 0.5577 2.11 ± 0.6441 0.0756
again, and the measured endurance scores of the 20 athletes showed Total score 13.95 ± 1.7666 13.83 ± 2.1017 0.0707
significant changes after the experiment. Especially in the experimental
group that used the excessive recovery training method, the endurance
level was significantly improved. Table 5
The sensitivity scores of the two groups of athletes after the experi­ Effect of Excessive Recovery on Improving Athletes’ FMS Test Performance
ment are shown in Table 3. The hexagonal jump scores of the experi­ Level.
mental group’s 10 athletes are 11.61 ± 0.5781 s, the “T” test scores are Test indicators Test group(N = 10) Control group(N = 10) P value
5.55 ± 0.3329 s, the Nebraska test scores are 11.39 ± 0.3718 s, the Pro
Squat 2.22 ± 0.4455 1.61 ± 0.7747 0.0630
test scores are 5.40 ± 0.1994s, the front and back slide scores are 6.08 ± Hurdle step 2.18 ± 0.3748 1.93 ± 0.4924 0.0850
0.5631 s, and the left and right slide scores are 6.18 ± 0.4668 s; The Straight lunge 2.16 ± 0.3793 1.67 ± 0.6254 0.0739
hexagonal jump performance of the control group of 10 athletes was Shoulder mobility 2.36 ± 0.5084 1.95 ± 0.7636 0.0631
12.27 ± 0.5382 s, the T-shaped test performance was 6.13 ± 0.3177 s, Total score 16.41 ± 1.1184 13.06 ± 2.6448 0.0612

the Nebraska test performance was 12.24 ± 0.3517 s, the Pro test per­
formance was 5.73 ± 0.3342 s, the front and back sliding step perfor­ When athletes are training, the training content and the physical load
mance was 6.84 ± 0.754 s, and the left and right sliding step they bear should follow scientific principles. Instructors should develop
performance was 6.67 ± 0.9270 s. After the experiment, the sensitivity corresponding exercise plans based on the specific situation of each
scores of 20 athletes were measured again, and the sensitivity scores of athlete, gradually increasing the load they bear, and making specific
the measured 20 athletes showed significant changes after the experi­ arrangements for the amount of exercise they exercise. Secondly, ath­
ment. Especially in the experimental group using the excessive recovery letes may also cause damage to some parts of their bodies during
training method, the sensitivity level was qualitatively improved. training. Therefore, specialized preparation activities should be carried
out before the athlete’s training to reduce the damage caused to the
4.3. Prevention effect of excessive recovery on athletes’ sports injuries athlete’s body, allowing the stressed parts to move and distribute the
load. Some athletes also need to practice extensively when performing
Measure the basic FMS scores of two groups of athletes before the physically demanding movements to prevent physical exhaustion and
experiment, as shown in Table 4. Before the experiment, the basic FMS injury during exercise. And athletes themselves should also pay atten­
scores of 20 athletes were measured, and there was no significant dif­ tion to their physical fitness and health. During training, they should
ference in the measured basic FMS scores of the 20 athletes. follow the coach’s instructions and take every training session seriously.
After the experiment, the basic FMS scores of the two groups of In order to prevent physical injury to athletes, improving their various
athletes were measured again, as shown in Table 5. After the experi­ skills is a very good way. It is not advisable to pursue the skills of one
ment, the basic FMS scores of the 10 athletes in the experimental group part while neglecting the development of another part of their physical
showed significant differences compared to before the experiment, fitness. If it continues for a long time, it will cause excessive local
resulting in a significant improvement. burden, leading to unnecessary sports injuries. At the same time, ath­
letes should strengthen the exercise of weak and easily injured parts of
Table 2 the body, increase the firmness of body joints, and thus improve their
The effect of excessive recovery on improving athletes’ endurance performance physical functions. In competitions and practice, many athletes do not
level. pay attention to the practice of basic skills, which leads to poor mastery
of many technical movements and many errors, resulting in incorrect
Test indicators Test group(N = Control group(N P value
10) = 10) and non-standard technical movements. Some athletes like to try some
very difficult movements, but their own physical fitness level limitations
Lmin whip leg (times) 106.30 ± 101.65 ± 4.9737 0.0775
4.7514
and many unreasonable spontaneous movements can easily cause
1 min punching (times) 111.27 ± 110.98 ± 4.0694 0.0722 physical damage. So athletes should have an awareness of the impor­
3.7679 tance of basic skills themselves, and at the same time, they should follow
30s change direction sliding 19.34 ± 3.7677 17.30 ± 3.3827 0.0818 the arrangements and training of coaches, diligently practice the
step (times)
movement norms of basic skills.
About 20s straight punch 90.79 ± 7.1032 85.50 ± 10.6027 0.0854
(times)
About 20s to whip the legs 48.03 ± 1.8570 43.26 ± 1.1263 0.0708
(times)

4
J. Cui et al. Preventive Medicine 173 (2023) 107589

5. Conclusion References

The changes and upgrades in competitive rules have set many trial Alom, M.Z., Taha, T.M., Yakopcic, C., Westberg, S., Sidike, P., Nasrin, M.S., Asari, V.K.,
2018. The History Began from Alexnet: A Comprehensive Survey on Deep Learning
standards for athletes for difficult movements, which will enable them to Approaches. arXiv preprint. arXiv:1803.01164.
break their limits time and time again, thereby demonstrating their Bhushan, S., Alshehri, M., Keshta, I., Chakraverti, A.K., Rajpurohit, J., Abugabah, A.,
superior technical level. The setting of these rules enables athletes to 2022. An experimental analysis of various machine learning algorithms for hand
gesture recognition. Electronics 11 (6), 968.
better demonstrate their strength. More effective training methods need Deng, Z., Zhu, X., Cheng, D., Zong, M., Zhang, S., 2016. Efficient kNN classification
to be explored, and the over recovery training method has been vali­ algorithm for big data. Neurocomputing 195, 143–148.
dated in this experiment to be more effective. Gu, S., Ding, L., Yang, Y., Chen, X., 2017, November. A new deep learning method based
on AlexNet model and SSD model for tennis ball recognition. In: In 2017 IEEE 10th
International Workshop on Computational Intelligence and Applications (IWCIA).
Ethical approval IEEE, pp. 159–164.
Huifeng, W., Kadry, S.N., Raj, E.D., 2020. Continuous health monitoring of sportsperson
using IoT devices based wearable technology. Comput Commun 160, 588–595.
This article does not contain any studies with human participants
Kuang, Q., Zhao, L., 2009. A practical GPU based kNN algorithm. In: Proceedings. The
performed by any of the authors. 2009 International Symposium on Computer Science and Computational Technology
(ISCSCI 2009). Academy Publisher, p. 151.
CRediT authorship contribution statement Maciejczyk, M., Wiecek, M., Szymura, J., Cempla, J., Wiecha, S., Szygula, Z., Brown, L.E.,
2014. Effect of body composition on respiratory compensation point during an
incremental test. J Strength Condition Res 28 (7), 2071–2077.
Jiayi Cui: Conceptualization, Data curation, Formal analysis, Moldagulova, A., Sulaiman, R.B., 2017, May. Using KNN algorithm for classification of
Writing – original draft. Hongyi Du: Investigation, Methodology, Vali­ textual documents. In: In 2017 8th International Conference on Information
Technology (ICIT). IEEE, pp. 665–671.
dation. Xiangzhi Wu: Supervision, Validation, Writing – review & Nahavandi, D., Alizadehsani, R., Khosravi, A., Acharya, U.R., 2022. Application of
editing. artificial intelligence in wearable devices: opportunities and challenges. Comput
Methods Prog Biomed 213, 106541.
Oono, K., Suzuki, T., 2019, May. Approximation and non-parametric estimation of
Declaration of Competing Interest ResNet-type convolutional neural networks. In: International Conference on
Machine Learning. PMLR, pp. 4922–4931.
The authors declare that they have no conflict of interests. Pan, H., Pang, Z., Wang, Y., Wang, Y., Chen, L., 2020. A new image recognition and
classification method combining transfer learning algorithm and mobilenet model
for welding defects. Ieee Access 8, 119951–119960.
Data availability Van Tuyckom, C., Scheerder, J., 2010. Sport for all? Insight into stratification and
compensation mechanisms of sporting activity in the 27 European Union member
states. Sport Educ Soc 15 (4), 495–512.
Data will be made available on request.
Verhagen, E.A., van Stralen, M.M., Van Mechelen, W., 2010. Behaviour, the key factor
for sports injury prevention. Sports Med 40, 899–906.
Acknowledgments Xue, B., Liu, T., 2018. Physical health data mining of college students based on DRF
algorithm. Wirel Pers Commun 102, 3791–3801.
Yang, C.C., Prasher, S.O., Landry, J.A., DiTommaso, A., 2000. Application of artificial
This paper was supported by China National Social Science Fund neural networks in image recognition and classification of crop and weeds. Can Agric
Project: Typical Cases, Model Refinement and Promotion Path Research Eng 42 (3), 147–152.
of the Integrated Development of Sports and Medicine in my country
(21BTY090).