1 Introduction

While artificial intelligence (AI) has triggered the development of the scientific and technological revolution and the industrial revolution, it has had a great impact on people's life, learning, and ways of thinking. It has promoted the structural reorganization of people's behavior patterns, personalized customization, and intelligent manufacturing, so as to promote medical and transportation. Breakthrough revolutions have been created in various fields such as industry and agriculture. As an important part of society, education profoundly affects people's present and future, and a new system needs to be built to adapt to social development (Foulquier et al. 2019). With the continuous reform of education, colleges hope to use advanced technology to promote the development of education, so the combination of AI and education has become the development direction of intelligent education (Parker and Forster 2018).

The reform of college sports training mainly includes teaching ideas, teaching theories, training methods, practice system and teaching materials. The new teaching mode needs to adapt to the development of new teaching ideas, and focus on individual development. In the traditional teaching methods, the training mode is relatively old, and it overemphasizes the students' mastery of sports skills. Meanwhile, the unified and more random training methods are adopted for all students, which greatly inhibits the development of students, and makes students have a psychological conflict with sports training (Beal et al.2019), thus affecting the overall training atmosphere. For colleges, lots of training contents are also relatively rigid and only aim to complete the unified teaching task of the school; moreover, the training content is relatively simple, that is, simple running, long jump and throwing and so on. In this way, there are still many problems in the traditional and unified teaching methods in the cultivation of high-quality and professional ability talents. In particular, training equipment can only meet the basic training, and more professional guidance and training need more professional coaches for manual guidance (Fister et al. 2018).

In the information technology center, information collection, transmission and processing technology has become the basic part of modern information technology, namely sensor technology, communication technology and computer technology. In many studies, it is also pointed out that wearable devices can obtain more basic information of the human body, and then provide more data for related devices for processing (Tian et al. 2019). In the past long time, there are more studies on sports, mainly focusing on obtaining students' physical health data, sports performance and analyzing and processing these data, so as to provide a basis for the design of relevant intelligent equipment. In China, college students' sports and training are mostly completed through classroom learning and amateur training. Moreover, in the training process, in addition to a part of theoretical knowledge, more is carried out by experience, and lack of support of big data and more intuitive decision-making basis (Novatchkov and Baca 2013). As the continuous development of the Internet and technology, more and more scholars in the field of sports begin to think whether AI can change people's long-time work and lifestyle in the future, and has a certain impact on sports training, especially whether the sports teaching in colleges can provide better training mode for teachers and students, so as to improve students' physical quality as well as the quality of sports training.

To sum up, "AI + education" mode has become the trend of future development, so in this context, improving the quality of college sports training has become an important work of sports teaching. The application of AI technology in the design of sports training and teaching system in colleges can meet the needs of different students and realize the development goal of personalized customized service, and further provide the basis for the development of AI in sports training and teaching quality.

2 Literature review

As the continuous development of AI, many scholars have applied it to the field of sports. (Galily 2018) pointed out that the application of AI to sports could obtain certain data; moreover, as for the production of sports content, it would change the behavior of relevant stakeholders, thus realizing the development of automation (Galily 2018). (Claudino et al. 2019) used PubMed, Scopus and WebofScience online databases to search studies on sports and AI technology and methods. The analysis proves that when AI is applied in sports training, artificial neural network, decision tree classifier, support vector machine and Markov process are mostly used, and they all have good performance indicators, which plays an important role in building specific AI technology and method (Claudino et al. 2019).

Rajšp and Fister 2020) conducted a systematic literature review on intelligent sports training, and put forward the verified analysis method. Then, they pointed out that intelligent sports training had been applied well in the training stage, which had certain guiding significance for group training and professional training in the future development (Rajšp and Fister 2020). (Patel et al. 2020) proposed that the sports industry was considered to be one of the most promising application fields of modern technologies such as big data, AI, data analysis, machine learning and neural network, and they had a certain impact on sports business and the experience of fans. Therefore, in the future research, the impact of these modern technologies on the whole sports industry will be also explored (Patel et al. 2020).

To sum up, in the previous research, scholars tended to analyze the application literature and methods of AI in sports, and rarely constructed sports training system. Therefore, AI technology is used in college sports training to provide support for improving the training quality and promoting the development of technology and education reform.

3 Methods and experiments

3.1 Human–computer interaction (HCI) and sports training

HCI refers to the technology of realizing the dialogue between humans and computers effectively through computer input and output devices. HCI technology includes that the machine provides lots of relevant information and prompts for instructions through the output or display device, and human input relevant information to the machine through the input device, answering questions and asking for instructions (Santos 2019).

In the process of interaction, it is necessary to recognize the user's gesture and shape, locate and analyze the voice through speech recognition, let the machine complete the feature extraction, and complete the auxiliary guidance by searching the knowledge base. When gesture information is acquired, camera, lidar and other equipment need to be used to collect video or image, and then hand features should be extracted to complete gesture recognition. This identification method does not need any external facilities, which can greatly reduce the number of external devices, thus simplifying the use process of users and making HCI more natural (Nadikattu 2020; Xue and Liu 2019).

In the process of body recognition, the sensor device needs to be used to locate the bone nodes of the human body, and the position of each bone node needs to be taken as a piece of data to form a frame, so as to realize the body recognition. The last is speech recognition. The device can also locate the position of the voice after acquiring the voice, improve the signal-to-noise ratio through the sound filtering algorithm of the device, remove redundant noise, accurately identify the position of the human body, and carry out word segmentation recognition of the user's output (Ahir et al. 2020).

Gesture recognition, body recognition, and speech recognition are combined with HCI, and the corresponding sample database can be established. It can form data background together with knowledge theory, and provide data support for students to obtain relevant professional knowledge and training guidance.

3.2 Expert knowledge base

In the construction of the expert knowledge base, the relevant knowledge of sports training needs to be stored into the system, including the knowledge base of sports strategy research and the knowledge base of sports organization management, which belong to the basic principles and theories. In addition, expertise based on direct and indirect experience needs to be collected (Wang et al. 2018, 2019). This is because the knowledge of experts is not always obtained from experience. If there is no solid theoretical basis, it is difficult to carry out the practice, and it is difficult to have corresponding experience content. Therefore, in the construction of the expert knowledge base, in addition to the basic theory, it also needs the support of practical data. A good expert system needs not only professional knowledge in a certain field, but also deep knowledge of the basic theory to deal with complex problems.

3.3 Demand analysis

The main function of the system is to improve the quality of college students' sports training. Therefore, according to the actual situation of college students' sports training, the sports training system is analyzed with the focus on sports training. The system is mainly implemented from two parts: functional requirements and technical requirements.

3.3.1 Functional requirements analysis

The first is functional requirements. When users use the system, the system should be able to analyze individual students, and assist teachers to understand and help students. Therefore, for students, the system should have the recognition function, which can identify individual activities and voice representation, and then give corresponding training suggestions; then, it also needs to have the coach side and the equipment side. Furthermore, it also needs an expert knowledge base system. It is provided by experts and related intelligent technology, and can give certain professional guidance to students in consulting (Umek and Kos 2018; Haoyang 2018). Finally, in the auxiliary training, different training tasks and corresponding training programs should be formulated according to different students. Also, in the actual training, the scheme needs to be analyzed, evaluated, and adjusted according to the situation of the students; the training information needs to be fed back; the effect needs to be evaluated, and the collection and management of training cases and training knowledge are completed.

3.3.2 Technical requirement analysis

The second is the technical requirements. In gesture and body recognition, small and light sensors need to be used to achieve information acquisition. Kinect platform is used to complete gesture and body recognition (Ling and Lihua 2015; Qing and Changhua 2017). Kinect, developed by Microsoft, is originally used as a game peripheral on the Xbox 360 game console, making the player operate the game only through action and voice interaction. Subsequently, it is favored by more researchers, and can be used in behavior recognition, face recognition and modeling because it can provide deep image information. Besides, it can also capture individual morphological actions and complete morphological recognition (Huan 2019; Penghai et al. 2019).

3.4 Module design

According to the above functional requirements, the design of the sports training system mainly includes the student module, coach module, information management module, knowledge base module, information processing module and teaching module (Wang et al. 2019).

Student module: students can log in to the system by inputting the account and password, and then select the required services, such as knowledge query and training guidance.

Coach module: the coach can log in the system through this module, view the training data of corresponding students, and give training suggestions according to the system prompts.

Information management module: students' historical training data, query data and corresponding system recommendation data will be stored in a database. Moreover, the coach's experience data and corresponding student data will also be retained.

Knowledge base module: this module is mainly the experience and suggestions provided by the experts in the sports field, and the related theoretical knowledge base is composed of industry experts and typical cases.

Information processing module: in this module, it is necessary to rely on the cloud platform to process the information data collected by the equipment, and further provide more suitable training programs for students and coaches.

Teaching module: theoretical teaching is inseparable in the training process. Therefore, basic theoretical knowledge and teaching progress management should be integrated under this module, so that coaches can have more time to formulate training programs and record each student's training information.

Fig. 1
figure 1

Sports training teaching system model with AI equipment

Full size image

This system is different from face recognition, traffic violations and other scenes of machine vision technology. Machine vision in moving scenes needs more complex algorithms with more accurate reasoning results. The general deep learning algorithm gives the confidence result, while the algorithm needs to give accurate and definite results in the moving scene. The motion vision technology of this system can calculate the continuous motion process, and analyze the data and performance of athletes. The products based on motion vision technology of this system overturn the traditional product forms such as wearable devices based on infrared and pressure sensor technology, so that athletes can abandon wearable devices and move, train and compete in the scene without equipment. The system mainly collects video and analyzes motion data based on camera and machine vision algorithms. In the standardized intelligent playground, the 50-m dash is taken as an example. The camera on the playground first identifies and authenticates the identity of the students, and the system issues voice commands such as "prepare", "gunshot (start)". After the students finish the 50-m dash, they can immediately hear the system broadcast their results. In the process of the movement, the system also analyzes whether the students rush and cross the road. In the standardized intelligent sports classroom, the camera can supervise and guide the whole process of students' training and testing. After each action is completed, students can see their own sports process data, sports performance and exercise prescription on the scene screen.

Therefore, Fig. 1 shows the sports training teaching system model based on AI technology.

Fig. 2
figure 2

Method for obtaining personalized training mode

Full size image

Moreover, individualized sports training can meet the individual training needs of different students. This is because each individual has different physical fitness, and the different exercise intensity in the past needs different training programs in sports training, which is more suitable for individual development. Figure 2 is the framework of individual personalized training.

Figure 2 suggests that sports training needs to be designed according to individual circumstances, including students’ emotions, goals, and values. AI equipment is used for recording and data analysis, and continuous training can be used to obtain the personalized characteristics of the students. Furthermore, a targeted and personalized training model is formulated to further improve the training efficiency of students and reduce unnecessary training.

3.5 System operating environment

Table 1 shows the hardware environment of the system at runtime.

Table 1 Hardware environment
Full size table

Besides, the software environment of the system is as follows: Windows 8/10 operating system, Kinect somatosensory device with Kinect SDK for Windows software; development language: C +  + ; Visual Studio software.

4 System implementation and test analysis

4.1 Analysis of test results

After the analysis of the sports training system, the detection content includes the student module, coach module, information management module, knowledge base module, the information processing module and teaching module. They can comprehensively reflect the operation efficiency and performance of the system. Table 2 presents the final system performance test results.

Table 2 System performance test results
Full size table

Table 2 show that each module of the system can complete the test content well, and has good system stability, indicating that the sports training teaching system designed has good operation efficiency and performance.

4.2 System implementation

The sports training and teaching system can analyze the students' psychology according to the collected information, and then formulate a more suitable training method, so as to improve the training efficiency. After students' personalized data are collected by interactive equipment, the training data and recommended scheme are fed back to the training end through data recording and analysis, which provides a reference for coaches to formulate personalized training programs. Besides, when the system obtains enough sample data through interactive technology, it can give the specific training content, so that students can complete the corresponding activities according to the given training content after class. Figure 3 presents its implementation.

Fig. 3
figure 3

Implementation Content of Physical Education Teaching System Based on AI

Full size image

4.3 HCI experimental results and analysis

In the process of HCI between students and the system, there may be some differences in the movement. Therefore, the stroke gesture in volleyball is used to test the students with different body shapes and weights to further test the accuracy of the system, and the data in Table 3 are obtained.

Table 3 Interactive test results
Full size table

Table 3 shows that the system is not affected by height and weight. However, considering the factors of different body shapes, their gesture recognition and accuracy are very similar. The teacher who teaches the course is the same coach, so there will be no difference in the content of coaching. It suggests that the recognition function of the system is less affected by different individual differences. From the recognition time, the system meets the requirements of real-time. For the system recognition function, the recognition time has a great relationship with the standard degree of students' actions. If the similarity is high, the recognition time is relatively long. When the similarity between the action and the given standard model is low, the recognition time is shorter.

To sum up, the HCI technology of AI technology is applied to the design of teaching system of sports training environment in colleges. On the one hand, students can learn by querying relevant basic knowledge and case content. On the other hand, through gesture, body recognition, speech recognition and other functions, the data collection work is completed; then, the corresponding training scheme is formulated according to the individual characteristics of the students and the training historical data, which is fed back to the coach so as to make a more detailed training plan according to the situation of the students. Moreover, it is convenient for the students to view the training tasks at any time and complete the training activities. The test results show that the system has good performance, high accuracy and short recognition time. It reveals that the application of AI technology in the design of college sports training and teaching system is of great significance to improve the training efficiency of students and the quality of sports teaching.

5 Conclusion

The HCI technology of AI technology is used to design the college sports training environment teaching system. Through the system, students can query the relevant basic theoretical knowledge and case content. Moreover, when interactive technology is used to identify students' gesture, body, voice and relevant information is collected; the system will process and analyze the information content, feedback the personalized characteristic data and training data to the coach, provide personalized training scheme for the coach, provide training tasks for students, and further provide reference for coaches to develop personalized training programs and tasks according to different students. However, there is little research on identification technology, and there is no comparison of identification technology. Therefore, future research will focus on the analysis of identification technology and carrying platform. In the future development, the integration of AI and education is inevitable, and the security of relevant information obtained by AI technology will become the focus of future research. Besides, the training system provides some ideas for the construction and improvement of the training system of athletes in the future major competitions.