Introduction

With the rapid development of multimedia and network technologies, open educational resources (OER) have flourished during the past decades (Butcher 2015; Ehlers 2011). Beyond time-space constraints, online learners are provided opportunities to build knowledge and skills by various teaching aids, including lecture videos, PowerPoint slides, discussion forums, quizzes, exams, and other additional resources (Dixson 2010). There are two typical teaching models for online courses: recorded broadcast and live broadcast. Figure 1 describes the main platforms for online teaching in America and China, which benefit millions of online learners each semester. Online courses make modern education beyond time–space constraints and borderless. Particularly, the massive open online course (MOOC) is the most popular type of OER, in which hundreds of thousands of students can participate in an online course. Similar to movies, most lecture videos for MOOCs are made in professional studios by photographic specialists, preparing “lines” in the early stage and editing the video in the later stage. Massive MOOCs are provided by the online teaching platforms depicted in Fig. 1, including Coursera, edX, Chinese university MOOC, and Zhihuishu. Some other platforms for video communications, such as ZOOM, Tencent QQ Qunketang, Tencent Meeting, are appropriate for live lectures and exam monitoring. The real-time interactivity of live lectures makes the teaching effectiveness of online courses close to that of face-to-face teaching in classrooms. Meanwhile, ZOOM and Tencent QQ Qunketang provide the function of recording live lectures for playback and review.

Fig. 1
figure 1

Main online teaching platforms used in America and China.

Full size image

However, current technologies of constructing online teaching platforms cannot be widely applied because of their drawbacks, such as high economic cost, poor user experience, and low learning effectiveness. To alleviate these problems, a framework of 5G-based Meta Classrooms is proposed in this study. The main contributions of this study are threefold:

  1. (1)

    Analyzing the development status of the typical online teaching platforms and the shortcomings of the existing online teaching technologies.

  2. (2)

    Proposing a framework of Meta Classrooms to reduce the costs of constructing online teaching platforms and improve the experience and learning efficiency of online learners via a commonly used device, 5 G wireless communication network, and virtual reality (VR) technology.

  3. (3)

    Giving experimental results obtained from achieving the short-term objectives for constructing Meta Classrooms and key technologies of the medium-term and long-term objectives to verify the feasibility of Meta Classrooms.

The rest of this paper is organized as follows. Chapter 2 summarizes the development status of the main platforms for online teaching in America and China, and analyzes the problems in current MOOCs. Chapter 3 presents the framework and objectives of 5G-based Meta Classrooms and key technologies involved. The concluding remark is given in Chapter 4.

Development status of MOOCs

In the past few years, MOOCs have been provided in many countries based on the rapid development of information technologies. According to the latest data published on UNESCO (UNESCO 2024), WORLDOMETER (WORLDOMETER 2024) and World Bank Group (World Bank Group 2024), the expenditures on education of the U.S. government and the Chinese government in 2022 exceeded $1300 billion and $700 billion, respectively, which are 10 times and 5 times that of the third place in the top ten countries worldwide by GDP (Gross Domestic Product). Combined, they account for 66% of the total investments from the top ten GDP countries. The huge investment of the two governments has promoted the rapid improvement of their educational environment and technology. The combined number of students from America and China accounts for over 46% of the total number of students in the top ten countries worldwide by enrollment in tertiary education. These students share the educational resources brought by substantial government investment, particularly the online teaching resources supported by high-tech technologies. This study focuses on analyzing the development and problems of MOOCs in America and China, which are the most representative countries, and constructs 5G-based Meta Classrooms to solve these problems and improve the effectiveness of online and offline teaching.

The top-ranked universities in America, such as Stanford University, Massachusetts Institute of Technology (MIT), and Harvard University, founded their MOOC platforms in 2012, which is called “The Year of the MOOC” (Pappano 2012). Following by Chinese university MOOC and Zhihuishu which are typical and early platforms for online learning in China, a lot of open online courses are provided for Chinese universities, from “ Double First-Class ” universities which are focused on the construction of world-class universities and first-class disciplines, such as Peking University and Tsinghua University, to local colleges, such as Nanchang Institute of Science and Technology and Heilongjiang Institute of Technology. These platforms not only benefit students in universities, who can acquire knowledge in a more flexible way, but also provide opportunities for white workers to make rapid progress. Especially, during the COVID-19 (Coronavirus Disease 2019) prevention and control phase in 2020, Chinese universities started the new semester online in March. Similarly, some universities in other countries have gradually transferred face-to-face teaching to online teaching.

Main platforms for online teaching in America

In 2012, the Massachusetts Institute of Technology (MIT), and Harvard University conducted edX (Harvard and MIT 2023), which is a free online course platform and has provided more than 4000 open online courses from 250 institutions at present, covering 31 subjects including computer programming, art, architecture, etc. The most popular Python course is “Introduction to Computer Science and Programming Using Python,” provided by MIT and enrolled by 1,605,508 learners all over the world by August 2023.

At the same time, Coursera (Stanford 2023), a profitable online course platform, was launched by two professors at Stanford University, Andrew Ng and Daphne Koller. According to the official statistics published in August 2023 (Stanford 2023), Coursera has opened more than 5400 courses with 318 partners from 54 countries. Up to now, more than 124 million users have registered on Coursera.

Main platforms for online teaching in China

Chinese university MOOC (NETEASE and Higher Education Press 2023) was launched by NETEASE and Higher Education Press in 2014. At present, 804 institutions of higher learning, including universities supported by Project 985, provide more than 16,800 courses on the Chinese University MOOC. Similar to edX, the most popular course on Chinese University MOOC is “Python Language Programming” offered by Beijing Institute of Technology, with more than 4.9 million online learners in total since 2015. Figure 2 shows the number of participants in each semester. The horizontal and vertical coordinates of Fig. 2 represent the semester and the number of participants in the semester, respectively, as do those of Fig. 3.

Fig. 2
figure 2

Number of participants in each semester of “Python Programming”.

Full size image
Fig. 3
figure 3

Number of participants in each semester of “Situation and Policy”.

Full size image

As shown in Fig. 2, the number of participants in “Python Language Programming” demonstrates an overall upward trend except for two short summer semesters (9 and 12). Especially, the number of participants in the 11th semester, which is the period from February 18, 2020, to May 12, 2020, called the COVID-19 prevention and control phase, reaches 743,634 and is 84.3% higher than that of the previous semester.

Zhihuishu (Shanghai Zhuoyueruixin 2023), another large-scale online course platform in China, has also opened close to 15,000 courses to more than 3000 universities and schools, and allows mutual recognition of credits for member universities. More than 160 million students have taken cross-school courses and earned credits. According to the official website, more than 10000 courses will be provided in the autumn-winter semester of 2023. The most popular course on Zhihuishu is “ Situation and Policy “ offered by Renmin University of China and Peking University. Figure 3 describes the number of participants in each semester of “ Situation and Policy “, showing an upward trend except for a short summer semester. Especially, the number of participants in the 10th semester, which is the spring-summer semester of 2020, reaches 1,348,100 and is 42.42% higher than that of the previous semester. Similar to “Python Language Programming” mentioned above, the number of participants in “ Situation and Policy “ has significantly increased during the COVID-19 prevention and control phase. This shows that online courses have played an important role in extraordinary periods.

As shown in Figs. 2 and 3, both the numbers of participants in the most popular courses on Chinese University MOOC and Zhihuishu show a general upward trend in the first ten semesters. In the following periods, the numbers of online learners show a stable trend because the number of online students has reached its peak, and most students have returned to the face-to-face teaching mode, while online learning still provides assistance for both teaching and learning. In the future metaverse world, the number of online learners is bound to rise.

Problems in current MOOCs

Traditionally, there are two modes for recording online course videos: recording lecture videos in professional studios or offices and recording lecture videos in an ordinary classroom. Generally, lecture videos are recorded in studios or offices. Nevertheless, photographic specialists and expensive photographic equipment are involved in producing the online courses mentioned above, and most of the lecture videos for MOOCs are recorded in studios. This recording mode not only leads to a high economic cost associated with special photographic equipment (Kellogg 2013), but also requires a lengthy period of producing high-quality videos (Kolowich 2013; Guo et al. 2014). Meanwhile, the instructors who are used to face-to-face teaching in classrooms are distracted by a small screen, and lack enthusiasm due to the absence of students. Consequently, this will increase the sense of isolation for online learners, which is an inherent problem in online courses (Dixson 2010; Martin et al. 2018). At the same time, due to the asynchronism of teaching and learning, lecturers cannot observe online learners’ responses to the current teaching, lack interaction with students, resulting in an inability to determine the teaching effectiveness. There is no direct connection and effective interaction throughout the entire teaching process, either for lecturers or for online learners, which is very important for improving teaching effectiveness. The statistical results (Martin et al. 2018) show that lecture videos are strongly related to instructor presence and instructor connection.

When the online students feel more connected to the lecturer, they dispel fears and anxieties about the contactless online environment. In addition, the accessibility and passion of the lecturer a critical factors that could promote students’ engagement in open online courses (Hew 2016). Some lecture videos of online courses are recorded in classrooms directly. For example, the videos of the Stanford CS231n course taught in the Spring of 2017 have been uploaded to YouTube (Stanford 2019) and are available to the public. Some Open Yale Courses are made up of videos of the instructor lecturing in traditional classrooms (Toven-Lindsey et al. 2015). However, the scope of the lecturer’s movement is limited by the fixed camera. In addition, the recording process requires manual or semi-manual intervention.

Some capturing systems have been developed to adjust the shooting angles according to the lecturer’s position, while these systems are not suitable for general use in practical applications due to the requirement for specific equipment. Automatic lecture recording systems are proposed to capture the lecturer in a classroom via advanced cameras that can automatically pan, tilt, and zoom to track a moving object (Liao et al. 2015; Chou et al. 2010). A PTZ camera (Chou et al. 2010) is used to shoot a lecture, but the quality of the recorded video is very poor because of frequent zooming. The same situation exists in (Liao et al. 2015). As soon as the object is lost, the camera zooms in to detect the object in the entire area. After detecting the object, the camera zooms out to focus on the lecturer. Various devices are utilized in lecturer tracking systems to enhance the quality of the recorded videos and maintain the stability of the system. A wide-angle camera is involved in GaliTracker, a lecturer tracking system designed in (González-Agulla et al. 2013), to cover a larger area. In order to achieve real-time performance, two machines are required to capture and track separately. In (Ketterl et al. 2014), a lecture recording system using a graphics processing unit to analyze the movement of the object to be tracked is proposed. A depth camera and multiple video cameras are used to capture a presenter during a talk (Winkler et al. 2012), and the calibration based on the actual body proportions of the presenter is performed before tracking.

Due to the requirements of specific equipment and limitations in frequently changing the focus of cameras, the automatic lecturer tracking systems mentioned above are not suitable for general use in practical applications. Moreover, all primary, secondary, and university schools in China adopted online learning to replace face-to-face learning during the first half of 2020 to cope with the impact of COVID-19. Due to the limitation of network bandwidth, it is difficult to realize large-scale online live teaching. Although many schools adopt recorded broadcasting, live broadcasting to promote teaching via online course platforms, such as Chinese university MOOC, Chaoxing Xuexitong, Tencent Meeting, etc., the effectiveness of online courses is far less than that of offline courses because of the lack of face-to-face interaction.

To resolve the problems mentioned above, artificial intelligence technology is combined with wireless communication sensing technology in this paper to design a low-cost and stable lecturer tracking system. Further, the 5 G wireless communication network is involved in constructing smart classrooms, realizing the synchronization between offline and online teaching. The proposed Meta Classrooms based on VR technology in this paper enhance the immersion experiences of online learners.

5G-based Meta Classrooms

Research motivation

The MOOCs provide opportunities for undergraduate students, adult learners, and citizens to share the OERs provided by the top-ranked universities and to pursue their lifelong learning. However, online courses must perform the basic teaching function of pedagogy before performing other functions. As stated by Daphne Koller, one of the founders of Coursera, “We don’t believe that computers should replace teachers. We think computers can enhance the work of teachers” (Kolowich, 2013). In other words, some instructional behaviors of the teachers in traditional face-to-face classrooms should be exhibited in MOOCs. One of the most important behaviors is the presentation of knowledge and techniques. Lecture videos are carefully recorded and embedded in most MOOC platforms.

Due to the requirements of specific equipment and limitations in frequently changing the focus of cameras mentioned above, a lecturer tracking system based on human location and wireless sensing technologies is proposed in this study, which can effectively track and capture the lecturer by one camera and be executed on a general-purpose CPU in real-time. Furthermore, the fifth-generation mobile wireless communication system is utilized to improve the interactivity and synchronization of online courses. The effectiveness of online and offline teaching is also improved by detecting offline learners’ behavior and assessing online learners’ emotions.

Implementation scheme

Framework of 5G-based Meta Classrooms

The smart classroom proposed in this paper involves the following technologies: lecturer image collection and analysis, human detection network architecture design, wireless communication network construction based on Arduino, Infrared (IR) sensing technologies, and 5 G wireless communication network. Figure 4 describes the framework of the 5G-based Meta Classroom proposed in this paper.

Fig. 4
figure 4

Framework of 5G-based Meta Classroom integrated with lecturer tracking system.

Full size image

The following four functions are involved in constructing the 5G-based Meta Classroom as shown in Fig. 4:

  1. (1)

    Lecturer and students’ image collection and analysis: capturing and labeling different teaching scenarios to construct the training dataset, introducing other existing related datasets to enlarge the training dataset, and preprocessing the captured video data to deal with the problem of the repetitiveness of the training dataset.

  2. (2)

    Lecturer detecting and tracking system: extracting the feature of the lecturer via CNN, RNN, or Attention Model as shown in Fig. 4, outputting the location of the lecturer by a deep neural network, and constructing wireless communication network by rotating a servo motor to track the lecturer by an Arduino microcontroller board, which also works as an Access Point and receives data collected by IR thermal sensors via wireless communication devices. IR sensors are utilized to locate the lecturer when the deep neural network fails to detect the lecturer and ensure the stability of the system. As a result, the lecturer is kept in the center of the screen during the capturing process.

  3. (3)

    Synchronous online and offline teaching: the fifth-generation mobile wireless communication system is utilized to connect the lecturer tracking system to the internet and achieve the synchronization between offline and online teaching via the advantages of 5 G in higher data rates and massive connectivity.

  4. (4)

    Meta Classroom in the digital world: constructing a virtual classroom that can accommodate digital avatars of the lecturer and all students, including both students in the classroom and online students. The experience and learning efficiency of online learners are greatly improved.

Key technologies

According to the framework of 5G-based Meta Classroom mentioned in the previous section, the following key technologies are to be addressed in subsequent research:

  1. (1)

    Cost of human detection: considering the computational cost of the detection method and the economic cost, which are related to the application value of the smart classroom.

  2. (2)

    Balance between the depth and breadth of data collection: collecting image data in different scenes to prevent the neural network from overfitting with certain time and economic cost.

  3. (3)

    Cooperation between different devices: covering the whole platform in a classroom via two IR thermal imaging sensors. Due to the memory shortage of sensors, it is difficult to realize synchronization between Arduino UNO WiFi and two WiPy 3.0 modules, which are connected with IR thermal sensors.

  4. (4)

    Real-time performance of 5G-based Meta Classroom: ensure synchronization between lecturer tracking and online broadcasting after connecting to a 5 G wireless communication network, which is key for this study, moving from theory to practice.

  5. (5)

    Virtual classrooms and digital humans: matching each entity in real teaching scenarios and his/her digital avatars in Meta Classroom and developing virtual teaching assistants to make online courses more engaging and provide timely help to students in self-study.

Objectives and methods

To ensure the successful implementation of this study and make progress step by step, the objectives are divided into three phases:

  1. (1)

    Short-term objectives: locating the lecturer based on human detection and sensors, and capturing the lecturer after tracking automatically.

  2. (2)

    Medium-term objectives: achieving the synchronization between offline and online teaching by the use of 5 G and VR technologies.

  3. (3)

    Long-term objectives: improving the interactivity of online courses via action recognition and analysis technologies.

Each phase is a relatively independent part with a certain function. As a result, the objectives of Meta Classrooms can be gradually achieved, and the results of each phase can be independently developed, tested, and improved.

Method to achieve short-term objectives

Learning-based human detection methods have received much attention in recent years (Li et al. 2018; Zhang et al. 2016; Liu et al. 2023). Tracking by human detection is a common method for human tracking. As shown in Fig. 4, a human is detected and located in the images captured by the camera via convolutional neural networks (CNN). Then, a control module is involved to rotate the camera and track the lecturer. However, abrupt and rapid movements of objects can cause detection failure in the learning-based human location methods (Nam and Han, 2016; Wei et al. 2023; Yun et al. 2017). The human position obtained by the IR thermal sensors is transmitted to the system. As a result, detection failure caused by abrupt and rapid movements is prevented, and the non-real-time problem for thermal sensors is solved via the combination of human detection and IR thermal sensing technology.

The experiments in a classroom and a laboratory are conducted to evaluate the efficacy of the locating lecturer method based on human detection and sensors. As shown in Fig. 5, there are four distances related to the experiment, including (a) the distance between two IR thermal sensors, (b) the height of the sensor, (c) the distance between the camera and the platform, and (d) the range of the lecturer’s movement. Figure 5 gives the values of the distances of the experiment conducted in the classroom. Those of the experiment conducted in the laboratory are 2.02 m, 3.30 m,1.52 m, and 2.45 m, respectively.

Fig. 5
figure 5

The experiment was conducted in a classroom.

Full size image

Table 1 gives the comparative results of the method with IR thermal sensors to locate the lecturer and the method without IR thermal sensors. F_Num represents the number of total frames in the video captured by the camera. C_Num represents the number of frames in which the lecturer is in the center of the screen. I_Num represents the number of frames in which the lecturer appears on the screen. C_Rate represents the ratio of C_Num to F_Num. The higher the C_Rate, the higher the efficacy of the proposed method to locate the lecturer. If the lecturer cannot be kept in the center of the screen, the second-best thing is to ensure successful tracking, and the lecturer appears on the screen. Therefore, I_Rate is considered to represent the ratio of I_Num to F_Num.

Table 1 Comparative results of the ratios in the two methods.
Full size table

As shown in Table 1, the average C_rate of the lecturer locating method with sensors is higher than that of the method without sensors by 15.61%, and its average I_rate is 21.40% higher. Consequently, the involvement of IR thermal sensors significantly improves the efficacy of the method to locate the lecturer.

The proposed lecturer detecting and tracking system is compared with the existing localization and tracking methods. Table 2 describes the pros and cons of these methods. Some of these methods are device-based methods and the non-real-time system (Hu et al. 2020; Neupane et al. 2021; Xu et al. 2017). A pedestrian positioning method based on magnetic perturbation and decision tree delivers higher accuracy (Hu et al. 2020). Multiple antennas and Random Forest learning are used to realize indoor human localization (Neupane et al. 2021). An indoor human localization method is proposed to improve the localization robustness by ultra-wideband (UWB) and extended finite impulse response (EFIR) technologies (Xu et al. 2017). Some other learning-based human localization and tracking methods have been proposed in recent years (Nam and Han, 2016; Wei et al. 2023; Yun et al. 2017). A pretrained CNN network consisting of shared layers and multiple domain-specific layers is used to realize visual tracking (Nam and Han 2016). Autoregressive Visual Tracking (ARTrack) is an encoder-decoder architecture to simplify the tracking pipeline, obtaining more coherent tracking results (Wei et al. 2023). A visual tracker with light computation and high accuracy is designed to pursue the target object by deep reinforcement learning (Yun et al. 2017). For these learning-based human localization and tracking methods, a GPU is essential to achieve high accuracy and performance. As shown in Table 2, the proposed method is the optimal scheme in the compared localization and tracking methods because of low cost, real-time performance, and contactless devices.

Table 2 Comparison of the existing localization and tracking methods.
Full size table

Real-time locating and tracking of the lecturer realizes synchronous online and offline teaching, allowing online learners to obtain the teacher’s teaching information from the first perspective, shortening the distance between the lecturer and students, and reducing the isolation of online learners.

Method to achieve medium-term objectives

The coming 5 G mobile technologies have the following features: massive device connectivity, higher capacity, higher data rate, lower E2E latency, lower cost, and consistent quality of experience (QoE) provisioning (Attaran 2023; Agyapong et al. 2014). According to the official statistics published in May 2024 on the GSA(Global mobile Suppliers Association) website (GSA 2024), there are 585 operators in 175 countries and territories that have invested in 5 G by the end of March 2024. In addition to these operators, nearly 200 other companies have acquired priority access licenses in the US auction of CBRS spectrum to potentially be used for 5 G. The number of announced 5 G devices increases by 107% compared to that the start of 2022.

Figure 6 illustrates the top 20 regions with the highest 5 G Connectivity Index (5 GI) in 2023 compiled by GSMA (Global System for Mobile communications Association) Intelligence (GSMA Intelligence 2024). To evaluate the deployment of 5 G, the Connectivity Index is proposed as an indicator with a value between 0 and 100, the higher the better. 5 G is constructed by two categories, 5 G infrastructure and 5 G services, considering 17 factors, such as band spectrum, download speeds, data affordability, and data traffic per user. As shown in Fig. 6, there are 13 regions with an index greater than 50%.

Fig. 6
figure 6

The top 20 regions with the highest 5 GI in 2023, published by GSMA Intelligence (GSMA Intelligence 2024).

Full size image

Meanwhile, GSMA publishes the overview of the development of 5 G (GSMA 2024). The report gives a lot of comparative data about the number of mobile internet users, the percentage of connections of 5 G, the number of licensed cellular IoT connections, etc., including real data for 2023 and predicted data for 2030. Figure 7 indicates the penetration of 5 G in the mobile communications market in eight regions. Both Figs. 6 and 7 imply that 5 G communication technology has a rapid development trend.

Fig. 7
figure 7

The penetration of 5 G in the mobile communications market, published by GSMA (GSMA 2024).

Full size image

The development and deployment of 5 G technology provides a breeding ground for constructing intelligent education. In March 2019, China Unicom and Shanghai University of Engineering Science jointly established China’s first 5 G university. In May 2019, Baidu VR, Shanghai Telecom, and Shanghai Yuyuan Road No.1 Primary School reached a cooperation to realize VR teaching in a 5 G environment in the future via combining Baidu VR classrooms with cloud VR and 5 G infrastructure. In 2020, Vodafone and Coventry University launched the UK’s first 5 G Standalone (5 G SA) media innovation lab to enable students majoring in healthcare to experience the immersive XR tour of the human body.

Key technologies for 5 G wireless communication networks can accelerate the construction of the smart classroom proposed in this paper. By integrating the lecturer tracking system with 5 G, the synchronization between offline and online teaching will be achieved via the advantages of 5 G in higher data rates and massive connectivity. Online learners can experience the teaching effects of a face-to-face course by watching the lecture video captured by the lecturer tracking system as if they are in the classroom and place themselves in a virtual classroom, including digital avatars of all students in the real classroom and online students. Furthermore, 5 G communication technology provides a more reliable guarantee for the operation of Meta Classroom. Figure 8 gives the diagram for 5G-enabled Meta Classroom by utilizing network slicing (NS) and mobile edge computing (MEC), which are essential technologies of 5 G communication networks. The intelligent education ecosystem grafted on Meta Classroom can provide functions such as digital educational resources, virtual classrooms, and real-time learning surveillance. On the one hand, NS technology makes the campus network more secure and stable by setting up a customized virtual logical network with high speed and high stability for independent services. On the one hand, MEC provided by edge cloud servers can enhance the security of local data by avoiding transmission over public networks and reduce the data delivery latency. The investment of governments and leading companies around the world in the infrastructure of 5 G and VR technologies provides a breeding ground for the application of Meta Classrooms and reduces costs for the widespread construction of Meta Classrooms.

Fig. 8
figure 8

Diagram for 5G-enabled Meta Classroom based on NS and MEC.

Full size image

In order to determine the acceptability of the synchronization between offline and online teaching, we surveyed 288 students from different colleges on the acceptance of different online teaching modes and concerns of participating in online learning. The three typical online teaching modes include:

  1. (1)

    Synchronous online and offline teaching: Interactive online live teaching, synchronized with the classroom teaching.

  2. (2)

    Combination of online and offline teaching: being accompanied by a small number of revision lessons and a certain number of online quizzes.

  3. (3)

    Online non-credit courses: self learning, previewing, and reviewing of face-to-face learning.

As shown in Fig. 9, 89.58% of the respondents are willing to accept the synchronous online and offline teaching mode. In contrast, less than 50% of respondents can accept the online and offline blended teaching mode.

Fig. 9
figure 9

The acceptance of different online teaching modes.

Full size image

Concerns of participating in online learning include Learning effects, Earning credits, Portability of the device, Interactive experience, Network transfer speed, and Investment in purchasing devices. The biggest concern is Learning effects as described in Fig. 10, which is followed by more than 76.04% of respondents. In addition to Earning credits and Portability of the device, Interactive experience and Network transfer speed are also noted by more than 55% of respondents. Meta Classrooms can solve the above problems mentioned in section 2.3 by improving learning effectiveness via realizing synchronous online and offline teaching, and enhancing interactive experience based on 5 G and VR technologies.

Fig. 10
figure 10

Concerns of participating in online learning.

Full size image

Method to achieve long-term objectives

Emotions and sentiments play an important role in daily learning and communication (Poria et al. 2017; Liu et al. 2013). Sentiments are usually divided into positive, negative, and neutral (Poria et al. 2017). Affective Computing is benefit in solving the problem of the sense of isolation for online students in distance education (Liu et al. 2013). It is also useful to observe the state of an individual in a large number of offline learners.

For credit courses, learning supervision is essential to improve learning effectiveness. In Meta Classrooms, the recognition of students’ behaviors and expressions will serve as a reference for monitoring their learning status, while clearly stating the purposes, means, scope of processing biometric information and the length of time for storeing biometric information in accordance with local law and with the authorization of students or their guardians before collecting data.

Action recognition has great potential for applications in video surveillance, human computer interaction, and video content analysis (Wang et al. 2015; Simonyan and Zisserman 2014). For further applying the lecturer tracking system in MOOCs, action recognition will be involved to improve the interaction between the lecturer and the online learner. Furthermore, the control commands can be sent to the system via lecturer action recognition. There are three objectives in applying action recognition in Meta Classroom:

  1. (1)

    Distinguishing the lecturer’s activity: standing, demonstrating, waving a hand, or writing on the blackboard.

  2. (2)

    Analyzing the activities of the students in the classroom: writing, standing, sitting, raising a hand, or lying on the desk.

  3. (3)

    Observing the response of online learners: nodding head, shaking head, or raising hand.

Affective computing, as described in Fig. 11, maps the posture of offline learners, and the expression of online learners is mapped to six emotions in three dimensions: Expectation, Excitement, and Attention. After calculating student emotions based on fuzzy classification methods, the correlation model between emotions and learning states is constructed to provide a reference for the comprehensive assessment of the effectiveness of instruction.

Fig. 11
figure 11

Assessment of student learning state.

Full size image

Hidden Markov Model (HMM) statistical model used to construct the correlation model between emotions and students’ learning status. As shown in Fig. 12, the results of emotion calculation obtained after clustering are fed into the HMM statistical model, taking the six emotions of negative, positive, calm, excited, inattentive, and attentive as observation sequences, and taking the three categories of learning status of relaxed, stressed, and opportune as hidden states. TP in Fig. 12 represents the probability of state transition, and OP represents the probability of generating observed states. The results of the state analysis can provide a reference for lecturers to adjust the teaching progress and method.

Fig. 12
figure 12

Clustering of emotions to learning status.

Full size image

Recently, many virtual classroom schemes have been proposed to enhance the experiences and outcomes of online learning. Some schemes increase the immersion of online learning by utilizing wearable devices (Makransky and Mayer 2022; Seufert et al. 2022). Another popular virtual classroom solution is online live streaming implemented with digital communication platforms (Islam et al. 2023; Florez 2022; Ho et al. 2023), such as Zoom, Google Meet, or Skype. As shown in Table 3, Meta Classrooms is superior to other compared virtual classroom schemes, improving the experience of online learners via using 5 G and virtual technologies, further ensuring high learning outcomes by monitoring and interacting with learners.

Table 3 Comparison of the existing virtual classroom schemes.
Full size table

Concretely, through real-time observation and rapid analysis of online and offline students’ behaviors and expressions, the lecturer can receive feedback on students’ learning status, and then timely adjust teaching progress and content, or even pause to interact with students. Therefore, Meta Classrooms can effectively improve teaching outcomes, especially for large-scale courses with large audiences.

Conclusion

This paper discusses the development status of MOOCs in America and China in the past decade, and analyzes the main MOOC platforms in both countries, such as edX, Coursera, Chinese University MOOC, and Zhihuishu. In order to resolve the problems of current MOOCs in recording online course videos and the experience of online learners, the framework of 5G-based Meta Classroom integrated with lecturer tracking system is proposed in this study to ensure the teaching effectiveness of online and offline learning via synchronous online and offline teaching, accurate assessment of students’ learning status, and interesting interactive experience. The objectives, key technologies of the three stages for constructing Meta Classrooms, and experimental results obtained from achieving the short-term objectives are described in detail. In addition, the scheme of Meta Classrooms is also compared with the existing virtual classroom. The survey on the acceptance of synchronous online and offline teaching modes verifies that constructing Meta Classrooms has a broad application prospect.

The low-cost, real-time 5G-based Meta Classroom will be constructed to realize the synchronization between offline and online teaching by combining human detection based on neural network, contactless thermal sensing technologies, 5G wireless communication network and VR technology, which will create an immersive virtual learning environment, suitable for a widely applicable student-centeredness, personalized teaching and learning ecosystem and accelerate the development of intelligent education.