Digital media recording and broadcasting classroom using Internet intelligent image positioning and opinion monitoring in communication

Abstract

Purpose

The digital media recording and broadcasting classroom using Internet real-time intelligent image positioning and opinion monitoring in communication teaching is researched and analyzed.

Design/methodology/approach

First, spatial grid positioning and monitoring and image intelligent recognition technologies were used to extract and analyze teaching images by mastering Internet of Things (IoT) technology and establishing an intelligent image positioning and opinion monitoring digital media recording and broadcasting system framework. Next, a positioning node algorithm was utilized to measure the image distance, and then a moving node location model under the IoT was established. In addition, a radial basis function (RBF) neural network was used to realize the signal transmission function. The experimental data of the adopted RBF based on the optimization of the adaptive cuckoo search (ACS-RBF) neural network, particle swarm algorithm neural network, and method of least squares optimization were compared and analyzed. In addition, a more efficient RBF neural network was adopted. Finally, the digital media recording and broadcasting classroom scheme of real-time intelligent image positioning and opinion monitoring was designed. In addition, the application environment of digital media actual teacher teaching was detected, and recording and broadcasting pictures were analyzed and researched.

Findings

The actual value, predicted value, and the number of predicted samples of the ACS-RBF model were all better than those of the two other neural networks. According to the analysis and comparison of the sampling optimization Monte Carlo localization (SOMCL), Monte Carlo, and genetic algorithm optimization-based Monte Carlo positioning algorithms, the SOMCL algorithm showed better robustness, and its positioning efficiency was superior to that of the two other algorithms. In addition, the SOMCL algorithm greatly reduced the positioning and monitoring energy consumption.

Originality/value

The application of real-time intelligent image positioning and monitoring technology in actual communication teaching was realized in the study.

1. Introduction

With the continuous development of information technology and multimedia technology, the teaching field has also greatly developed. In addition, the combination of information technology and multimedia technology can better disseminate knowledge and promote the change in current teaching methods of teachers with the change of times rather than the application in classrooms only. Consequently, better teaching for the new generation of students is achieved (Li et al., 2019). In this context, teachers are no longer the only people who undertake teaching tasks, but they also promote the establishment of some educational institutions. These institutions make full use of information methods to realize the teaching for students and become the precursor of modern education. In addition, they actively introduce new technology and modes into teaching (Waleed et al., 2020). Because of the above innovations and attempts, the current teaching mode becomes more efficient than before. Enrolled students are no longer the only educated population, and employees and retirees also receive education. In addition, a new teaching mode is comprehensively developed in actual application with higher alterability. This mode has become an academic branch of computer in the direction of vision with a pivotal role and the application subject researched by many teams (Khanh et al., 2020). At present, there are an increasing number of industries of the application of identification detection technology and real-time tracking technology with increasing maturity. The application of the two technologies can be effectively applied in each industry, such as face recognition, video surveillance in public areas, enterprise staff management, military analysis, and many other industries. Their widespread application is credited to the rapid development of Internet technology in recent years (Yuta et al., 2020). Because of multimedia technology faster information technology and adaptable multimedia technology, multimedia recording and broadcasting system with real-time image tracking and positioning as the core technology is valued by many technical teams and popular among a large number of online educational institutions. Hence, this teaching mode has become the hot research direction of many technical teams at present and a reference for many colleges and universities. Although the current practical application of this teaching mode is not mature enough, a technical reference is provided for the mode in considerable studies.

Wang et al. (2020a) combined the Internet of Things (IoT) technology and artificial intelligence algorithm in their study to establish the autonomous control model of unmanned aerial vehicle. They designed an unmanned aerial vehicle formation control law from the two aspects of interference and non-interference. According to the basic architecture of the IoT, the ad hoc network of ZigBee was simulated to adopt the working mode of IoT modules in node equipment to propose an IoT-based self-adaptive networking scheme. Asghari et al. (2019) discussed the achievements and disadvantages of each research into IoT and put forward some suggestions on the solution of the disadvantages as well as future challenges and problems to be addressed in IoT application. Mushtaq et al. (2019) researched the performance of a position-based routing protocol Geocast self-adaptive Mesh environment. The environment was used to collect the routing for context information from the global positioning system and internal navigation and discovery framework. These systems were assessed according to various indexes, including accuracy, data package delivery rate, and data package cost. The research achievements could be applied in different fields, such as building navigation, hospital patient tracking, situational context awareness service supply in smart cities, and the deployment of Industry 4.0.

The research focused on the application of IoT technology in communication teaching in digital media recording and broadcasting classrooms to realize the positioning and monitoring of real-time intelligent images for the acquisition of more accurate and clearer images. The innovations of the research lie in the real-time positioning and monitoring of video images by IoT technology and the effective image nodes by the sampling optimization Monte Carlo localization (SOMCL) algorithm. As the technical traction of communication teaching, human behavior identification and moving target detection is a revolution to traditional teaching. In addition, this technology can also be applied in the teaching of other disciplines to show excellent teaching effects. It enables students to understand the teaching contents through digital media recording and broadcasting video. This research not only provided technical support for the optimization of English teaching process, but also contributed to the development of education and teaching in current society.

2. Scheme design

2.1 Perception layer of IoT technology

The perception layer is the core of IoT in sensing objects, which has also become an important part of the development process of IoT. For the first time, the International Telecommunication Union incorporated sensing technology and radio frequency identification technology into key technologies related to IoT (Gu and Jia, 2021).

1. The sensor is the most important conduction device in sensing technology because it is the key conduction device that completes the data interaction between people and objects or objects and objects in the IoT. It is also definitely the most important data collector in the IoT. The sensor can collect the relevant data of the surrounding environment and use the communication method to transmit the sensed data to the network layer and the application layer for data processing. Finally, the diversification of the application is realized (Li et al., 2020; Abro et al., 2019).

2. Radio frequency identification technology

Radio frequency identification technology, as an important part of communication technology, is mainly used to develop IoT technology. The composition of radio frequency identification systems is relatively simple, generally including electronic tags and readers. Radio signals are mainly used in the radio frequency identification system to acquire data of target objects to make it convenient for users to identify various objects (Chiwamba et al., 2019).

1. Wireless sensor network

In general, as an important technology of the IoT, the wireless sensor network belongs to the perception layer. The composition of the wireless sensor network is relatively complex, consisting of many microsensor nodes. These dynamic or static nodes are connected to each other to form a network (Luo et al., 2020). The nodes sense the target object and collect data through the wireless network to realize identification. Then, the routing protocol transmits the collected information to the node. Finally, the wireless data receiving device sends the data to the control center to complete the data transmission and interaction. Figure 1 shows the structure of the wireless sensor network.

In wireless sensor networks, the computing, communication, and energy of nodes are very limited. The nodes in micro embedded wireless sensor networks generally have terminal and routing functions. It can not only process local information but can also forward the data and information sent from other nodes that work together with other nodes to complete tasks (Liu, 2021). The sink node closely connects the wireless sensor network and the internet as well as other extranets as the link. In addition, it has the function of forwarding the acquired data and publishing detection tasks. In general, the computing, communication, and data processing of the sink node are stronger than those of common nodes. The sensor network node consists of four modules, including a sensing module, processing module, communication module, and power component, as shown in Figure 2 below.

The sensing module is used to collect the information and data within the monitoring environmental range and then convert the data into A/D. The processing module computes and processes the collected data based on certain algorithms or rules and publishes work instructions for the interior of the sensor node itself. In addition, the supply of the power module is adjusted in a timely manner based on the energy consumption of the node. The communication module uses a wireless communication device to realize the communication between sensor nodes as well as between sensor nodes and user node management control nodes, interactively controls information, and receives and sens business data. The power component provides energy for the above three function modules to ensure the continuous and stable operation of the node system.

2.2 The framework of the digital media recording and broadcasting system

2.2.1 Theory

From the perspective of system function, the digital media recording and broadcasting system for classroom teaching can be divided into multiple functional modules, including the traditional digital media classroom function module, the external environment module (including lighting system and sound absorption system), and the recording and broadcasting module (Dimililer et al., 2021). Its brief framework is shown in Figure 3.

The digital media classroom function module is the basic function module facing normal course teaching. It is derived from the need for digital media demonstration in classroom teaching at the beginning of system development and the auxiliary demonstration in the classroom environment based on traditional blackboard teaching, including the platform containing the control main engine and project equipment. The external environment module contains the function modules added to improve the digital media audio-visual effects in class. To achieve clearer classroom recording, a reasonable indoor lighting configuration is required to solve the problem of uneven picture brightness. In addition, a special lighting system is equipped. To address a series of problems in the entire video recording process, including noise and crosstalk of multiple sound sources, a sound-absorbing system and sound pick-up equipment distributed in classrooms are adopted. In addition, sound-absorbing systems can address the problem of various kinds of indoor sound echoes. However, the above equipment is just icing on the cake for digital media recording and broadcasting systems realizing basic video recording functions. They can be used as the supplementary function module to achieve recording and broadcasting effects perfectly. Nevertheless, not all systems need necessary modules for widely applied recording and broadcasting systems.

The digital media recording and broadcasting module is the core module and core function module of the whole recording and broadcasting system. It includes complex submodules. In most cases, the recording and broadcasting module of a digital media recording and broadcasting system includes classroom image collection, teaching audio sound collection, video graphics array (VGA) collection, automatic course image tracking, multisource image control switch, information digitization processing, sound amplification, and other submodules. From the perspective of functional realization, the aforementioned sound pick-up equipment distributed densely inside classrooms can also be included in digital media recording and broadcasting system. In brief, the main functional framework of the recording module is the collection of images, sounds, and computer VGA signals, a series of digital processing of the captured original materials, and the centralized display (either real-time live or provision after postediting).

2.2.2 The positioning and monitoring technology in the digital media recording and broadcasting system

1. Spatial grid positioning and monitoring technology

The spatial grid is the small three-dimensional grid divided from a three-dimensional picture in the recording process, which is then marked and located (Islam et al., 2020). Through conducting the spatial grid on the real-time teaching of teachers, the digital media recording and broadcasting system can quickly analyze the three-dimensional coordinates of teachers and students in the classroom to realize real-time and comprehensive tracking.

1. Real-time image recognition technology

Real-time image recognition technology uses a positioning camera to locate a moving object. Students are regarded as the research object. When students make some actions, there are differences in the characteristic images in the camera of students in a series of actions. A real-time image recognition system can exactly cover the height difference in the image (Rahman, 2021; Chen and Yang, 2021; Lin, 2020). When the positioning camera successfully captures a student beyond the preset height, it will make the judgment that the student is answering questions. At this time, the camera will accurately track and locate the standing students.

2.3 Node positioning algorithm of the IoT

2.3.1 Static node positioning algorithm

The node positioning algorithm based on distance measurement can obtain a better positioning effect in general. The typical node positioning algorithms are TOA (Time of Arrive) and AOA (Angle of Arrive) (Zhou and Gao, 2019).

1. TOA algorithm

TOA is a node positioning technology based on signal transmission. The distance between nodes is obtained by multiplying the radio wave propagation time and propagation velocity between unknown nodes and anchor nodes. This algorithm requires accurate time synchronization between nodes and at least three nodes in two-dimensional space to confirm the position of an unknown node (Zhang et al., 2020). TOA positioning technology is shown in Figure 4.

TOA usually uses the local clock to know the propagation time. However, TOA has higher hardware requirements due to the potential error of the local clock because the wireless wave propagation time is fast.

1. AOA algorithm

This algorithm judges the distance between nodes mainly by the angle between the signal and the antenna array when the signal reaches the node. The AOA algorithm is not suitable for node localization due to the large overhead of the loaded antenna array in AOA and the great influence on the algorithm from the external environment (Khan and Al-Badi, 2020; Omar, 2021).

2.3.2 Mobile node positioning algorithm

The mobile anchor node positioning algorithm based on RSSI (Received Signal Strength Indication) is used to broadcast the data information of anchor nodes periodically during the movement of anchor nodes. After receiving the information, the unknown node obtains a position estimation of its own coordinates according to the relationship between signal strength and distance (Wang et al., 2020b).(1)$C(x,y)=P(d((x,y),\left(x_B,y_B\right)))\forall (x,y)\in \left[\left(x_{\min },y_{\min }\right)\times \left(x_{\max },y_{\max }\right)\right]$

In Equation (1), P is the probability density function of the distance corresponding to RSSI. In addition, d is the Euclidean distance between the virtual beacon (x_B, y_B) and the unknown node (x, y), while x_min, y_min, x_max, and y_max are the boundaries of the monitoring area (Rho and Chen, 2019). The mobile anchor node positioning algorithm based on RSSI has signal attenuation and a multipath effect, which indicates that the monitoring environment has a great influence on the algorithm.

2.3.3 Node position prediction technology

The design of the mobile node position prediction model consists of three parts, including data processing, network modelling, and position prediction.

The Markov model is adopted due to its strong ability to process time series data, which is widely used in the field of position prediction. As a mobility model based on probability reasoning, the Markov model uses a historical observation sequence to predict the future position of nodes and regards the process of node position movement as the change process of the system state (Jian et al., 2021). For example, the current node position is S_k, so the prediction position of the node is S_k+1 at the next moment. The possible position set of nodes at the next moment is presented in Equation (2).(2)$S=\left\{S_1,S_2,···,S_i,···,S_n\right\}$

The probability of the node moving from the current position S_k to the next position S_k+1 is calculated according to Equation (3).(3)$P_{\max }=\max p\left(\frac{S_{k+1}}{S_k}\right)$

The Markov model believes that the location of the node at the next moment is only related to the location of the node at the current moment and independent of other historical locations, which is obviously unrealistic (Ren et al., 2021). Therefore, the K-order Markov Model is proposed, as shown in Equation (4).(4)$P_{\max }=\max P\left(\frac{S_{k+1}}{S_1{S}_2···S_i{S}_{i+1}···S_k}\right)$

The K-order Markov model inputs historical information into the calculation process so that the position information of the next moment of the node is closely related to the historical information of the node. Accordingly, the prediction accuracy of the model is greatly improved.

2.3.4 MCL (Monte Carlo Localization) algorithm

The core content of the MCL algorithm is to give a certain weight to all particles in the three-dimensional space. By predicting the particle sample set, the particle distribution in the next period is predicted. Additionally, the probability density function is improved by the recursive algorithm (Gomez et al., 2019). When there are enough particle samples, the probability density of the object can be accurately predicted by the MCL algorithm. The mathematical model of the MCL algorithm is presented as follows.

x_r is set as the dynamic variable in the algorithm model, and z_r is the observed data. The corresponding state transition equation is Equation (5).(5)$x_r=f_r\left(x_{r-1},v_{r-1}\right)$

Equation (6) shows the calculation of the observed data.(6)$z_r=h_x\left(x_r,n_r\right)$

In Equation (5) and Equation (6), f_r represents the state transition function, while r is the sampling number in the data observation process. In Equation (6), h_x is the observation function, and n_r is the detection of noise. The whole system state is related to that of the previous period.

2.3.5 Differential Evolution (DE) algorithm

The DE algorithm is more efficient than other evolutionary algorithms. It realizes the screening of excellent biological groups by studying the characteristics of organisms in the evolution process and according to the survival law of “natural selection, survival of the fittest” (Khaw et al., 2020). With an iterative process in the form of an expression, the algorithm obtains the optimal solution by selecting the initial species community and using the iterative algorithm.

2.3.6 SOMCL algorithm

The SOMCL algorithm first predicts the possible direction and speed of the node through the Newton interpolation method to reduce the survey range of the movement region of the node. Then, the DE algorithm is used to improve the probability density of particles to reduce the collection rate of invalid samples and indirectly improve the accuracy of the SOMCL algorithm.

1. Initial stage

N sample points are randomly collected in the monitoring area to form the position sample set of nodes at the initial stage, as shown in Equation (7).(7)$L_0=\left\{l_0^1,l_0^2,\cdot \cdot \cdot ,l_0^N\right\}$

1. Prediction stage

According to the coordinates at the latest three moments in the historical information queue of nodes, the velocity V of the node at the moment is calculated by the Newton interpolation method. In addition, the sampling area is a circle made with the position of the node at the previous moment as the center and min (V, V_max) as the radius. N sample points are randomly selected in the sampling area.

1. Filtrable stage

The position samples of particles obtained in the prediction stage cannot achieve accurate positioning. Therefore, particles failing to meet the research scope should be filtered. Equation (8) represents the filtrable stage. In Equation (8), S₁ is the first node set, while S₂ represents the second node set. Meanwhile, r is the allowable range of communication, and d (l_k, s) is the distance between the unknown node and the target node.(8)$Filter\left(l_t\right)=\left(\forall s\in S_1,d(l_t,s)\le r\right)\cap \left(\forall s\in S_2,d(l_t,s)\le 2r\right)$

1. Sample optimization stage

The sample particle set after the above stages is cross-operated and updated. After sample optimization, the particles in the sample are filtered again until N sample particle sets with experimental significance are obtained.

1. Location prediction

The average position coordinates of N valid samples are used as the estimated location of nodes. The flow chart of the SOMCL algorithm is shown in Figure 5.

2.4 Design of the mobile node positioning prediction model

2.4.1 RBF (Radial Basis Function) network theory

An RBF network is a neural network with a single hidden layer, which also includes an input layer and an output layer. The input layer becomes linear through a series of transformations in the hidden layer, while the mapping relationship between the hidden layer and the output layer becomes nonlinear (Arboleda, 2019).

As a forward feedback neural network, the RBF network has good nonlinear mapping ability through a three-layer network structure. The input layer and hidden layer of the RBF network are connected directly. In addition, various types of radial basis functions can be applied to the hidden layer. The common radial basis functions are shown in the following three equations.

1. The Gaussian function is shown in Equation (9).

(9)$R_i\left(x\right)=\exp \left[\frac{\Vert x_k-c_i\Vert }{2{\sigma }_i^2}\right]\left(i=\mathrm{1,2},\mathrm{...},P\right)$

1. The multiquadratic function is shown in Equation (10).

(10)$R_i\left(x\right)=\sqrt{{\Vert x_k-c_i\Vert }^2+c_i^2}\left(i=\mathrm{1,2},\mathrm{...},P\right)$

1. The thin plate spline function is shown in Equation (11).

(11)$R_i\left(x\right)={\Vert x_k-c_i\Vert }^2\lg \Vert x_k-c_i\Vert \left(i=\mathrm{1,2},\mathrm{...},P\right)$

In the above three equations, x_k denotes the N-dimensional input vector, while P is the number of neurons in the hidden layer, and ||x_k-c_i|| represents the European norm of x_k-c_i.

2.4.2 Establishment of the RBF network

The RBF network has excellent nonlinear mapping ability and generalization ability. The time sequence of the given mobile node position is used as training samples, and the RBF network trains the nonlinear mapping relationship between the fitting time and the corresponding mobile node position (Robert, 2018). Equation (12) is the calculation of the output of the RBF network.(12)$Y\left(X_n\right)={\displaystyle \sum _{i=1}^n{\omega }_i\phi \left(X_n,k_i\right)}+b_0$

In Equation (12), ω_i (i = 1,2 …, k) is the weight of the ith hidden unit before output, while b₀ is the offset of the output node. The base function φ (X_n, k_i) represents the output of the kth hidden unit, and k_i (i = 1, 2 …, k) denotes the center of the base function. The independent variable of the RBF network is the distance between the input vector and the threshold vector ||X_n-C_j||, which is the product of the input vector and the row vector of the weighted matrix C. The RBF network can transfer parameters in many forms.

The RBF network has a compact topology and structure parameters to realize separation learning, making its convergence speed fast. The nonlinear mapping ability of the RBF network is mainly determined by the center of the basis function. Usually, the data points are randomly selected as the center, which will make the performance of the RBF network poor and make it difficult to reflect the real input and output relationship of the system.

The prediction results of the ACS-RBF (radial basis function based on the optimization of adaptive cuckoo search) network, PSO-RBF (radial basis function based on particle swarm optimization) network, and OLS-RBF (radial basis function based on the optimization of ordinary least squares) network are compared to test the prediction effect of the RBF network. Other parameter settings are N_SIZE = 30; P_a = 0.25; G_MAX = 300; the input sample is the movement time of mobile node Y(X_n), among which the dimension n = 1, and the output Y(X_n) is the coordinate of the node at that time. In Equation (13), the dimension m = 2. The number of hidden layer nodes L is 3, calculated according to Equation (13).(13)$L=\sqrt{0.43nm+0.12m^2+2.54n+0.77m+0.35}+0.51$

2.5 Design of a real-time intelligent image positioning and opinion monitoring system for a digital media recording and broadcasting classroom

2.5.1 The whole digital media recording and broadcasting classroom mainly includes the hardware platform, the main control function, and the intelligent positioning and monitoring strategy

The complete system architecture is shown in Figure 6

The VP6802 digital media processing module under the Linux operating system ensures the accuracy of the results, which is better than the performance of the TMS320-DM8168 module on the application of high-definition images. In addition, the driver of the board in the VP6802 module can fully meet the requirements of 1080P60VGA/DVI video acquisition and support the application of h. 264 video decoder encoder.

The following is a description of the master control server. A mature commercial recording and broadcasting system is selected as the hardware. Figure 7 shows the overall logic structure of the hardware.

2.5.2 The target monitoring strategy is the core of the recording and broadcasting system, including three-dimensional coordinate positioning and opinion monitoring for teachers and students

1. Detection of students

Generally, students stay in a stable range in the communication classroom. Therefore, the system only needs to judge the condition that students disappear from the picture when communicating with the teacher.

1. Detection of the teacher

The teacher usually moves back and forth in communication teaching within the platform, which limits the detection range of the teacher. It is only necessary to complete the data monitoring of the horizontal position and contour height of the journalism and communication teacher on the platform. The monitoring camera adjusts the rotation angle corresponding to the teacher's activity range in teaching to successfully locate the teacher.

2.5.3 Logic of the algorithm

The image detection of the teacher or students is only a small target in the whole video image. Therefore, the operation of image segmentation is restricted to images of the teacher and students. In addition, only the useful background is selected as the segmented image to improve the comparison background. Figure 8 shows the process of target detection.

3. Design and analysis of the simulation experiment

3.1 Results of the SOMCL algorithm

Figure 9 shows the comparison of the sampling times over time of the MCL, GMMCL (Monte Carlo Localization based on genetic algorithm optimization) algorithm, and SOMCL algorithm.

From Figure 8, at first, the sampling times of the three algorithms gradually decrease with increasing time. When the timeslot number is greater than 80 units, the change in the sampling times tends to be stable. The SOMCL algorithm has the fewest sampling times among the three algorithms. This indicates that the SOMCL algorithm makes full use of the advantage of the Newton interpolation method in reducing the sampling area, making the sampling success rate higher, and then reducing the number of samplings. In addition, the SOMCL algorithm improves the target positioning efficiency and reduces the energy consumption in the calculation process.

Figure 10 shows the robustness analysis of the three algorithms.

Among the three algorithms, the MCL algorithm has the largest average positioning error and the most obvious fluctuation of positioning error. The positioning error and fluctuation of the SOMCL algorithm are significantly less than those of the two other algorithms. Figure 11 displays the error fluctuation comparison of the three algorithms.

The expected value and MSE (mean square error) of the positioning error of the GMMCL algorithm are 0.3243 and 0.330, respectively, while those of the IMCL (improved Monte Carlo node localization) algorithm are 0.6345 and 0.0479. In contrast, the expected value and MSE of the positioning error of the SOMCL algorithm are 0.2226 and 0.0164, respectively. The SOMCL algorithm has the strongest robustness because the DE algorithm used in the SOMCL algorithm has strong robustness, which increases the diversity of node location, enabling nodes to obtain more effective observation information.

3.2 Implementation of the RBF prediction model

Figure 12 displays the prediction effects of the ACS-RBF, PSO-RBF, and OLS-RBF models below, which are used to predict the moving node location. Figure 13 shows the comparison of prediction errors of the three prediction models below. Figure 14 demonstrates the comparison results of the three algorithms below.

Figure 12a–c describe the prediction effects of the ACS-RBF, PSO-RBF, and OLS-RBF prediction models, respectively. The results indicate that the three models all can accurately predict the location of moving nodes in the last 20 moments. According to Figure 12, it can be intuitively found that the ACS-RBF model showed higher prediction accuracy than the PSO-RBF and OLS-RBF models. To highlight the effects of the predication model more accurately, Figure 13 shows that the prediction error of the ACS-RBF model is greater than that of the PSO-RBF model only at the 1st and 2nd seconds and is the smallest among the three models during the whole time period.

Based on Figure 14, it is obvious that the three measure index values of the ACS-RBF model are optimal compared with those of the PSO-RBF and OLS-RBF models. According to the above experiments, the prediction accuracy of the ACS-RBF model for moving node location is the highest among the three prediction algorithms, which results from the strong global search capability and convergence speed of the adaptive cuckoo search algorithm. As a result, it can quickly and accurately find the most suitable parameter for the RBF neural network. The optimized prediction model can predict node location more accurately and realize the positioning and monitoring of moving nodes.

3.3 Demonstration of communication teaching in digital media recording and broadcasting classrooms

Figure 15 shows the target positioning and monitoring images of the communication teacher in the teaching video.

The teaching picture of the communication teacher can be seen clearly through the recorded video picture. The coverage, accuracy, and real-time performance of the teaching contents on the blackboard are all excellent. Because of the difference between domestic and foreign teaching modes, most teachers still follow traditional teaching methods at present. They give lessons in front of platform almost every class. Hence, the range of the target location is limited, and the model should be more appropriate for domestic communication teaching.

Figure 16 denotes the accuracy statistics of the real-time intelligent image positioning and opinion monitoring system for the digital media recording and broadcasting classroom.

According to Figure 16, the control accuracy of the system is close to 90%, which meets the basic requirements of the whole system. Meanwhile, the monitoring accuracy of students is higher than that of the teacher because it is easier to monitor the range of students' activities, which is smaller than that of the teacher.

4. Conclusion

IoT technology was used for the real-time positioning and monitoring of communication teaching pictures in digital media recording and broadcasting classrooms in the research. The analysis and comparison of the SOMCL, IMCL, and MMCL algorithms demonstrated that the SOMCL algorithms showed more significant robustness and higher positioning efficiency than the other two algorithms. The simulation experiment on the SO-RBF, OLS-RBF, and ACS-RBF models showed that the prediction accuracy of the ACS-RBF model for moving node location was the highest among the three prediction algorithms. It can predict node location more accurately and realize the positioning and monitoring of moving nodes in real-time images.

The theoretical significance of this research was as follows. A sampling optimization-based MCL algorithm was proposed. The prediction of the traditional MCL algorithm was optimized in the IoT and unknown node movement scenarios. The sampling area was reduced, while the diversity of node location was increased. In addition, the accuracy and efficiency of node positioning were improved, and the robustness of the algorithm was enhanced, whose practical significance lies in the promotion of the application of IoT technology in the research and development of communication teaching and digital media education recording and broadcasting system in an increasing number of colleges and universities. At present, some digital media recording and broadcasting systems have been successfully implemented in schools, which not only broadens the dissemination of teaching resources and the scope of teaching audience but also increases the diversity of teaching activities. Consequently, teaching activities are developed towards a more perfect direction. The research not only provided technical support for the development of informatization teaching but also contributed to the comprehensive development of social teaching.

However, there are some flaws in the research. Both pictures and images are two-dimensional so that the needs of the positioning and monitoring of three-dimensional images cannot be met. Therefore, three-dimensional image positioning will be the main research purpose of future studies, especially dynamic tracking under high-speed cameras. In addition, the stability and robustness of the algorithms will also be researched more extensively and profoundly in future.

Figure 1

Structure of the wireless sensor network

[Figure omitted. See PDF]

Figure 2

Node structure

[Figure omitted. See PDF]

Figure 3

Framework of the digital media recording and broadcasting system

[Figure omitted. See PDF]

Figure 4

TOA

[Figure omitted. See PDF]

Figure 5

Process of the SOMCL algorithm

[Figure omitted. See PDF]

Figure 6

Architecture of the recording and broadcasting classroom system

[Figure omitted. See PDF]

Figure 7

Hardware structure of the recording and broadcasting system

[Figure omitted. See PDF]

Figure 8

Process of the target detection algorithm

[Figure omitted. See PDF]

Figure 9

Comparison of the sampling times of the three algorithms (a: MCL algorithm; b: GMMCL algorithm; c: SOMCL algorithm)

[Figure omitted. See PDF]

Figure 10

Comparison of the positioning error of the three algorithms (a: MCL algorithm; b: GMMCL algorithm; c: SOMCL algorithm)

[Figure omitted. See PDF]

Figure 11

Error fluctuation comparison of the three algorithms

[Figure omitted. See PDF]

Figure 12

Forecasting results of the three prediction models (a: ACS-RBF model; b: PSO-RBF model; c: OLS-RBF model)

[Figure omitted. See PDF]

Figure 13

Prediction errors of the three models (a: ACS-RBF model; b: PSO-RBF model; c: OLS-RBF model)

[Figure omitted. See PDF]

Figure 14

Comparison of the effects of the three models

[Figure omitted. See PDF]

Figure 15

Images of the teaching video for the target monitoring algorithm

[Figure omitted. See PDF]

Figure 16

Accuracy statistics of the system (a: monitoring of the teacher; b: monitoring of students; c: picture switching; A: total frame number; B: number of monitoring frames; C: monitoring accuracy; D: number of events; E: switching times; F: switching accuracy)

[Figure omitted. See PDF]