Remote internet teaching system operation method

Abstract

The invention discloses a remote internet teaching system operation method. The method comprises the following steps: S1, executing edge caching and dynamic distribution; s2, implementing adaptive transmission protocol optimization; s3, P2P collaborative learning resource sharing is carried out; s4, implementing cloud edge collaborative intelligent interaction; and S5, constructing secure tunnel encryption transmission. According to the method, interaction data such as characters, voices and expressions generated by a user are preprocessed in real time through edge nodes, key information is extracted by means of a professional algorithm, the data volume is compressed, invalid information is filtered, and only necessary data is uploaded to the cloud. In a high-concurrency access scene, cloud computing and storage pressure is effectively reduced, service abnormity caused by data overload of the cloud is avoided, stable operation of cloud service is ensured, and a foundation is built for development of large-scale teaching activities.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet teaching, and in particular to a method for operating a remote Internet teaching system. Background Art

[0002] With the innovation of Internet technology, remote Internet teaching systems have broken through the time and space limitations of education, enabling students to access rich teaching resources without leaving home, and becoming an important force in promoting educational equity and expanding the boundaries of learning.

[0003] However, during the operation of such systems, many key problems that hinder the improvement of teaching experience and quality have emerged. Under the traditional centralized architecture, teaching resources are mostly stored on the central server. When a large number of students access popular resources simultaneously, the server load is too high, the transmission delay increases significantly, and the resource loading is slow. Existing systems mostly use a single transmission protocol and cannot flexibly switch according to the real-time network conditions. When the network is unstable, live courses frequently freeze and the audio and video are out of sync. In addition, existing platforms cannot comprehensively collect and analyze students' learning behavior data, making it difficult to accurately grasp students' learning characteristics and needs. The learning plans provided cannot meet the personalized learning requirements of different students, resulting in low student learning interest and efficiency, and it is difficult to achieve the expected learning effects.

[0004] Accordingly, this application proposes a method for operating a remote Internet teaching system. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies in the prior art and propose a method for operating a remote Internet teaching system.

[0006] To achieve the above objective, the present invention adopts the following technical solutions:

[0007] A method for operating a remote Internet teaching system includes the following steps:

[0008] S1. Execute edge caching and dynamic distribution: Deploy cache servers at network edge nodes, monitor students' learning requests in real time, and based on content popularity and network conditions, pre-cache popular teaching resources, such as high-definition course videos, common documents, etc., to edge nodes closer to students. When students initiate requests, directly obtain resources from the edge cache, effectively reducing data transmission delay. At the same time, according to the load conditions of each edge node, dynamically adjust the resource distribution strategy to ensure the stable performance of the overall system;

[0009] S2. Implement adaptive transmission protocol optimization: Automatically select the most suitable transmission protocol based on real-time parameters such as network bandwidth, packet loss rate, and latency. When the network condition is good, use a fast transmission protocol based on UDP to improve the data transmission speed; when the network is unstable, switch to the TCP protocol to ensure data reliability. In addition, optimize the transmission protocol, such as improving the congestion control algorithm, to achieve stable and efficient transmission of teaching data in a complex network environment.

[0010] S3. Carry out P2P collaborative learning resource sharing: Build a learning resource sharing network based on P2P technology. During the learning process, students can not only obtain resources from the server but also directly share learning materials such as notes and homework examples with other students in the same network. In this way, the load pressure on the server is reduced, the efficiency of resource acquisition is improved, and at the same time, learning communication and collaboration among students are promoted.

[0011] S4. Implement cloud-edge collaborative intelligent interaction: Sink some intelligent interaction functions to the edge nodes to achieve cloud-edge collaboration. When students have real-time interactions, such as online questioning and group discussions, the edge nodes perform preliminary processing on interaction data such as voice and text, filter out invalid information, extract key content, and then upload it to the cloud. The cloud is responsible for more complex data analysis and decision-making, such as evaluating the learning status based on students' interaction behaviors and feeding back the results to the edge nodes to achieve a smoother and more personalized interaction experience.

[0012] S5. Build a secure tunnel for encrypted transmission: Establish an end-to-end secure tunnel and use encryption protocols such as SSL / TLS to encrypt the transmission of teaching data to prevent data from being stolen or tampered with during transmission. At the same time, perform integrity verification on the transmitted data to ensure the accuracy and integrity of the data. In addition, update the encryption key regularly to improve the security of the system.

[0013] Preferably, in the edge caching and dynamic distribution step, use machine learning algorithms to predict students' learning needs and pre-cache relevant teaching resources to further improve the resource acquisition speed.

[0014] Preferably, in the adaptive transmission protocol optimization step, monitor the status of multiple network links in real time, and integrate multiple links through link aggregation technology to increase the data transmission bandwidth.

[0015] Preferably, in the P2P collaborative learning resource sharing step, use a reputation mechanism to evaluate students participating in sharing and encourage students to actively share high-quality resources.

[0016] Preferably, in the implementation of the cloud-edge collaborative intelligent interaction step, the edge node preprocesses the historical interaction data of students stored locally to reduce the amount of data uploaded to the cloud and relieve the computing and storage pressure on the cloud server.

[0017] Preferably, in the implementation of the cloud-edge collaborative intelligent interaction step, through the rapid processing of interaction data by the local edge node, the latency of data transmission and processing is reduced, enabling students to obtain near-real-time interaction responses and enhancing the fluency of learning.

[0018] Preferably, in the implementation of the cloud-edge collaborative intelligent interaction step, by combining the big data analysis in the cloud and the local data of the edge node, a comprehensive understanding of the learning characteristics and needs of students is achieved, providing highly personalized interaction services for students and enhancing the learning effect.

[0019] The present invention has the following beneficial effects:

[0020] 1. Through the edge node, real-time preprocessing is carried out on the interaction data such as text, voice, and expressions generated by users. Key information is refined with professional algorithms, the data volume is compressed, invalid information is filtered, and only the necessary data is uploaded to the cloud. In high-concurrency access scenarios, the computing and storage pressure on the cloud is effectively reduced, preventing service anomalies on the cloud due to data overload, ensuring the stable operation of cloud services, and laying a solid foundation for the implementation of large-scale teaching activities.

[0021] 2. By adopting multi-threading technology on the student-side device, various types of interaction data are processed in parallel. Different threads use targeted algorithms to quickly process information such as text, voice, and video. The interaction response time is significantly shortened, and the phenomenon of freezing is reduced. In scenarios such as interactive questions and group discussions, the user operation is smoother, creating an immersive learning atmosphere and enhancing the enthusiasm and participation of students in learning.

[0022] 3. By deploying a variety of sensors at the edge node, user learning behavior data is collected frequently at high frequencies, covering dimensions such as learning duration, answering performance, and interaction participation. Through professional algorithms, real-time data is compared and analyzed with historical data, and appropriate courses are selected from the massive course library to generate a personalized learning plan that fits the learning rhythm and characteristics of students, stimulating students' learning interest, improving learning efficiency, and helping students achieve better learning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of the steps of a method for operating a remote Internet teaching system proposed by the present invention;

[0024] Figure 2 It is a schematic diagram of the resource caching and distribution process in the first embodiment of a method for operating a remote Internet teaching system proposed by the present invention;

[0025] Figure 3It is the analytical diagram of the working process of the system interaction link in the second embodiment of the operation method of a remote Internet teaching system proposed by the present invention;

[0026] Figure 4 It is the display diagram of the user learning behavior analysis process in the third embodiment of the operation method of a remote Internet teaching system proposed by the present invention. Specific implementation manners

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0028] Embodiment 1:

[0029] Step 1, perform edge caching and dynamic distribution

[0030] Build a resource popularity evaluation system based on Spark on the platform, and use the content popularity calculation formula At 3 o'clock every morning, the system analyzes the access logs of various types of course resources in the past 7 days to obtain the access times N access , and synchronize the scoring data R of students for the resources rating . Set the access time weight W recency = 0.6, and the scoring weight W rating = 0.4. According to the evaluation results, classify and store the popular resources on 1500 edge cache servers deployed in 28 key cities across the country. For example, classify the K12 mathematics animation teaching videos by grade and the music courses by style. When a student initiates a resource request, the system uses the GeoIP library to resolve the IP address of the student device into longitude and latitude coordinates (x1, y1), and at the same time obtains the longitude and latitude coordinates (x2, y2) of the edge server, and uses the distance calculation formula to locate the nearest server and shorten the resource loading duration.

[0031] Step 2, implement adaptive transmission protocol optimization

[0032] Deploy a network status monitoring system based on Prometheus on the platform, and collect network bandwidth data every 5 seconds. Calculate the bandwidth according to the formula , where S is the amount of data transmitted within 10 seconds and T is the transmission time. Count the packet loss rate every 30 seconds, and use the formula N lost is the number of lost data packets, and N sent is the number of data packets sent. Calculate the delay through the formula D latency = T receiued - T sent where T receiued is the reception time, and Tsent is the sending time. During the peak learning hours from 7 pm to 9 pm, the packet loss rate in some areas exceeds the threshold. The system automatically switches to the TCP protocol with the help of the Netty framework to ensure the stable playback of live courses. During other periods, the network load is low, and the system uses the UDP protocol to accelerate the data download speed.

[0033] Step 3: Carry out P2P collaborative learning resource sharing

[0034] The platform integrates the P2P sharing function based on the BitTorrent protocol at the user side, encouraging users to carry out collaborative learning in study groups and share materials such as study notes, project documents, and works. The system monitors the server load every 10 minutes through the server load calculation formula where N requests is the number of requests received by the server, and N capacity is the server processing capacity. By introducing P2P technology, the server load pressure is reduced.

[0035] Step 4: Implement cloud-edge collaborative intelligent interaction

[0036] In the live classroom and interaction sessions, the edge node is equipped with an 8-core high-performance processor with a main frequency of 3.2 GHz. It performs real-time preprocessing on user interaction data such as text, voice, and expressions. When the user sends a text message, the open-source natural language processing framework HanLP is called to extract keywords based on the TextRank algorithm, analyze the semantic tendency of the text, and streamline the long text into key information. For voice information, the Baidu Speech Recognition API is used to translate the voice at a sampling rate of 16 kHz and a quantization precision of 16 bits, with an accuracy of over 95%. The OpenCV image recognition library is used to detect the facial expression feature points of the user and identify the expression type. Using the data compression ratio calculation formula through the Huffman coding algorithm, the data upload volume is compressed by 75%, and only the key information is uploaded to the cloud. In high-concurrency scenarios, the edge node processes the interaction data at high speed, reducing the computing and storage pressure on the cloud and ensuring the stable operation of the cloud service to support the large-scale teaching activities of the platform.

[0037] Step 5: Build a secure tunnel for encrypted transmission

[0038] The platform uses the SSL / TLS encryption protocol to encrypt the transmission of data such as user personal information and learning records. Using the encryption strength calculation formula the data is encrypted through the SHA-256 hash algorithm to ensure data security. The encryption key is updated every 30 days to prevent the data from being cracked.

[0039] It should be noted that in the first embodiment, by constructing an execution edge caching and dynamic distribution system, with the help of content popularity evaluation and distance positioning algorithms, rapid resource distribution is achieved, and edge nodes are used to preprocess data to relieve the pressure on the cloud. In Comparative Example 1, the traditional centralized caching and distribution method is adopted, and all resources are obtained from the central server. In Comparative Example 2, a caching and distribution mechanism is provided, but there is a lack of a real-time network monitoring mechanism. The specific test data is shown in Table 1.

[0040] Table 1: Comparison Table of Resource Caching and Distribution Efficiency

[0041]

[0042]

[0043] Specifically, the first embodiment constructs a complete resource popularity evaluation system. With the help of big data analysis technology, the access and scoring data of course resources are continuously tracked and analyzed. Based on the analysis results, edge caching servers are deployed in multiple key cities across the country to achieve the near storage and rapid distribution of popular resources. This strategy enables the resource loading success rate to reach 99.2%, and the average server response time is shortened to 48.7 ms.

[0044] Comparative Example 1 adopts the traditional centralized caching and distribution mode, and all resource requests from users need to be processed by the central server. During large-scale user concurrent access, the processing and transmission capabilities of the central server are extremely easy to saturate, resulting in a resource loading success rate of only 90.5% and an average server response time as long as 203.4 ms. Compared with Comparative Example 1, the resource loading success rate of the present invention is increased by 8.7%. This means that in 1000 resource requests, the present invention can successfully respond 87 more times, greatly reducing the time for users to wait for resource loading. The average server response time is shortened by 154.7 ms, significantly improving the timeliness for users to obtain resources and providing a strong guarantee for the smooth progress of teaching activities.

[0045] Although Comparative Example 2 sets up a caching and distribution mechanism, due to the lack of real-time network monitoring, the transmission strategy cannot be flexibly adjusted according to the network status. During network congestion periods, resource transmission is blocked, resulting in a resource loading success rate dropping to 85.3% and an average server response time of 201.9 ms. Compared with Comparative Example 2, the resource loading success rate of the present invention is increased by 13.9%, and the average server response time is shortened by 153.2 ms, effectively avoiding resource acquisition failures and long waiting times caused by network problems.

[0046] In terms of cloud data processing, the present invention deploys a data preprocessing mechanism at the edge node. In live classrooms and interactive sessions, the user's text, voice, expression and other interactive data are intelligently analyzed and refined, and an efficient data compression algorithm is used to reduce the amount of data uploaded to the cloud by 75%, so that the daily cloud data processing volume is only 10.2TB. In comparison example 1, since all resources are obtained from the central server, a large amount of unprocessed raw data pours into the cloud, and the daily data processing volume reaches 40.8TB. In comparison example 2, due to poor network transmission strategy, the daily data processing volume in the cloud is 40.6TB. Compared with comparison example 1, the present invention reduces the daily data processing volume in the cloud by 30.6TB, and reduces it by 30.4TB compared to comparison example 2. This not only reduces the hardware cost and operation and maintenance pressure of the cloud server, but also improves the stability and reliability of cloud services, ensuring that the platform can continue to provide users with high-quality teaching services.

[0047] Embodiment 2:

[0048] Step 1: Implement edge caching and dynamic distribution

[0049] The platform builds a resource heat evaluation system based on Spark, using the content heat calculation formula At 3 a.m. every day, the system analyzes the access logs of various types of course resources in the past 7 days and obtains the number of accesses N. access , and synchronize students' rating data on resources rating . Set the access time weight W recency =0.6, scoring weight W rating =0.4. Based on the evaluation results, popular resources are classified and stored on 1,500 edge cache servers deployed in 28 key cities across the country. For example, K12 math animation teaching videos are classified by grade, and music courses are classified by style. When a student initiates a resource request, the system resolves the IP address of the student's device into longitude and latitude coordinates (x1, y1) through the GeoIP library, and obtains the longitude and latitude coordinates of the edge server (x2, y2), and uses the distance calculation formula Locate the nearest server to shorten resource loading time.

[0050] Step 2: Implement adaptive transmission protocol optimization

[0051] The platform deploys a Prometheus-based network status monitoring system, which collects network bandwidth data every 5 seconds. Calculate the bandwidth, where S is the amount of data transmitted in 10 seconds and T is the transmission time. Count the packet loss rate every 30 seconds and use the formula N lost is the number of lost packets, N sent is the number of packets sent. latency =Treceiued -T sent Computing latency, where T receiued is the reception time, and T sent is the transmission time. During the peak learning hours from 7 pm to 9 pm, the packet loss rate in some areas exceeds the threshold. The system automatically switches to the TCP protocol with the help of the Netty framework to ensure the stable playback of live courses. During other periods, the network load is low, and the system uses the UDP protocol to accelerate the data download speed.

[0052] Step 3: Carry out P2P collaborative learning resource sharing

[0053] The platform integrates the P2P sharing function based on the BitTorrent protocol at the user end, encouraging users to carry out collaborative learning in study groups and share materials such as study notes, project documents, and works. The system monitors the server load every 10 minutes through the server load calculation formula where N requests is the number of requests received by the server, and N capacity is the server processing capacity. By introducing P2P technology, the server load pressure is reduced.

[0054] Step 4: Implement cloud-edge collaborative intelligent interaction

[0055] The edge device at the student end is equipped with a dedicated interactive processing chip of model NVIDIA A30, which has 3,840 CUDA cores. Using multithreading technology, data processing tasks such as text, voice, and video are allocated to different threads for parallel processing. When the user inputs text in the interactive question scenario, the text processing thread uses the Aho-Corasick algorithm to quickly match common questions in the knowledge base to achieve a quick response. In the group discussion scenario, the voice processing thread uses VAD (Voice Activity Detection) technology to accurately identify the start and end times of the voice, reducing ineffective voice processing. The video processing thread optimizes video data using the YUV420 format. Through the interactive response time calculation formula T response = T processed - T initiated , the interactive response time is significantly shortened. In scenarios such as interactive questions and group discussions, user interaction operations are processed at high speed, improving the user operation fluency and creating an immersive learning experience.

[0056] Step 5: Build a secure tunnel for encrypted transmission

[0057] The platform uses the SSL / TLS encryption protocol to encrypt the transmission of data such as user personal information and learning records. Using the encryption strength calculation formula the data is encrypted through the SHA-256 hash algorithm to ensure data security. The encryption key is updated every 30 days to prevent data from being cracked.

[0058] It should be noted that in the second embodiment, multi-threading technology and data optimization algorithms are used to improve the accuracy of speech recognition, reduce the occupancy of video transmission bandwidth, and reduce the lag of the user interface. In Comparative Example 1, single-thread processing is used to handle interactive data, and data such as text, speech, and video are processed sequentially. In Comparative Example 2, the video data transmitted is not format-optimized, as shown in Table 2 specifically.

[0059] Table 2: Comparison Table of Performance Optimization in the Interaction Link

[0060]

[0061] Specifically, in the interaction process of online teaching, the accuracy of speech recognition, the occupancy of video transmission bandwidth, and the number of lags of the user interface are the key factors affecting the user experience. In the edge device of the student side, the present invention adopts multi-threading technology to allocate data processing tasks such as text, speech, and video to different threads for synchronous processing, greatly improving the data processing efficiency. In terms of speech recognition, professional speech recognition tools and voice activity detection technology are used to accurately translate and recognize speech, and the accuracy of speech recognition reaches 95.3%. In video processing, the video data is converted into a specific high-definition format to reduce the data volume, and the occupancy of video transmission bandwidth is only 2.1 Mbps. Through multi-thread parallel processing and data optimization, the number of lags of the user interface is only 1.2 times per hour.

[0062] In Comparative Example 1, single-thread processing is used to handle interactive data, and data such as text, speech, and video need to be processed sequentially, resulting in low processing efficiency. The speech recognition process is easily interfered by other tasks, resulting in the accuracy of speech recognition dropping to 80.2%. Delayed data processing also causes video transmission delay, and the occupancy of video transmission bandwidth reaches 5.2 Mbps, which in turn leads to 9.8 lags of the user interface per hour. Compared with Comparative Example 1, the accuracy of speech recognition of the present invention is increased by 15.1%, which enables more accurate understanding of the user's intention and reduces communication errors during the speech interaction process. The occupancy of video transmission bandwidth is reduced by 3.1 Mbps, reducing the network transmission pressure and ensuring smooth video playback. The number of lags of the user interface is reduced by 8.6 times per hour, significantly improving the fluency of user operations and the interaction experience.

[0063] In Comparative Example 2, since the video data format was not optimized, the volume of the transmitted video data was large, which not only occupied more bandwidth (the video transmission bandwidth occupancy was 5.3 Mbps), but also increased the difficulty of device decoding and processing, resulting in video transmission delays, affecting the synchronization of speech recognition, and making the speech recognition accuracy 80.1%, with the user interface freezing 14.6 times per hour. Compared with Comparative Example 2, the present invention has a 15.2% increase in speech recognition accuracy, a 3.2 Mbps reduction in video transmission bandwidth occupancy, and a 13.4 - time reduction in the number of freezes of the user interface per hour. Through multi - aspect optimizations, the present invention creates a smooth and efficient interaction environment for users, promoting interaction and communication in the teaching process.

[0064] Example 3:

[0065] Step 1: Implement edge caching and dynamic distribution

[0066] Build a resource popularity evaluation system based on Spark on the platform, and use the content popularity calculation formula Every day at 3:00 am, the system analyzes the access logs of various types of course resources in the past 7 days to obtain the number of accesses N access , and synchronizes the scoring data R of students for the resources rating . Set the access time weight W recency = 0.6, and the scoring weight W rating = 0.4. According to the evaluation results, classify and store popular resources on 1500 edge caching servers deployed in 28 key cities across the country. For example, classify K12 math animation teaching videos by grade and music courses by style. When a student initiates a resource request, the system uses the GeoIP library to resolve the IP address of the student's device into longitude and latitude coordinates (x1, y1), and at the same time obtains the longitude and latitude coordinates (x2, y2) of the edge server, and uses the distance calculation formula to locate the nearest server and shorten the resource loading time.

[0067] Step 2: Implement adaptive transmission protocol optimization

[0068] Deploy a network status monitoring system based on Prometheus on the platform, and collect network bandwidth data every 5 seconds. Calculate the bandwidth according to the formula , where S is the amount of data transmitted within 10 seconds and T is the transmission time. Statistically calculate the packet loss rate every 30 seconds, and use the formula N lost is the number of lost data packets, and N sent is the number of data packets sent. Calculate the delay through the formula D latency = T receiued - T sent where T receiued is the reception time, and T sentis the sending time. During the peak learning hours from 7 pm to 9 pm, the packet loss rate in some areas exceeds the threshold. The system automatically switches to the TCP protocol with the help of the Netty framework to ensure the stable playback of live courses. During other periods, the network load is low, and the system uses the UDP protocol to accelerate the data download speed.

[0069] Step 3: Carry out P2P collaborative learning resource sharing

[0070] The platform integrates the P2P sharing function based on the BitTorrent protocol at the user end, encourages users to carry out collaborative learning in study groups, and share materials such as study notes, project documents, and works. The system calculates the server load through the formula Monitors the server load every 10 minutes, where N requests is the number of requests received by the server, and N capacity is the server processing capacity. Introduce P2P technology to reduce the server load pressure.

[0071] Step 4: Implement cloud-edge collaborative intelligent interaction

[0072] The edge node is configured with 6 types of data collection sensors, including cameras, microphones, gravity sensors, etc. Collects user learning behavior data at a frequency of 10 Hz, covering learning duration, answering questions, interaction participation, etc. Stores the collected behavior data in the local cache according to the time series. Uses the learning behavior similarity calculation formula:

[0073]

[0074] Adopts the cosine similarity algorithm to compare the user's real-time learning behavior data with the historical data of the past 30 days. Through the comparison results, screen out suitable courses for the user from 100,000 courses in the platform course library, generate a personalized learning plan, and improve the user's learning interest and effect.

[0075] Step 5: Build a secure tunnel for encrypted transmission

[0076] The platform uses the SSL / TLS encryption protocol to encrypt and transmit data such as user personal information and learning records. Uses the encryption strength calculation formula Encrypts the data through the SHA-256 hash algorithm to ensure data security. Updates the encryption key every 30 days to prevent the data from being cracked.

[0077] It should be noted that in Example 3, by collecting multi-dimensional learning behavior data, analyzing with the cosine similarity algorithm, generating a personalized learning plan, the user retention rate is improved, the learning duration is increased, and the learning content reuse rate is increased. In Comparative Example 1, learning behavior data is not collected, and in Comparative Example 2, simple matching rules are used to analyze the learning behavior data, as shown in Table 3 specifically.

[0078] Table 3: Comparison Table of the Effects Driven by User Learning Behaviors

[0079] Comparison Items Example 3 Comparative Example 1 Comparative Example 2 Weekly User Retention Rate 78.6％ 39.4％ 59.3％ Average Daily Learning Duration of Users 102.3 minutes 41.7 minutes 63.2 minutes Learning Content Reuse Rate 62.5％ 21.3％ 32.8％

[0080] Specifically, in online teaching, understanding user learning behaviors and providing personalized learning plans for users are the keys to improving user retention rates, increasing learning durations, and enhancing the reuse rate of learning content. In this invention, a variety of data collection sensors are configured at edge nodes to collect user learning behavior data at a frequency of 10 Hz, covering multiple dimensions such as learning duration, answering question situations, and interaction participation. Using the cosine similarity algorithm, the real-time user learning behavior data is deeply compared with the historical data of the past 30 days, and the appropriate courses are screened from the large course library of the platform to generate personalized learning plans for users. After implementation, the weekly user retention rate reaches 78.6%, the average daily learning duration is 102.3 minutes, and the reuse rate of learning content reaches 62.5%.

[0081] In Comparative Example 1, no learning behavior data is collected, and it is impossible to recommend learning content based on user characteristics. The learning plan lacks pertinence and is difficult to meet the personalized needs of users, resulting in a weekly user retention rate of only 39.4%, an average daily learning duration of 41.7 minutes, and a reuse rate of learning content of 21.3%. Compared with Comparative Example 1, the weekly user retention rate of this invention is increased by 39.2%, which means that among 100 users, this invention can retain 39 more users, enhancing the stickiness of users to the platform. The average daily learning duration increases by 60.6 minutes, indicating that users are more willing to spend time learning on this platform. The reuse rate of learning content is increased by 41.2%, improving the utilization efficiency of learning resources and reducing the operating costs of the platform.

[0082] In Comparative Example 2, although learning behavior data is collected, simple matching rules are used for analysis, and the generated learning plan has poor accuracy and cannot well fit the user's learning path. The weekly user retention rate is 59.3%, the average daily learning duration is 63.2 minutes, and the reuse rate of learning content is 32.8%. Compared with Comparative Example 2, the weekly user retention rate of this invention is increased by 19.3%, the average daily learning duration is increased by 39.1 minutes, and the reuse rate of learning content is increased by 29.7%. Through multi-dimensional data collection and precise data analysis, this invention realizes the generation of personalized learning plans, greatly improves the effectiveness of user learning behavior drive, provides users with a better-quality and more efficient learning experience, and promotes the improvement of learning effects.

[0083] Furthermore, as Figure 2As shown, in the first embodiment, the system automatically starts a data collection task to collect the access logs of course resources in more than 20 fields such as painting, music, and K12 education in the past 7 days, obtain the number of accesses, and synchronously collect the scoring data of students on the course resources. Using the content popularity evaluation algorithm, a weight of 0.6 is set for the access time and a weight of 0.4 is set for the score to accurately evaluate the resource popularity. According to the evaluation results, the popular resources are classified by course type and stored on 1500 edge cache servers deployed in 28 key cities across the country. For example, the K12 math animation teaching videos are stored by grade, and the music courses are stored by style.

[0084] When a student initiates a resource request, the system calls the GeoIP library to resolve the IP address of the student device into longitude and latitude coordinates. At the same time, it obtains the longitude and latitude coordinates of the edge server and locates the server closest to the student through the distance calculation formula. After testing, this measure reduces the resource loading speed from 2.5 seconds to 0.4 seconds, greatly improving the resource acquisition efficiency.

[0085] In live classrooms and interactive sessions, the edge nodes collect the interactive data sent by students in real time, such as text, voice, and expressions, at a millisecond-level response speed. When processing text messages, the open-source natural language processing framework HanLP is called to extract keywords based on the TextRank algorithm, analyze the semantic tendency of the text, and streamline the long text into key information. For voice information, the Baidu Speech Recognition API is used to translate the voice at a sampling rate of 16kHz and a quantization precision of 16 bits, with an identification accuracy rate of over 95%. The OpenCV image recognition library is used to detect the facial expression feature points of students and identify the expression types. With the Huffman coding compression algorithm, the data upload volume is reduced by 75%. In high-concurrency scenarios, the edge nodes can process 1000 interactive data per second, greatly reducing the cloud computing and storage pressure and ensuring that the platform can stably support large-scale teaching activities with a daily course access volume of 500,000 times.

[0086] Furthermore, as Figure 3 shown, when a student inputs interactive data such as voice, text, and video in scenarios such as interactive questions and group discussions, the NVIDIA A30 dedicated interactive processing chip installed on the edge device of the student side immediately starts a multi-threaded mechanism and distributes different types of data processing tasks to 3840 CUDA cores for parallel processing.

[0087] In the speech processing thread, the VAD (Voice Activity Detection) technology is adopted to accurately identify the start and end times of speech, reducing the processing of invalid speech. In the interactive question scenario, the text processing thread uses the Aho-Corasick algorithm to quickly match common questions in the knowledge base, achieving fast response. The video processing thread converts the video data into the YUV420 format, reducing the data volume by adjusting the color space and sampling method. After each thread finishes processing, the system integrates the processing results and feeds them back to the student interaction interface. After testing, the system can process 500 user interaction operations per second, reducing the interaction response time from 0.6 seconds to 0.15 seconds, greatly improving the operation fluency, creating an immersive learning environment for students, and enabling students to interact efficiently with teachers and classmates in scenarios such as group discussions and live Q&A sessions.

[0088] Furthermore, as Figure 4 shown, the edge node is configured with six types of data collection sensors such as cameras, microphones, and gravity sensors, continuously collecting students' learning behavior data at a frequency of 10Hz. The data covers multiple dimensions such as students' daily learning duration, the correctness of each answer, and their interaction participation in live classes. The collected data is stored in the local cache in time series.

[0089] The system regularly extracts students' real-time learning behavior data every week and uses the cosine similarity algorithm to compare and analyze it with the historical data of the past 30 days. Through the comparison results, 5 - 10 suitable courses are selected from the platform's 100,000-course library. For example, if the system detects that a certain student has shown strong interest in the landscape photography direction during the recent study of a photography course and has a high quality of assignment completion, the system will select relevant advanced landscape photography courses from the course library and recommend them to the student. After implementing this solution, through investigation, students' learning interest has been significantly improved, the reuse rate of learning content has been greatly increased, effectively enhancing students' learning enthusiasm and learning effect, and strengthening students' willingness to use the platform in the long term.

[0090] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A method for operating a remote Internet teaching system, characterized in that, It includes the following steps: S1. Execute edge caching and dynamic distribution: Deploy cache servers at network edge nodes, monitor students' learning requests in real time, and based on content popularity and network conditions, pre-cache popular teaching resources such as high-definition course videos and common documents to edge nodes closer to students. When students initiate requests, they can directly obtain resources from the edge cache, effectively reducing data transmission latency. At the same time, according to the load conditions of each edge node, dynamically adjust the resource distribution strategy to ensure the stable performance of the overall system; S2. Implement adaptive transmission protocol optimization: Automatically select the most suitable transmission protocol based on real-time parameters such as network bandwidth, packet loss rate, and latency. When the network condition is good, adopt a fast transmission protocol based on UDP to improve data transmission speed; when the network is unstable, switch to the TCP protocol to ensure data reliability. In addition, optimize the transmission protocol, such as improving the congestion control algorithm, to achieve stable and efficient transmission of teaching data in complex network environments; S3. Carry out P2P collaborative learning resource sharing: Build a learning resource sharing network based on P2P technology. During the learning process, students can not only obtain resources from the server but also directly share learning materials such as notes and homework examples with other students in the same network. In this way, reduce the load pressure on the server, improve the efficiency of resource acquisition, and promote learning communication and collaboration among students; S4. Implement cloud-edge collaborative intelligent interaction: Sink some intelligent interaction functions to edge nodes to achieve cloud-edge collaboration. When students have real-time interactions such as online questions and group discussions, edge nodes perform preliminary processing on interaction data such as voice and text, filter out invalid information, extract key content and upload it to the cloud. The cloud is responsible for more complex data analysis and decision-making, such as evaluating the learning status based on students' interaction behaviors and feedbacking the results to edge nodes to achieve a smoother and more personalized interaction experience; S5. Build a secure tunnel for encrypted transmission: Establish an end-to-end secure tunnel, use encryption protocols such as SSL / TLS to encrypt the transmission of teaching data, prevent data from being stolen or tampered with during transmission. At the same time, perform integrity verification on the transmitted data to ensure the accuracy and integrity of the data. In addition, update the encryption key regularly to improve the security of the system.

2. A method for operating a remote Internet teaching system according to claim 1, characterized in that, In the step of executing edge caching and dynamic distribution, use machine learning algorithms to predict students' learning needs and pre-cache relevant teaching resources to further improve the speed of resource acquisition.

3. A method for operating a remote Internet teaching system according to claim 1, characterized in that, In the step of implementing adaptive transmission protocol optimization, monitor the status of multiple network links in real time, and through link aggregation technology, integrate multiple links to increase the data transmission bandwidth.

4. A method for operating a remote Internet teaching system according to claim 1, characterized in that In the step of carrying out P2P collaborative learning resource sharing, adopt a reputation mechanism to evaluate students participating in sharing and encourage students to actively share high-quality resources.

5. A method for operating a remote Internet teaching system according to claim 1, characterized in that, In the step of implementing cloud-edge collaborative intelligent interaction, edge nodes use the historical interaction data of students stored locally for preprocessing, reduce the amount of data uploaded to the cloud, and reduce the computing and storage pressure on the cloud server.

6. A method for operating a remote Internet teaching system according to claim 1, characterized in that, In the process of implementing cloud-edge collaborative intelligent interaction, by quickly processing interaction data through local edge nodes, the latency of data transmission and processing is reduced, enabling students to obtain near-real-time interaction responses and enhancing the fluency of learning.

7. A method for operating a remote Internet teaching system according to claim 1, characterized in that, In the process of implementing cloud-edge collaborative intelligent interaction, by combining big data analysis in the cloud and local data at edge nodes, a comprehensive understanding of students' learning characteristics and needs can be obtained, providing highly personalized interaction services for students and enhancing learning effects.

Images