Smart campus converged communication system and information interaction method thereof

Abstract

The invention discloses a smart campus converged communication system and an information interaction method thereof, and relates to the technical field of campus communication. The system comprises a terminal access module, a protocol conversion module, a hierarchical transmission module and a cross-modal interaction module. The terminal access module acquires a terminal data stream in real time by using a multi-protocol adapter; the protocol conversion module performs protocol decoupling and standardized packaging on the data flow by using a QoS bandwidth prediction model, and generates a bandwidth allocation scheme according to a dynamic bandwidth allocation algorithm in combination with a network topology state matrix; the hierarchical transmission module extracts service information data by using a data semantic analysis model, calculates Shannon entropy and constructs a transmission queue mapping table; and the cross-modal interaction module generates an interaction protocol stack adaptive to the service by using a multi-modal feature collaborative extraction mechanism and an attention weight fusion mechanism based on the transmission queue mapping table.

Description

Technical Field

[0001] The present invention belongs to the technical field of campus communication, and particularly relates to a smart campus integrated communication system and an information interaction method thereof. Background Art

[0002] With the in-depth construction of smart campuses, traditional campus networks face problems such as incompatible access protocols for multi-source heterogeneous terminals, low transmission efficiency of teaching data streams, and rigid scheduling of cross-modal teaching resources. Existing systems generally have three deficiencies: data islands are caused by differences in communication protocols among Internet of Things devices, mobile terminals, and teaching facilities; there is a lack of a dynamic bandwidth allocation mechanism based on service characteristics during network congestion; the allocation of teaching resources and course arrangements rely on manual experience and it is difficult to achieve spatio-temporal dynamic optimization. Especially in multi-modal teaching scenarios that integrate voice, video, and text, traditional solutions cannot meet new requirements such as real-time interaction quality assurance, intelligent scheduling of teaching resources, and precise push of personalized services. Summary of the Invention

[0003] Based on this, it is necessary to provide a smart campus integrated communication system and an information interaction method thereof that can solve the above technical problems.

[0004] In a first aspect, the present application provides a smart campus integrated communication system, including:

[0005] A terminal access module, a protocol conversion module, a hierarchical transmission module, and a cross-modal interaction module:

[0006] The terminal access module is used for:

[0007] Real-time acquisition of the data stream of the access terminal through a multi-protocol adapter;

[0008] The protocol conversion module is used for:

[0009] Adopting a QoS-based bandwidth prediction model to perform protocol decoupling and standard encapsulation on the data stream, and generating a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix;

[0010] The hierarchical transmission module is used for:

[0011] Extracting the service information data contained in the decoupled data stream through a data semantic parsing model;

[0012] Calculating the Shannon entropy of the service information data, and constructing a transmission queue mapping table according to the Shannon entropy and the bandwidth allocation scheme; the transmission queue mapping table is used for dynamically allocating transmission queues according to the priority attributes of the data stream;

[0013] The cross-modal interaction module is used for:

[0014] Based on the transmission queue mapping table, use the multi-modal feature collaborative extraction mechanism to extract the feature vectors of business information data, and generate an interaction protocol stack adapted to the business through the attention weight fusion mechanism.

[0015] In one embodiment, the hierarchical transmission module is further configured to:

[0016] Calculate the Shannon entropy value using the following formula:

[0017]

[0018] where p i (t) represents the traffic proportion of the i-th type of data stream within the window time t, and w i (t) represents the dynamic weight factor determined by data priority and real-time network status; Priority i represents the fixed priority, and Delay i (t) represents the current average delay of the i-th type of data stream.

[0019] In one embodiment, the cross-modal interaction module includes a course coordination unit, which is used to:

[0020] Obtain student progress data, interest preference data, and teacher available time data in real time;

[0021] Calculate the course demand value of the student according to the student progress data and interest preference data, and in combination with the time window and subject type;

[0022] Based on the teacher available time data, calculate the teacher spatio-temporal matching degree tensor and generate a teacher time matching matrix;

[0023] Calculate the matching degree between the course demand value and the teacher time matching matrix through a matching degree algorithm;

[0024] If the matching degree is higher than the preset threshold, generate a course arrangement plan;

[0025] If the matching degree does not meet the threshold, adjust the time window and recalculate the matching degree;

[0026] According to the course arrangement plan, generate course schedule data, store it in the course database, and feedback it to the student and teacher terminals at the same time.

[0027] In one embodiment, the course coordination unit is further configured to:

[0028] Calculate the course demand value of the student using the following formula:

[0029]

[0030] Among them, P(t) represents the progress completion rate at time t, P0 represents the midpoint of the Sigmoid function, and k p represents the progress sensitivity coefficient, where T avg represents the average learning duration, α(t) represents the time-varying weight, and I i (t) represents the multi-dimensional interest vector, and S k represents the subject type tensor, and C represents the teaching resource feature matrix.

[0031] In one embodiment, the course coordination unit is further configured to:

[0032] Calculate the teacher spatio-temporal matching degree tensor using the following formula:

[0033]

[0034] where u i and σ i represent the Gaussian distribution parameters of the teaching time preference of teacher i, Entropy(T k ) represents the global conflict entropy of the time window K, represents the recursive penalty factor, i represents the teacher, j represents the time window, and k represents the subject type.

[0035] In one embodiment, the cross-modal interaction module further includes a learning resource allocation unit, which is configured to:

[0036] Based on the course arrangement plan, use a resource prediction model to predict the total amount of teaching resources required in different time periods, and generate a resource prediction result;

[0037] According to the resource prediction result, divide the teaching resources into two parts: dynamic allocation and fixed allocation; the dynamically allocated part is used to handle course adjustments, and the fixed allocation part is used to ensure the stability of basic resources;

[0038] Obtain the resource allocation information in real time, and detect resource overload or idleness;

[0039] If resource overload or idleness is detected, use a resource reallocation mechanism to reallocate the resources in the dynamically allocated part;

[0040] Match the reallocated teaching resources with the course arrangement plan through a resource scheduling algorithm to generate a resource allocation table.

[0041] In one embodiment, the cross-modal interaction module further includes a personalized push unit, which is configured to:

[0042] Obtain the user's historical behavior and feature data to form a user preference database;

[0043] Generate a push content set based on the user preference database and using preset matching rules;

[0044] Based on the push content set, send the push content to the user terminal using the distribution channels.

[0045] In one embodiment, the personalized push unit is further configured to:

[0046] Calculate the matching degree between the user and the push content based on the push content set through a data model;

[0047] If the matching degree is higher than the preset threshold, include the content in the push queue;

[0048] Select the push content with the highest priority from the push queue according to the delivery strategy;

[0049] Use the distribution channels to send the push content with the highest priority to the user terminal;

[0050] Record the user click and feedback data, and update the user preference database and the preset matching rules.

[0051] In one embodiment, the cross-modal interaction module further includes a teaching achievement analysis unit for:

[0052] Obtain teaching interaction data, extract the interaction volume and eigenvalue of the teaching interaction data, and construct an interaction feature set;

[0053] Based on the interaction feature set, use the random forest algorithm to train and generate an analysis model;

[0054] Based on the teaching interaction data, output an analysis value using the analysis model;

[0055] Generate a teaching interaction trend analysis result according to the output analysis value.

[0056] In a second aspect, the present application further provides a smart campus integrated communication information interaction method, including:

[0057] The method includes:

[0058] Real-time obtain the data stream of the access terminal through a multi-protocol adapter;

[0059] Adopt a QoS-based bandwidth prediction model to perform protocol decoupling and standard encapsulation on the data stream, and generate a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix;

[0060] Extract the service information data included in the decoupled data stream through a data semantic parsing model;

[0061] Calculate the Shannon entropy of the service information data, and construct a transmission queue mapping table according to the Shannon entropy and the bandwidth allocation scheme; the transmission queue mapping table is used to dynamically allocate transmission queues according to the priority attributes of data streams;

[0062] Based on the transmission queue mapping table, use the multi-modal feature collaborative extraction mechanism to extract the feature vectors of the service information data, and generate an interaction protocol stack adapted to the service through the attention weight fusion mechanism.

[0063] In a third aspect, the present application also provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned intelligent campus fusion communication information interaction method are implemented.

[0064] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned intelligent campus fusion communication information interaction method are implemented.

[0065] The above-mentioned intelligent campus fusion communication system and its information interaction method realize the real-time acquisition of multi-source heterogeneous terminal data streams through a multi-protocol adapter, and solve the data island problem caused by communication protocol differences in traditional campus networks; adopt technologies such as a QoS-based bandwidth prediction model to achieve dynamic bandwidth allocation based on service characteristics and avoid low transmission efficiency during network congestion; construct a transmission queue mapping table through a data semantic parsing model, etc., and use a multi-modal feature collaborative extraction mechanism and an attention weight fusion mechanism to realize the intelligent scheduling of cross-modal teaching resources and the generation of an interaction protocol stack adapted to the service, which can meet new requirements such as real-time interaction quality guarantee, intelligent scheduling of teaching resources, and precise push of personalized services in multi-modal teaching scenarios integrating voice, video, and text, improve the overall performance and functions of the intelligent campus fusion communication system, and promote the further development of the construction of the intelligent campus. Description of the Drawings

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0067] Figure 1 It is a schematic structural diagram of an intelligent campus fusion communication system of the present invention;

[0068] Figure 2 It is a flowchart of an intelligent campus fusion communication information interaction method of the present invention. Detailed Embodiments

[0069] To make the objectives, technical solutions, and advantages of this application clearer and more understandable, the following further details this application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely used to explain this application and are not used to limit this application.

[0070] A smart campus integrated communication system of this application is composed of terminal devices, Internet of Things gateways, network switches, routers, servers, etc. When multi-source heterogeneous terminals access, the terminal devices send data according to their respective protocols. The Internet of Things gateway collects and preprocesses the data and then transmits it to the server. After operations such as protocol conversion, the data can be normally transmitted. When a large amount of teaching data is transmitted, the terminal devices transmit the data to the network switch and then to the server. The server analyzes the data to construct a transmission queue mapping table, allocates transmission queues according to priorities, and transmits the data in sequence through the network switch and router. When scheduling cross-modal teaching resources, the terminal devices send multi-modal data to the server. The server extracts feature vectors, etc., to generate an interactive protocol stack, coordinates the resource server to obtain corresponding teaching resources from the resource library, and pushes them to the terminal devices through network devices.

[0071] In one embodiment, as Figure 1 shown, a smart campus integrated communication system is provided. In this embodiment, this system is exemplified by being deployed on a terminal. It can be understood that this system can also be independently deployed on the server side, or can be deployed in the entire network architecture including terminals and servers, and its functions are realized through collaborative interactions between terminals and servers. In this embodiment, this system includes:

[0072] A terminal access module 101, a protocol conversion module 102, a hierarchical transmission module 103, and a cross-modal interaction module 104:

[0073] The terminal access module 101 is used for:

[0074] Obtaining the data stream of the access terminal in real time through a multi-protocol adapter.

[0075] Among them, the multi - protocol adapter supports docking with communication protocols of various heterogeneous terminal devices, such as TCP / IP, HTTP, MQTT, CoAP, WebSocket and other communication protocols, solves the problem of data intercommunication caused by protocol differences among different terminals, and can use an efficient polling mechanism such as epoll / kqueue or an event - driven model such as the asynchronous I / O of Node.js to monitor the network ports of terminal devices in real - time, ensuring that the processing process is triggered immediately after the data packet arrives. It adopts a streaming data processing framework to segment - parse and dynamically buffer the continuously incoming data stream, avoiding data loss or delay caused by burst traffic. It can shield the underlying protocol differences, provide a unified data entry for the upper layer, and use real - time stream processing technology to meet the high - concurrency and low - latency data interaction requirements in the smart campus scenario.

[0076] The protocol conversion module 102 is used for:

[0077] Adopt a QoS - based bandwidth prediction model to decouple the protocol and standardize the encapsulation of the data stream, and generate a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix.

[0078] Among them, the QoS (Quality of Service) - based bandwidth prediction model can define the QoS level according to service priorities such as video monitoring > environmental monitoring > log upload, fuse data stream characteristics such as data volume and packet frequency, terminal characteristics such as device type and historical bandwidth occupancy, and environmental characteristics such as time period and geographical location to construct a multi - dimensional feature vector, and combine with an LSTM neural network or a Transformer model to obtain a QoS - based bandwidth prediction model. It can identify the protocol type through technologies such as regular matching and packet - header parsing, split the mixed - protocol data stream into independent data units such as sensor values, control instructions, and video - stream segments according to business logic, and eliminate the protocol - layer coupling. Convert the decoupled data into a unified format such as JSON - LD or Protobuf, attach metadata such as device ID, timestamp, and QoS level to form a general data object within the system. Collect switch - port traffic, BGP routing - table updates, and SDN controller status through the SNMP protocol, construct a network topology state matrix containing node load, link - bandwidth utilization, and latency, and generate a policy table containing terminal ID, priority, allocated bandwidth, and transmission path, and inject it to obtain a transmission - queue mapping table.

[0079] The hierarchical transmission module 103 is used for:

[0080] Extract the service - information data contained in the decoupled data stream through a data semantic parsing model.

[0081] Among them, the standardized data stream output by the protocol conversion module is deeply parsed, and different data types adopt different parsing strategies. For example, for structured data such as sensor values, keywords are extracted through regular expressions or JSON Schema, such as temperature: 25°C; for unstructured data such as videos / images, pre-trained models such as YOLOv8 and ResNet are used for object detection or content classification. It can be based on an NLP model to parse the semantic intention of text instructions, such as turning on the laboratory lighting, and identifying the semantic types of data, such as: sensor data: environmental monitoring values such as temperature, humidity, PM2.5, etc.; video stream: real-time camera footage or video clips; control instructions: operation commands such as device on / off and parameter adjustment; emergency notifications such as smoke alarms and intrusion detections.

[0082] Calculate the Shannon entropy of the service information data, and construct a transmission queue mapping table according to the Shannon entropy and the bandwidth allocation scheme; the transmission queue mapping table is used to dynamically allocate transmission queues according to the priority attributes of the data stream.

[0083] Among them, calculating the Shannon entropy of the service information data can be to perform statistical analysis on the standardized data stream output by the protocol conversion module, extract the occurrence frequency characteristics of types such as sensor values, video frames, and control instructions, count the proportion of different data packet lengths, calculate the difference degree between adjacent data packets such as the mutation frequency of sensor values, and obtain the Shannon entropy. Preset the initial priority according to the service type, such as high priority for video streams, medium priority for environmental monitoring, and low priority for control instructions. If the entropy value of a certain data stream is significantly higher than the historical average, such as a video stream suddenly having a high entropy, then temporarily increase its priority; otherwise, decrease its priority. In this way, the dynamic priority is obtained, and the transmission queue mapping table is generated by combining the bandwidth allocation and the dynamic priority.

[0084] The interaction module 104 is used for:

[0085] Based on the transmission queue mapping table, use the multi-modal feature collaborative extraction mechanism to extract the feature vectors of the service information data, and generate an interaction protocol stack adapted to the service through the attention weight fusion mechanism.

[0086] Among them, based on semantic tags in the transmission queue mapping table such as video stream and control instruction, inherent attributes related to the service type are extracted. For example, for video stream: resolution 1080p / 4K, frame rate 30fps / 60fps, encoding format H.264 / H.265. Sensor data: data type temperature / humidity, sampling frequency 1Hz / 10Hz, value range 0 - 100°C. Control instruction: instruction type switch / regulation, target device air conditioner / lighting, execution priority. Combining the calculation result of Shannon entropy, the dynamic nature of the data is quantified. For high-entropy data such as video stream: with drastic content changes, high bandwidth and low compression ratio are adopted; for low-entropy data such as environmental sensors: with stable content, high compression ratio transmission is used; the importance weights of each feature are dynamically calculated through the self-attention mechanism to generate a protocol stack configuration adapted to the service. For example, for video stream: bandwidth is preferentially guaranteed, H.265 encoding is enabled, and UDP transmission is used; for environmental monitoring: 80% compression ratio is adopted, and CoAP transmission is used; for control instruction: HTTP transmission, without encryption, and latency constraint <10ms.

[0087] A smart campus integrated communication system and its information interaction method of the present application can obtain the data stream of the access terminal in real time through a multi-protocol adapter, perform protocol decoupling and standard encapsulation on the data stream by using a QoS-based bandwidth prediction model, and generate a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix, enabling the system to be compatible with multiple communication protocols, solving the problem that different protocol devices are difficult to be compatible in the traditional system, improving the versatility and scalability of the system, and at the same time improving the efficiency of data transmission, ensuring that bandwidth can be reasonably allocated under limited network resources, avoiding data congestion, and enhancing the overall performance of the system. The service information data in the data stream is extracted through a data semantic parsing model, and its Shannon entropy is calculated. A transmission queue mapping table is constructed according to the Shannon entropy and the bandwidth allocation scheme to realize dynamic allocation of the transmission queue according to the priority attributes of the data stream, ensuring that high-priority data can be transmitted preferentially and enhancing the response speed of the system to urgent or important data. The feature vectors of the service information data are extracted by using a multi-modal feature collaborative extraction mechanism, and an interaction protocol stack adapted to the service is generated through an attention weight fusion mechanism, solving the compatibility and efficiency problems in cross-modal data interaction, enabling the system to better process and integrate data from different modalities, and enhancing the user experience and the intelligent level of the system.

[0088] In one of the embodiments, the hierarchical transmission module 103 is further configured to:

[0089] Calculate the Shannon entropy value using the following formula:

[0090]

[0091] Among them, p i (t) represents the traffic ratio of the i-th type of data stream within the window time t, wi (t) represents a dynamic weight factor determined by data priority and real-time network status; Priority i represents a fixed priority, Delay i (t) represents the current average delay of the i-th type of data stream.

[0092] Specifically, in the formula for calculating the Shannon entropy value, p i (t) reflects the relative traffic size of the i-th type of data stream within a specific time window, w i (t) is used to weight various types of data streams according to the priority of the data and the current network condition to adapt to different transmission requirements, Priority i is the preset priority value for various types of data streams, which can be determined according to the importance and urgency of the service, Delay i (t) reflects the delay situation of the data stream during the transmission process, which affects the real-time performance and transmission efficiency of the data. By substituting these parameters into the formula, the Shannon entropy value H can be calculated, and then the distribution characteristics of the data stream can be evaluated, providing a basis for subsequent transmission queue allocation and resource scheduling.

[0093] In one embodiment, the cross-modal interaction module 104 includes a course coordination unit for:

[0094] Obtaining student progress data, interest preference data, and teacher available time data in real time;

[0095] Calculating the course demand value of the student according to the student progress data and interest preference data, and combining the time window and subject type;

[0096] Calculating the teacher spatio-temporal matching degree tensor and generating a teacher time matching matrix based on the teacher available time data;

[0097] Calculating the matching degree between the course demand value and the teacher time matching matrix through a matching degree algorithm;

[0098] If the matching degree is higher than the preset threshold, generating a course arrangement plan;

[0099] If the matching degree does not meet the threshold, adjusting the time window and recalculating the matching degree;

[0100] Generating course schedule data according to the course arrangement plan, storing it in the course database, and simultaneously feeding it back to the student and teacher terminals.

[0101] Exemplarily, three types of data can be collected in real time, such as: the learning progress data of students, which helps to understand the learning stages of students in various subjects; the interest preference data of students, which reflects the degree of interest of students in different subjects; and the available time data of teachers, which determines the free time of teachers in different time periods. According to the learning progress and interest preference data of students, and combined with a set time window, such as a one-week time range and subject type, calculate the course demand value of each student in different subjects. The calculation of the demand value can be based on factors such as the time spent by the student on a certain subject, the number of tasks completed, or the intensity of interest shown by the student in a certain subject. Based on the available time data of teachers, construct a teacher spatio-temporal matching degree tensor of a multi-dimensional data structure, which is used to represent the matching degree of teachers in different times and spaces, such as different classrooms or teaching areas. Convert the teacher's spatio-temporal matching degree tensor into a teacher time matching matrix, which can intuitively display the matching situation of teachers in each time period. Use the matching degree algorithm to calculate the matching degree between the course demand value and the teacher time matching matrix. The course arrangement plan that is most suitable for the teacher's time arrangement while meeting the course needs of students can be found. If the calculated matching degree is higher than the preset threshold, it means that the current course arrangement plan is feasible and of high quality, and the system will generate a specific course arrangement plan. If the matching degree does not meet the threshold, the system will automatically adjust the time window, such as extending or shortening the time range, and recalculate the matching degree until a course arrangement plan that meets the conditions is found. If a course arrangement plan that meets the requirements is generated, it is converted into course schedule data and stored in the course database for subsequent query and management. At the same time, the generated course schedule is fed back to the terminal devices of students and teachers, so as to realize the automation and intelligence of the entire course coordination process.

[0102] In one of the embodiments, the course coordination unit is further configured to:

[0103] Calculate the course demand value of students using the following formula:

[0104]

[0105] where P(t) represents the progress completion rate at time t, P0 represents the midpoint of the Sigmoid function, and k p represents the progress sensitivity coefficient, where T avg represents the average learning duration, α(t) represents the time-varying weight, I i (t) represents the multi-dimensional interest vector, S k represents the subject type tensor, and C represents the teaching resource feature matrix.

[0106] Specifically, P(t) represents the completion degree of a student for a certain course at time t, which is adjusted by the Sigmoid function. Here, is the midpoint of the Sigmoid function, determining the position of the symmetric center of the function; is the progress sensitivity coefficient, determining the steepness of the function, and its value ranges from 0 to 1. T avg represents the average learning time of the student in this course, reflecting the degree of investment of the student in the course. α(t) represents the weight adjusted with time, which may be dynamically adjusted according to factors such as the difficulty of the course and the learning stage. I i (t) represents the interest preferences of the student in different dimensions, such as different subjects, different teaching methods, etc. S k is used to represent the characteristics and attributes of different subject types. C contains the characteristic information of various teaching resources such as textbooks, videos, experiments, etc., and is used to evaluate the matching degree between the teaching resources and the student needs. By comprehensively calculating these factors, the course demand value of the student is obtained, which can reflect the demand intensity of the student for a certain course at a specific time t, providing a basis for subsequent course arrangement and resource allocation.

[0107] In one of the embodiments, the course coordination unit is further configured to:

[0108] calculate the teacher space-time matching degree tensor using the following formula:

[0109]

[0110] where, u i and σ i represent the Gaussian distribution parameters of the teaching time preference of teacher i, Entropy(T k ) represents the global conflict entropy of the time window K, represents the recursive penalty factor, i represents the teacher, j represents the time window, and k represents the subject type.

[0111] Exemplarily, u i and σ i can reflect the time periods to which the teacher tends to teach. Entropy(T k ) represents the global conflict degree of the teaching times of all teachers within the time window K. The higher the entropy value, the more conflicts there are and the more complex the time arrangement is. It is used to punish mismatched situations during the recursive calculation process to adjust the calculation result of the matching degree, ensuring that the matching degree can reflect the actual feasibility and rationality. i is the index of the teacher, used to distinguish different teachers; j is the index of the time window, used to distinguish different time windows; k is the index of the subject type, used to distinguish different subject types. By considering the above parameters, the teacher spatio-temporal matching degree tensor can be calculated using this formula, which can intuitively show the matching degree of each teacher under different time windows and subject types, providing a scientific basis for subsequent course scheduling and resource allocation.

[0112] In one embodiment, the cross-modal interaction module 104 further includes a learning resource allocation unit, which is used for:

[0113] Based on the course scheduling plan, using a resource prediction model to predict the total amount of teaching resources required in different time periods and generate a resource prediction result;

[0114] According to the resource prediction result, divide the teaching resources into two parts: dynamic allocation and fixed allocation; the dynamically allocated part is used to cope with course adjustments, and the fixed allocated part is used to ensure the stability of basic resources;

[0115] Obtain resource allocation information in real time and detect resource overload or idleness;

[0116] If resource overload or idleness is detected, use a resource reallocation mechanism to reallocate the resources in the dynamically allocated part;

[0117] Through a resource scheduling algorithm, match the reallocated teaching resources with the course scheduling plan to generate a resource allocation table.

[0118] Specifically, it can be to use resource prediction models such as time series models, machine learning models, and deep learning models to predict the total amount of teaching resources required in different time periods, and generate resource prediction results. According to the resource prediction results, the resource requirements can be understood in advance, which is conducive to subsequent resource allocation. According to the resource prediction results, the teaching resources are divided into two parts: the dynamically allocated part and the fixed allocated part. Among them, the dynamically allocated part is mainly used to cope with the changes in resource requirements brought about by curriculum adjustments, and the fixed allocated part is used to ensure the stability of basic resources and maintain the normal progress of daily teaching activities. Obtain resource allocation information in real time, and detect resource overload or idleness, that is, continuously monitor the usage of resources to determine whether there is a problem of uneven resource allocation, such as some course resources being over-occupied or some resources not being fully utilized. When resource overload or idleness is detected, use the resource reallocation mechanism to reallocate the resources in the dynamically allocated part, which can optimize the resource utilization efficiency and ensure that the teaching resources can be flexibly adjusted according to the actual needs. Through the resource scheduling algorithm, match the reallocated teaching resources with the curriculum arrangement plan to generate the final resource allocation table, ensuring that the allocation of teaching resources is consistent with the curriculum arrangement.

[0119] In one embodiment, the cross-modal interaction module 104 further includes a personalized push unit for:

[0120] Obtain the user's historical behavior and feature data to form a user preference database;

[0121] Based on the user preference database, use preset matching rules to generate a push content set;

[0122] Based on the push content set, use the distribution channel to send the push content to the user terminal.

[0123] Exemplarily, collect the user's operation history and feature information, such as browsing records, learning preferences, interest tags, etc., to build a user preference database, providing a user portrait basis for subsequent personalized pushes. Based on the user preference database, use preset matching rules to screen out a set of push content that meets the user's interests and needs from the push library, and send the generated push content set to the user's terminal device through distribution channels such as mobile applications and web pages, enabling the user to obtain content that matches their needs and interests.

[0124] In one embodiment, the personalized push unit is further used for:

[0125] Based on the push content set, calculate the matching degree between the user and the push content through a data model;

[0126] If the matching degree is higher than the preset threshold, include the content in the push queue;

[0127] Select the push content with the highest priority from the push queue according to the delivery strategy;

[0128] Use the distribution channel to send the push content with the highest priority to the user terminal;

[0129] Record the user click and feedback data, and update the user preference database and the preset matching rules.

[0130] Specifically, after the personalized push unit generates the push content set, it can calculate the matching degree between the user and the push content through models such as the content-based matching model and the collaborative filtering model. If the matching degree is higher than the preset threshold, the content is included in the push queue. According to the delivery strategy, select the push content with the highest priority from the push queue and send it to the user terminal through the distribution channel. Record the user click and feedback data, and update the user preference database and the preset matching rules, which can optimize the accuracy of the push content and the user experience.

[0131] In one embodiment, the cross-modal interaction module 104 further includes a teaching achievement analysis unit for:

[0132] Obtain the teaching interaction data, extract the interaction volume and eigenvalue of the teaching interaction data, and construct an interaction feature set;

[0133] Based on the interaction feature set, use the random forest algorithm to train and generate an analysis model;

[0134] Based on the teaching interaction data, use the analysis model to output an analysis value;

[0135] Generate a teaching interaction trend analysis result according to the output analysis value.

[0136] Exemplarily, by deeply analyzing the teaching interaction data, the trends of teaching outcomes are evaluated and predicted. Various interaction data in the teaching process are obtained, such as the communication frequency between students and teachers, the cooperation situation among students, the feedback of students on teaching content, etc. The interaction volume and eigenvalue are extracted from these data. The interaction volume can refer to the number of interaction behaviors that occur, and the eigenvalue can include the duration of interaction, participation, complexity of content, etc. The extracted information is integrated to generate an interaction feature set, which serves as the basis for subsequent analysis. The random forest algorithm is used to train the analysis model. Random forest is a machine learning algorithm that improves the accuracy and stability of prediction by constructing multiple decision trees and integrating their results. During the machine learning process, the interaction feature set is used as training data to provide the model with the potential relationship between features and teaching outcomes. After the analysis model is trained, new teaching interaction data is processed and an analysis value is output. This analysis value can be a quantitative evaluation of teaching outcomes such as teaching quality and students' learning effects, or a prediction of future teaching trends. Generating the teaching interaction trend analysis result based on this analysis value can help teachers understand the advantages and disadvantages in the teaching process, discover potential problems in students' learning, and thus timely adjust teaching strategies, optimize teaching content and methods to improve the overall teaching effect.

[0137] The above-mentioned intelligent campus integrated communication system and its information interaction method support multiple communication protocols through a multi-protocol adapter, solve the problem of data interconnection caused by protocol differences among different terminals, provide a unified data entry, and meet the high-concurrency and low-latency data interaction requirements in the intelligent campus scenario. It adopts a streaming data processing framework to segment and dynamically buffer the continuously incoming data stream, avoiding data loss or delay caused by sudden traffic. Through a QoS-based bandwidth prediction model and a dynamic bandwidth allocation algorithm, the bandwidth is reasonably allocated to ensure the efficiency and quality of data transmission. The data semantic parsing model is used to extract the business information data in the data stream and calculate its Shannon entropy. According to the Shannon entropy and the bandwidth allocation scheme, a transmission queue mapping table is constructed to realize the dynamic allocation of transmission queues according to the priority attributes of the data stream, ensuring the priority transmission of high-priority data. Through a multi-modal feature collaborative extraction mechanism and an attention weight fusion mechanism, an interactive protocol stack adapted to the service is generated to solve the compatibility and efficiency problems in cross-modal data interaction, and improve the user experience and the intelligent level of the system. It obtains the data of students and teachers in real time, calculates the course demand value and the teacher's space-time matching degree, generates a course arrangement plan and a resource allocation table, and realizes the automation and intelligence of course coordination and resource allocation. Based on the user preference database and preset matching rules, a push content set is generated and sent to the user terminal through the distribution channel to improve the personalization and accuracy of learning resources. It obtains the teaching interaction data, extracts the interaction features, uses the random forest algorithm to train and generate an analysis model, outputs the analysis value, and generates the teaching interaction trend analysis result to help teachers understand the advantages and disadvantages in the teaching process, adjust the teaching strategy in time, and optimize the teaching content and methods.

[0138] It should be understood that although the steps in the flowcharts involved in the above-mentioned embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0139] Based on the same inventive concept, an embodiment of the present application further provides a method for intelligent campus fusion communication information interaction for implementing an intelligent campus fusion communication system involved above. The implementation solutions provided by this method to solve problems are similar to the implementation solutions recorded in the above system. Therefore, the specific limitations in one or more embodiments of the intelligent campus fusion communication information interaction method provided below can refer to the limitations on the intelligent campus fusion communication system in the above text, and will not be elaborated here.

[0140] In an exemplary embodiment, as Figure 2 shown, a method for intelligent campus fusion communication information interaction is provided, including:

[0141] S1, obtaining the data stream of the access terminal in real time through a multi-protocol adapter;

[0142] S2, using a bandwidth prediction model based on QoS to perform protocol decoupling and standard encapsulation on the data stream, and generating a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix;

[0143] S3, extracting the service information data contained in the decoupled data stream through a data semantic parsing model;

[0144] S4, calculating the Shannon entropy of the service information data, and constructing a transmission queue mapping table according to the Shannon entropy and the bandwidth allocation scheme; the transmission queue mapping table is used to dynamically allocate transmission queues according to the priority attributes of the data stream;

[0145] S5, based on the transmission queue mapping table, using a multi-modal feature collaborative extraction mechanism to extract the feature vectors of the service information data, and generating an interaction protocol stack adapted to the service through an attention weight fusion mechanism.

[0146] In one of the embodiments, calculating the Shannon entropy of the service information data includes:

[0147] Calculating the Shannon entropy value using the following formula:

[0148]

[0149] where p i (t) represents the traffic ratio of the i-th type of data stream within the window time t, and w i (t) represents the dynamic weight factor determined by the data priority and the real-time network state; Priority i represents the fixed priority, and Delay i (t) represents the current average delay of the i-th type of data stream.

[0150] In one of the embodiments, this method further includes:

[0151] Real-time access to student progress data, interest preference data, and teacher availability data;

[0152] Calculate students' course demand based on their progress data and interest preference data, combined with time windows and subject types;

[0153] Based on the teacher's available time data, the teacher's spatiotemporal matching tensor is calculated and the teacher's time matching matrix is generated;

[0154] Calculate the matching degree between course demand value and teacher time matching matrix through matching degree algorithm;

[0155] If the matching degree is higher than the preset threshold, a course arrangement plan is generated;

[0156] If the matching degree does not meet the threshold, adjust the time window and recalculate the matching degree;

[0157] According to the course schedule plan, the course schedule data is generated and stored in the course database, and fed back to the student and teacher terminals at the same time.

[0158] In one embodiment, the method further comprises:

[0159] Calculate the course requirement value for a student using the following formula:

[0160]

[0161] Among them, P(t) represents the progress completion rate at time t, P0 represents the midpoint of the Sigmoid function, k p represents the progress sensitivity coefficient, where T avg represents the average learning time, α(t) represents the time-varying weight, I i (t) represents the multidimensional interest vector, S k Represents the subject type tensor, and C represents the teaching resource feature matrix.

[0162] In one embodiment, the method further includes: calculating the teacher spatiotemporal matching tensor using the following formula:

[0163]

[0164] Among them, u i and σ i represents the Gaussian distribution parameter of teacher i’s teaching time preference, Entropy(T k ) represents the global conflict entropy of time window K, represents the recursive penalty factor, i represents the teacher, j represents the time window, and k represents the subject type.

[0165] In one embodiment, the method further comprises:

[0166] Based on the curriculum arrangement plan, a resource prediction model is adopted to predict the total amount of teaching resources required in different time periods, and a resource prediction result is generated.

[0167] According to the resource prediction result, the teaching resources are divided into two parts: dynamic allocation and fixed allocation; the dynamically allocated part is used to cope with curriculum adjustments, and the fixed allocated part is used to ensure the stability of basic resources.

[0168] Obtain resource allocation information in real time, and detect resource overload or idleness.

[0169] If resource overload or idleness is detected, use the resource reallocation mechanism to reallocate the resources in the dynamically allocated part.

[0170] Through the resource scheduling algorithm, match the reallocated teaching resources with the curriculum arrangement plan to generate a resource allocation table.

[0171] In one embodiment, the method further includes:

[0172] Obtain user historical behavior and feature data to form a user preference database.

[0173] Based on the user preference database, use preset matching rules to generate a push content set.

[0174] Based on the push content set, use the distribution channel to send the push content to the user terminal.

[0175] In one embodiment, the method further includes:

[0176] Based on the push content set, calculate the matching degree between the user and the push content through a data model.

[0177] If the matching degree is higher than the preset threshold, include the content in the push queue.

[0178] According to the delivery strategy, select the push content with the highest priority from the push queue.

[0179] Use the distribution channel to send the push content with the highest priority to the user terminal.

[0180] Record the user click and feedback data, and update the user preference database and the preset matching rules.

[0181] In one embodiment, the method further includes:

[0182] Obtain teaching interaction data, extract the interaction volume and feature values of the teaching interaction data, and construct an interaction feature set.

[0183] Based on the interaction feature set, use the random forest algorithm to train and generate an analysis model.

[0184] Based on teaching interaction data, an analysis value is output using an analysis model;

[0185] According to the output analysis value, a teaching interaction trend analysis result is generated.

[0186] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of a smart campus integrated communication system and its information interaction method as described above are implemented.

[0187] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0188] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment. The device embodiments described above are merely illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0189] The above embodiments only represent several implementation manners of the embodiments of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the application embodiments. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A smart campus integrated communication system, characterized in that, The system includes a terminal access module, a protocol conversion module, a hierarchical transmission module, and a cross-modal interaction module: The terminal access module is used for: Obtaining the data stream of the access terminal in real time through a multi-protocol adapter; The protocol conversion module is used for: Adopting a bandwidth prediction model based on QoS to perform protocol decoupling and standard encapsulation on the data stream, and generating a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix; The hierarchical transmission module is used for: Obtaining the data stream of the access terminal in real time through a multi-protocol adapter; Adopting a bandwidth prediction model based on QoS to perform protocol decoupling and standard encapsulation on the data stream, and generating a bandwidth allocation scheme according to the dynamic bandwidth allocation algorithm in combination with the network topology state matrix; Extracting the service information data contained in the decoupled data stream through a data semantic parsing model; Calculating the Shannon entropy of the service information data, and constructing a transmission queue mapping table according to the Shannon entropy and the bandwidth allocation scheme; the transmission queue mapping table is used for dynamically allocating transmission queues according to the priority attributes of the data stream; Based on the transmission queue mapping table, using a multi-modal feature collaborative extraction mechanism to extract the feature vectors of the service information data, and generating an interaction protocol stack adapted to the service through an attention weight fusion mechanism.

2. The system according to claim 1, wherein The hierarchical transmission module is also used for: Calculating the Shannon entropy value using the following formula: Among them, p i (t) represents the traffic proportion of the i-th type of data flow within the window time t, and w i (t) represents the dynamic weight factor determined by data priority and real-time network status; Priority i represents the fixed priority, and Delay i (t) represents the current average delay of the i-th type of data flow.

3. The system according to claim 1, wherein The cross-modal interaction module includes a course coordination unit, which is used for: Obtaining student progress data, interest preference data, and teacher available time data in real time; Calculating the course demand value of the student according to the student progress data and interest preference data, and combining the time window and subject type; Based on the teacher available time data, calculating the teacher spatio-temporal matching degree tensor and generating a teacher time matching matrix; Calculating the matching degree between the course demand value and the teacher time matching matrix through a matching degree algorithm; If the matching degree is higher than the preset threshold, generating a course arrangement plan; If the matching degree does not meet the threshold, adjusting the time window and recalculating the matching degree; According to the course arrangement plan, generating course schedule data, storing it in the course database, and simultaneously feeding it back to the student and teacher terminals.

4. The system according to claim 3, wherein The course coordination unit is also used for: Calculating the course demand value of the student using the following formula: Among them, P(t) represents the progress completion rate at time t, P0 represents the midpoint of the Sigmoid function, and k p represents the progress sensitivity coefficient, where T avg represents the average learning duration, α(t) represents the time-varying weight, and I i (t) represents the multi-dimensional interest vector, and S k represents the subject type tensor, and C represents the teaching resource feature matrix.

5. The system according to claim 3, characterized in that, The course coordination unit is also used for: Calculating the teacher spatio-temporal matching degree tensor using the following formula: Among them, u i and σ i represent the Gaussian distribution parameters of teacher i's teaching time preference, Entropy(T k ) represents the global conflict entropy of time window K, represents the recursive penalty factor, i represents the teacher, j represents the time window, and k represents the subject type.

6. The system according to claim 3, wherein The cross-modal interaction module also includes a learning resource allocation unit, which is used for: Based on the course arrangement plan, adopting a resource prediction model to predict the total amount of teaching resources required in different time periods, and generating a resource prediction result; According to the resource prediction result, dividing the teaching resources into two parts: dynamic allocation and fixed allocation; the dynamic allocation part is used to cope with course adjustments, and the fixed allocation part is used to ensure the stability of basic resources; Obtaining resource allocation information in real time, and detecting resource overload or idleness; If resource overload or idleness is detected, using a resource reallocation mechanism to reallocate the resources in the dynamic allocation part; Through a resource scheduling algorithm, matching the reallocated teaching resources with the course arrangement plan to generate a resource allocation table.

7. The system according to claim 1, wherein The cross-modal interaction module further includes a personalized push unit, which is used for: Obtaining the user's historical behavior and feature data to form a user preference database; Based on the user preference database, using a preset matching rule to generate a push content set; Based on the push content set, using a distribution channel to send the push content to the user terminal.

8. The system according to claim 7, wherein The personalized push unit is further used for: Based on the push content set, calculating the matching degree between the user and the push content through a data model; If the matching degree is higher than a preset threshold, include the content in the push queue; According to the delivery strategy, select the push content with the highest priority from the push queue; Using a distribution channel to send the push content with the highest priority to the user terminal; Recording the user's click and feedback data, and updating the user preference database and the preset matching rule.

9. The system according to claim 1, wherein The cross-modal interaction module further includes a teaching achievement analysis unit, which is used for: Obtaining teaching interaction data, extracting the interaction volume and feature values of the teaching interaction data, and constructing an interaction feature set; Based on the interaction feature set, using a random forest algorithm to train and generate an analysis model; Based on the teaching interaction data, using the analysis model to output an analysis value; According to the output analysis value, generating an analysis result of the teaching interaction trend.

10. A method for intelligent campus integrated communication information interaction, characterized in that, The method includes: Real-time obtaining the data stream of the access terminal through a multi-protocol adapter; the access terminal includes an Internet of Things sensing device, a mobile terminal, and an intelligent teaching facility; Adopting a QoS-based bandwidth prediction model to perform protocol decoupling and standardization encapsulation on the data stream, and generating a bandwidth allocation scheme according to a dynamic bandwidth allocation algorithm in combination with a network topology state matrix; Extracting the service information data included in the decoupled data stream through a data semantic parsing model; Calculating the Shannon entropy of the service information data, and dynamically constructing a transmission queue mapping table according to the Shannon entropy; the transmission queue mapping table is used for dynamically allocating a transmission queue according to the priority attribute of the data stream; Based on the transmission queue mapping table, using a multi-modal feature collaborative extraction mechanism to extract the feature vector of the service information data, and generating an interaction protocol stack adapted to the service through an attention weight fusion mechanism.

Images