Information sharing method of online teaching platform based on Internet of Things

Abstract

The invention relates to the technical field of information sharing, in particular to an information sharing method of an online teaching platform based on the Internet of Things. The method comprises the following steps: collecting real-time user behavior data, and constructing a user dynamic participation degree model to obtain a user participation degree model; obtaining a course content library, carrying out course feature integration, and obtaining an actual learning content feature descriptor of the user; performing associated course content recommendation strategy analysis on the course content library to obtain a user course content recommendation strategy; acquiring real-time user platform interaction data, and performing user course acceptability evaluation to obtain user course acceptability data; and performing recommendation strategy iterative optimization on the user course content subsequent recommendation strategy to obtain an optimized user course content recommendation strategy. According to the invention, the learning effect of education and the user satisfaction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information sharing, and in particular to an information sharing method of an online teaching platform based on the Internet of Things. Background Art

[0002] With the rapid development of information technology, especially the maturity of Internet of Things technology, innovation in the field of education is constantly driving the transformation of teaching models. As a product of the information age, online education has gradually become an important force in promoting the reform of teaching content, methods and management methods. Traditional teaching often relies on classroom lectures and paper textbooks. The teaching content and teaching methods are relatively fixed, the interactivity is poor, and the learning effect is limited. Especially in the Internet age, traditional teaching methods have failed to give full play to their due advantages, it is difficult to stimulate users' interest in learning, and it is impossible to make timely adjustments based on users' learning feedback, which affects the popularity and depth of education. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide an information sharing method for an online teaching platform based on the Internet of Things to solve at least one of the above technical problems.

[0004] To achieve the above purpose, an information sharing method of an online teaching platform based on the Internet of Things includes the following steps: Step S1: using the Internet of Things platform to collect real-time user behavior data, and quantifying user behavior perception of the real-time user behavior data, thereby obtaining user behavior perception data; constructing a user dynamic engagement model based on the user behavior perception data, thereby obtaining a user engagement model; Step S2: Obtain the course content library through the online teaching platform, and locate the user's learning course in the course content library based on the user participation model, so as to obtain the user's actual learning course data; integrate the course features according to the user's actual learning course data, so as to obtain the user's actual learning content feature descriptor; Step S3: Performing actual learning content-related course analysis on the course content library according to the user's actual learning content feature descriptor, thereby obtaining the user's actual learning content-related course data; performing related course content recommendation strategy analysis based on the user's actual learning content-related course data, thereby obtaining the user's course content recommendation strategy, and uploading it to the online teaching platform to perform the content recommendation task; Step S4: obtaining real-time user platform interaction data through the online teaching platform, and performing user feedback sentiment analysis on the real-time user platform interaction data through the user engagement model, thereby obtaining user feedback sentiment data; performing user course acceptance evaluation based on the user feedback sentiment data, thereby obtaining user course acceptance data; Step S5: Iteratively optimize the subsequent recommendation strategy for user course content according to the user course acceptance data, so as to obtain the optimized user course content recommendation strategy, and upload it to the online teaching platform to perform the content recommendation task.

[0005] The present invention collects real-time user behavior data through the Internet of Things platform and quantifies the behavior perception, which can accurately capture the interaction and reaction of users in the learning process, thereby providing a reliable data basis for subsequent learning analysis. By establishing a user dynamic participation model, it can not only dynamically evaluate the user's learning input and interest, but also accurately identify which courses or teaching methods have a greater appeal to users, thereby optimizing the teaching strategy. On this basis, based on the user's actual learning data, the course feature integration can better understand the user's specific needs and preferences in the learning process, and lay the foundation for customized recommendation of course content. Through the association analysis with the course content library, personalized learning content push can be achieved, ensuring that the recommended courses are highly matched with the user's interests and needs, thereby significantly improving learning efficiency and the attractiveness of the course. In addition, real-time acquisition of user platform interaction data and feedback sentiment analysis can timely capture the user's emotional response to the course, identify positive and negative emotions, thereby accurately evaluating the user's acceptance of the course content, and providing an important basis for the adjustment and optimization of subsequent courses. After evaluating the course acceptance based on the user's feedback sentiment data, the recommendation strategy of the course content can be continuously improved to ensure that the recommended content not only meets the user's interests, but also effectively improves their learning effect and participation. This series of steps form a dynamic feedback loop through mutual collaboration, enabling the education platform to continuously adjust teaching content and methods based on real-time data and user feedback, continuously optimize course recommendation strategies, and ultimately improve education learning outcomes and user satisfaction.

[0006] Optionally, step S1 specifically includes: Step S11: using the Internet of Things to collect real-time user behavior data, and performing data preprocessing on the real-time user behavior data, thereby obtaining real-time user behavior data to be analyzed; Step S12: extracting behavior perception features from the real-time user behavior data to be analyzed, thereby obtaining user facial expression images, user action data, and user audio data; Step S13: performing emotion fluctuation sentiment analysis according to the user's facial expression image and the user's audio data, thereby obtaining the user's facial emotion data and the user's voice emotion data; Step S14: performing user action emotion analysis according to the user action data to obtain user action emotion data, and integrating the user action emotion data, the user facial emotion data, and the user voice emotion data into behavioral features to obtain user behavior perception data; Step S15: construct a user dynamic engagement model based on the user behavior perception data, thereby obtaining a user engagement model.

[0007] The present invention collects real-time user behavior data and performs preprocessing through the Internet of Things technology, effectively ensuring the accuracy and reliability of subsequent analysis data, and helping to dynamically monitor the user's learning behavior and its emotional changes. The behavior perception feature extraction makes it possible to capture the user's facial expressions, actions, sounds and other interactive methods, so as to fully understand the emotional fluctuations of the user in the learning process. Through the emotional fluctuations analysis of facial expression images and audio data, the emotional state of the user can be deeply excavated, and the emotional changes they have produced in the learning process can be evaluated, which helps to accurately identify which course content or teaching methods can cause the user's positive emotional response, and then provide a basis for personalized recommendation and teaching optimization. Through the emotional analysis of user action data, the user's body language and behavioral response are further understood, helping to identify their participation in the course content and interest changes. After integrating facial emotion data, voice emotion data and action emotion data, the user's overall emotional feedback and participation status can be obtained more comprehensively, and a more accurate behavior perception model can be established. This comprehensive analysis will provide key data support for the construction of the user's dynamic participation model, ensure that the user's learning participation can be grasped in real time, and provide a scientific basis for the personalized adjustment and optimization of teaching content.

[0008] Optionally, step S13 is specifically: Performing facial feature point alignment on the user's facial expression image to obtain a facial feature point aligned image; Perform facial feature point combination extraction on the facial feature point aligned image to obtain facial expression feature data; Classify the facial expression and emotional state according to the facial expression feature data, so as to obtain the user's facial emotion data; Performing audio data framing on the user audio data to obtain a user sound frame set, and performing Mel-frequency cepstral coefficient sound feature extraction on the user sound frame set to obtain user sound frame feature data; Audio emotion state classification is performed based on the user voice frame feature data to obtain user voice emotion data.

[0009] The present invention can accurately standardize the user's facial expression by aligning the feature points of the user's facial expression image, thereby eliminating the inconsistency caused by different facial angles and shooting conditions, and ensuring that high-quality data is obtained during the analysis process. This step provides clear and stable facial features for further emotional analysis and ensures the accuracy of facial expression feature extraction. Through the combined extraction of facial feature points, feature information highly related to the user's emotional state can be extracted, laying a solid foundation for subsequent emotional classification. The facial expression feature data extracted in this process can help to deeply analyze the user's emotional response and identify their emotional fluctuations such as pleasure, confusion, and interest during the learning process, thereby providing data support for the adjustment of personalized learning content. At the same time, the framing of audio data and the Mel Frequency Cepstral Coefficient (MFCC) feature extraction technology can effectively capture the key emotional information in the voice. The MFCC feature is the time-frequency feature of the audio signal, which can accurately characterize the emotional characteristics of the voice, thereby improving the accuracy of emotional state classification. These voice feature data are closely related to the user's emotional state, helping to analyze the changes in the user's psychological state during the interaction process. Combining the analysis of facial emotion data and voice emotion data, it is possible to comprehensively assess the user's emotional state, thereby improving the personalized service effect of education, ensuring that the course content and teaching methods are highly consistent with the user's emotional needs, and thus achieving more accurate teaching intervention and learning motivation.

[0010] Optionally, step S14 is specifically: Step S141: estimating the user's limb state based on the user's motion data, thereby obtaining the user's limb state data; Step S142: performing limb state time series analysis according to the user's limb state data, thereby obtaining limb state time series data; Step S143: performing time window limb state change statistics on the limb state time series data according to a preset time window, thereby obtaining periodic limb state change data and non-periodic limb state change data; Step S144: classifying the limb state change frequency according to the non-periodic limb state change data, thereby obtaining non-periodic high-frequency limb state change data and non-periodic low-frequency limb state change data; Step S145: performing a user gentle motion emotion analysis on the periodic limb state change data and the non-periodic low-frequency limb state change data, thereby obtaining the user gentle motion emotion data; performing a user tense motion emotion analysis on the non-periodic high-frequency limb state change data, thereby obtaining the user tense motion emotion data; Step S146: performing user action emotion pattern recognition based on the user's smooth action emotion data and the user's tense action emotion data, thereby obtaining user action emotion data; Step S147: Integrate the user action emotion data, the user facial emotion data, and the user voice emotion data into behavioral features to obtain user behavior perception data.

[0011] The present invention can fully capture the user's body language and posture changes by estimating the user's motion data, and then understand the possible emotions and participation levels in the education process. This data provides a basis for subsequent emotional analysis, so that the system can judge the user's emotional state based on the user's non-verbal behavior. Analyzing the limb state time series data can reveal the user's motion change pattern, help identify the user's state changes at different learning stages, and accurately capture the time series law of emotional fluctuations. The time window limb state change statistics can further quantify the user's limb movement characteristics, and distinguish between periodicity and non-periodicity, thereby providing a more refined data level for understanding the user's emotional pattern. This analysis helps to identify whether the user has continuous emotional fluctuations or recurring behavior patterns during the learning process, and then helps teachers or education platforms adjust teaching strategies. By frequency classification of non-periodic limb state change data, the user's emotional response types in different situations can be clearly distinguished, especially the high-frequency nervous emotions and low-frequency flat states shown by the user in the learning process can be accurately identified, which is of great significance for understanding the user's cognitive load and learning effect. Emotional analysis of these changes in body states can effectively assess whether users are anxious, confused, or relaxed during the learning process, providing more personalized data support for subsequent course recommendations and teaching interventions. By integrating these emotional data with facial and voice data through action emotion pattern recognition, a comprehensive map of the user's emotional state can be drawn. This integration of multimodal data helps to accurately identify the user's emotional patterns, thereby providing strong data support for providing teaching content and interactive design that better meets user needs.

[0012] Optionally, step S2 specifically includes: Step S21: Obtaining a course content library through an online teaching platform; Step S22: locating the user's learning course based on the real-time user behavior data to be analyzed according to the course content library, thereby obtaining the user's learning course data; Step S23: evaluating the user course learning data for user course learning participation through a user participation model, thereby obtaining user course participation data; Step S24: removing low-participation courses from the user's course data according to the user's course participation data, thereby obtaining the user's actual course data; Step S25: Integrate course features according to the user's actual course learning data, so as to obtain the user's actual learning content feature descriptor.

[0013] The present invention can provide a comprehensive teaching resource foundation for the online teaching platform by acquiring the course content library. These resources can accurately match the user's learning needs and interests and provide support for personalized teaching. Then, by positioning the learning course for the real-time user behavior data to be analyzed, the user's behavior is effectively associated with the course content, helping the platform to accurately identify the user's learning direction and interest preference. This process can not only ensure that the courses the user is exposed to are consistent with their learning needs, but also provide an accurate course data basis for subsequent data analysis. By combining the user engagement model to evaluate the user's learning course data, the user's participation in the learning process can be monitored and quantified in real time, including browsing time, interaction frequency, etc. This engagement evaluation helps to identify the user's learning investment, so as to determine whether there are problems such as learning fatigue and loss of interest. By eliminating low-engagement courses, the platform can optimize the matching of course content, so that users can focus more on courses that can stimulate learning interest and improve engagement. This measure effectively improves the relevance of learning content and avoids the interference of inefficient course content on the user's learning process. Finally, by integrating the course characteristics of the user's actual learning course data, the platform can extract the content characteristics that the user has been exposed to during the learning process, thereby forming a detailed user learning content feature descriptor. These descriptors provide refined data support for subsequent course recommendations, learning path planning, etc., enabling the teaching platform to design a personalized learning experience based on each user's unique learning characteristics, further improving teaching effectiveness and learning quality.

[0014] Optionally, step S22 is specifically: Step S221: obtaining a teaching platform page interaction log through the online teaching platform; Step S222: classifying the user behavior collection type of the real-time user behavior data to be analyzed, thereby obtaining the user behavior continuously collected data and the user behavior instantaneous collected data; Step S223: time matching the continuously collected data of user behavior and the teaching platform page interaction log, thereby obtaining the continuously interactive teaching page data; time matching the instantaneously collected data of user behavior and the teaching platform page interaction log, thereby obtaining the instantaneous interactive teaching page data; Step S224: extracting teaching page teaching course features from the continuous interactive teaching page data, thereby obtaining continuous interactive teaching course data; extracting teaching page teaching course features from the instantaneous interactive teaching page data, thereby obtaining instantaneous interactive teaching course data; Step S225: performing a teaching course intersection operation on the continuous interactive teaching course data and the instantaneous interactive teaching course data, thereby obtaining user interactive teaching course data; Step S226: Locate the teaching track of the user interactive teaching course data through the course content library, so as to obtain the user learning course data.

[0015] The present invention can accurately record every interaction of users on the learning platform by obtaining the teaching platform page interaction log, ensuring comprehensive tracking and analysis of user behavior. This log data provides a basis for further understanding user learning habits, interest preferences and usage patterns. On this basis, by classifying the collection type of real-time user behavior data, it is possible to clearly distinguish different types of user behaviors, distinguish between continuous collection data and instantaneous collection data, and then help the system to more accurately process and analyze the data to be analyzed. Time matching of these two types of data further integrates user behavior data and platform interaction logs, so that the system can clearly identify the user's interactive behavior at each time point, ensuring that the analysis results are more accurate and dynamic. In view of the different characteristics of continuous interaction and instantaneous interaction, course feature extraction is performed on teaching page data, which not only improves the depth of course content analysis, but also enables the system to understand the user's interactive performance on different teaching pages from multiple dimensions, which is helpful to refine the personalized recommendation of course content. By performing intersection operations on the two types of data, user interactive teaching course data is obtained, which further strengthens the association between users and course content, and facilitates accurate understanding of the course content that each user is exposed to during the learning process. By locating the teaching trajectory of these interaction data and the course content library, we can accurately identify the user's learning path and progress, ensuring that each user's learning experience is tailored, thereby improving learning outcomes and course participation, and achieving the goal of personalized education.

[0016] Optionally, step S226 is specifically: The course teaching trajectory is divided into course teaching trajectory datasets based on the course content library. Performing background audio extraction on the user audio data to obtain background audio data, and performing audio data framing on the background audio data to obtain background sound frames; Performing speech recognition on the background sound frame to obtain background sound text data, and removing stop words from the background sound text data to obtain background text data to be matched; Matching the course content keywords with the background text data to be matched through the course content library, thereby obtaining audio matching teaching content data; According to the course teaching trajectory dataset, the audio is matched with the teaching content data to identify the user's learning course, so as to obtain the user's learning course data.

[0017] The present invention can accurately track the learning path and behavior of users in the learning process by dividing the course teaching track of user interactive teaching course data, and help the teaching platform understand the learning progress and focus of each user. Such a track data set provides an important basis for subsequent learning analysis and optimization, and is convenient for implementing personalized teaching and dynamically adjusting course content. The step of background audio extraction can effectively identify background noise from user audio data, ensure that the analysis focuses on valuable audio content, and improve the accuracy of subsequent processing. By performing frame processing on background audio data, not only the detailed analysis of audio is enhanced, but also it is convenient to divide it into different time periods for processing, so that the audio data is clearer and more accurate. The application of speech recognition technology can convert the language information in the audio into text, so that the system can more easily process, analyze and match this information, further improving the intelligence of the teaching system. The stop word removal step of the background sound text data can remove irrelevant words, make the text data to be matched more concise and accurate, reduce possible interference in the subsequent matching process, and improve the efficiency and accuracy of matching. By matching the background text with course content keywords, the system can extract important information related to the course from the text, helping to achieve an effective association between audio and teaching content. This provides data support for personalized recommendations of course content and can also provide teachers with real-time feedback on student learning. By matching audio with teaching content data based on the course teaching trajectory dataset, user learning courses can be identified, which can provide an in-depth understanding of the user's learning behavior and course matching, ensuring that each user's learning content and progress are accurately identified and tracked, providing strong support for personalized course recommendations and teaching interventions.

[0018] Optionally, step S3 specifically includes: Step S31: Calculate the course content similarity of the user's actual learning content feature descriptor and the course content library, thereby obtaining course content similarity data; Step S32: extracting courses with high similarity in course content from the course content library based on the course content similarity data, thereby obtaining course data associated with the user's actual learning content; Step S33: classifying the course data associated with the user's actual learning content into course priorities according to the course teaching trajectory data set, thereby obtaining associated course priority data; Step S34: performing an associated course complexity evaluation on the associated course data of the user's actual learning content, thereby obtaining associated course complexity data; Step S35: Plan a course content recommendation strategy based on the associated course priority data and the associated course complexity data, so as to obtain a user course content recommendation strategy, and upload it to the online teaching platform to perform the content recommendation task.

[0019] The present invention can accurately evaluate the relationship between the user's actual learning content feature descriptor and the course content library by calculating the course content similarity, so as to find the content suitable for the user to learn. Such similarity data helps to understand the user's learning interests and needs, and provides a basis for subsequent course recommendations. Based on similarity data, extracting high-similarity courses can effectively reduce the interference of irrelevant courses, ensure that the recommended content is more in line with the user's learning trajectory and interests, improve learning efficiency and enhance the user's learning experience. By dividing the course priorities of the user's actual learning content-related course data, it can be ensured that the recommended courses give priority to those courses that are of great help to the user's current learning progress, thereby improving the learning effect and avoiding unnecessary learning burden. Evaluating the complexity of the associated courses can identify the difficulty of the course content, customize a reasonable learning plan for the user, ensure that they improve their abilities under the appropriate challenge, and avoid the influence of too simple or too complex courses on learning motivation. Finally, by combining the course priority and complexity data for course content recommendation strategy planning, a personalized course recommendation plan can be tailored to further optimize the learning path and improve the learning effect. After uploading the recommendation strategy to the online teaching platform, the content recommendation task can be executed immediately to provide users with accurate learning content and realize dynamic adjustment to ensure the continuous optimization of teaching content.

[0020] Optionally, step S4 specifically includes: Step S41: acquiring real-time user platform interaction data through the online teaching platform, and performing data preprocessing on the real-time user platform interaction data, thereby obtaining the real-time user platform interaction data to be analyzed; Step S42: extracting interaction features from the real-time user platform interaction data to be analyzed, thereby obtaining user platform browsing time data and user platform comment content data; Step S43: performing comment text sentiment polarity analysis based on the user platform comment content data, thereby obtaining user platform positive comment content data and user platform negative comment content data; Step S44: classifying the platform browsing time according to the user platform browsing time data, thereby obtaining the user's continuous browsing content data and the user's instantaneous browsing content data; Step S45: performing content-related sentiment analysis on the user platform's positive comment content data and the user's continuous browsing content data, thereby obtaining the user's positive feedback sentiment data; performing content-related sentiment analysis on the user platform's negative comment content data and the user's instant browsing content data, thereby obtaining the user's negative feedback sentiment data; Step S46: merging the user's positive feedback emotion data and the user's negative feedback emotion data to obtain the user's feedback emotion data; Step S47: Evaluate user course acceptance based on user feedback sentiment data, thereby obtaining user course acceptance data.

[0021] The present invention can ensure the quality and consistency of data by acquiring user platform interaction data in real time and performing data preprocessing, and provide an efficient and accurate basis for subsequent analysis. Feature extraction of interaction data can not only help analyze user behavior patterns, but also dig out valuable information such as browsing time and comment content. Emotional polarity analysis based on comment content can effectively identify users' emotional attitudes towards courses, thereby classifying positive or negative feedback, and further providing a basis for the adjustment of personalized teaching content. Classifying platform browsing time data can help distinguish users' long-term participation from short-term interests, and provide an important basis for subsequent content recommendation and course design. By performing emotional analysis on positive comments and continuous browsing data, the emotional state of positive feedback can be identified. Conversely, emotional analysis on negative comments and instantaneous browsing data can help identify users' negative emotions and potential problems. The integration of these emotional data can fully reflect users' emotional attitudes and learning acceptance, and provide accurate information for evaluating users' interest and satisfaction in courses. Based on users' feedback emotional data, course acceptance evaluation is performed to help optimize course content and recommendation strategies, ensure that teaching content can better connect with user needs, and improve learning experience and learning effects.

[0022] Optionally, step S47 is specifically: Step S471: calculating the emotion intensity of the user feedback emotion data, thereby obtaining the user feedback emotion intensity data; Step S472: Calculate the emotional polarity ratio according to the emotional intensity data fed back by the user, thereby obtaining emotional polarity ratio data; Step S473: marking the user feedback emotion data with emotion polarity proportions based on the emotion polarity proportions data, thereby obtaining positive feedback emotion polarity proportions data and negative feedback emotion polarity proportions data; Step S474: Perform sentiment weighted scoring on the recommended content based on the positive feedback sentiment polarity ratio data and the negative feedback sentiment polarity ratio data, thereby obtaining user course acceptance data.

[0023] The present invention can quantify the intensity of user emotions by calculating the emotional intensity of user feedback emotional data, thereby more accurately reflecting the user's emotional response. This helps to identify which feedback has a higher emotional intensity, and then optimize the course content or interaction method in a targeted manner. By calculating the emotional polarity ratio, the relative proportion of positive and negative emotions in user feedback can be revealed, providing directional guidance for course content adjustment. If the emotional polarity ratio tends to be negative, it may indicate that there is a problem with the course content or method, and appropriate adjustments need to be made. Marking the emotional polarity ratio not only helps to distinguish the tendency of user emotions, but also can associate positive and negative feedback with specific course content or learning stages, further deepening the understanding of user feedback. Based on this data, emotional weighted scoring of recommended content can dynamically adjust course recommendations according to the user's emotional attitude to ensure that the recommended content is more in line with the user's learning needs and emotional state, thereby effectively improving the user's course acceptance and optimizing the learning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1 It is a schematic diagram of the steps of the information sharing method of the online teaching platform based on the Internet of Things of the present invention; Figure 2 Detailed step flow diagram of step S1 in the present invention; Figure 3 Detailed step flow diagram of step S2 in the present invention; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0025] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0027] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve this, please refer to Figures 1 to 3 The present invention provides an information sharing method for an online teaching platform based on the Internet of Things, the method comprising the following steps: Step S1: using the Internet of Things platform to collect real-time user behavior data, and quantifying user behavior perception of the real-time user behavior data, thereby obtaining user behavior perception data; constructing a user dynamic engagement model based on the user behavior perception data, thereby obtaining a user engagement model; In this embodiment, the IoT platform collects real-time user behavior data, including behavioral features such as clicks, dwell time, and mouse tracks, uses sensors and smart devices to record various interactive behaviors of users on the teaching platform, and quantifies these behavioral data. For example, the time a user stays on a course page and the frequency of interaction with the content are quantified as engagement indicators. Based on these data, a dynamic engagement model is constructed, and a weighted scoring system is used to evaluate the depth of interaction between each user and the platform content during the learning process, thereby obtaining a user engagement model.

[0029] Step S2: Obtain the course content library through the online teaching platform, and locate the user's learning course in the course content library based on the user participation model, so as to obtain the user's actual learning course data; integrate the course features according to the user's actual learning course data, so as to obtain the user's actual learning content feature descriptor; In this embodiment, the online teaching platform extracts course-related content from the course content library, and performs content matching based on the user engagement model, compares the actual learning content with the user's learning progress, and accurately locates the courses that the user has learned. Based on the user's interactive behavior data, such as learning time, reading depth, question answers, etc., the platform automatically integrates the characteristics of the course, such as difficulty level, knowledge point coverage, interactivity, etc., to form the user's actual learning content feature descriptor. This descriptor will serve as the basis for the user's course needs and provide a basis for subsequent course recommendations.

[0030] Step S3: Performing actual learning content-related course analysis on the course content library according to the user's actual learning content feature descriptor, thereby obtaining the user's actual learning content-related course data; performing related course content recommendation strategy analysis based on the user's actual learning content-related course data, thereby obtaining the user's course content recommendation strategy, and uploading it to the online teaching platform to perform the content recommendation task; In this embodiment, the algorithm calculates the correlation between the user's actual learning content and the courses that have not been learned in the course library according to the similarity between the user's learning content feature descriptor and other courses in the course content library, and selects courses with higher correlation for recommendation. Through course recommendation strategy analysis, combined with learning needs and interests, the course recommendation strategy is adjusted to match the recommended content with the user's learning progress, emotional attitude and participation model, ensuring that the recommended course content meets the user's learning needs and is uploaded to the platform for automatic recommendation tasks.

[0031] Step S4: obtaining real-time user platform interaction data through the online teaching platform, and performing user feedback sentiment analysis on the real-time user platform interaction data through the user engagement model, thereby obtaining user feedback sentiment data; performing user course acceptance evaluation based on the user feedback sentiment data, thereby obtaining user course acceptance data; In this embodiment, the user's interactive data, including comments, ratings, clicks and other interactive behaviors, is obtained in real time through the online teaching platform, and the user's emotional feedback, such as the user's positive comments, negative comments and emotional fluctuations during course learning, is analyzed using the participation model. The analysis tool uses natural language processing technology to perform sentiment analysis on the comment content, extracts positive and negative emotional features, and combines behavioral data for comprehensive sentiment analysis to obtain user feedback sentiment data. This data can be further used to evaluate the user's acceptance of the course, such as the user's preference for certain teaching methods, course difficulty or interactive forms, and finally outputs the user's course acceptance data.

[0032] Step S5: Iteratively optimize the subsequent recommendation strategy for user course content according to the user course acceptance data, so as to obtain the optimized user course content recommendation strategy, and upload it to the online teaching platform to perform the content recommendation task.

[0033] In this embodiment, the course recommendation strategy is continuously optimized based on the user's course acceptance data. Through sentiment analysis of user feedback, the recommended content can be adjusted in real time and the recommendation strategy can be optimized. For example, if the user shows a strong interest in a certain type of course and has a high degree of participation, the recommendation priority of this type of course will be increased. After multiple rounds of iterations, the recommendation system continuously adjusts the strategy to ensure that it can more accurately meet the user's personalized learning needs, and finally uploads the optimized recommendation strategy to the online teaching platform to perform the automatic recommendation task.

[0034] Optionally, step S1 specifically includes: Step S11: using the Internet of Things to collect real-time user behavior data, and performing data preprocessing on the real-time user behavior data, thereby obtaining real-time user behavior data to be analyzed; In this embodiment, the user's behavior data is collected in real time through IoT devices (such as smart bracelets, computers, mobile phones, cameras, microphones, etc.), including but not limited to somatosensory data, audio data, video data, etc. The user's movements, expressions, sounds and other information are captured by sensors, and then the raw data is cleaned and formatted using data preprocessing techniques such as data denoising, missing value filling and data standardization to ensure data quality and consistency. The user's facial expressions are collected using cameras, and the audio equipment captures the user's speech sound. The user's movements, such as waving and nodding, are recorded in combination with sensors to form real-time user behavior data to be analyzed.

[0035] Step S12: extracting behavior perception features from the real-time user behavior data to be analyzed, thereby obtaining user facial expression images, user action data, and user audio data; In this embodiment, the facial feature points of the collected user facial image data are aligned through machine vision technology to ensure the consistency of facial data in time and space, and to extract the user's expression features. For example, the facial recognition algorithm can extract expression features by calibrating facial feature points and combining a deep learning model (such as a convolutional neural network). In terms of audio data, speech recognition technology is used to frame the user's audio, and Mel-frequency cepstral coefficients (MFCC) and other technologies are used to extract sound features. These features represent the user's facial expression data, action data, and audio data, respectively, and serve as the basis for subsequent sentiment analysis.

[0036] Step S13: performing emotion fluctuation sentiment analysis according to the user's facial expression image and the user's audio data, thereby obtaining the user's facial sentiment data and the user's voice sentiment data; In this embodiment, the emotion analysis algorithm is used to analyze the emotion fluctuations of the user's facial expression image and audio data respectively. Through the trained deep learning model (such as LSTM network or emotion analysis convolutional neural network), the user's emotion fluctuations, such as changes in basic emotions such as joy, anger, sorrow, and happiness, are analyzed based on the facial feature data. The audio data analyzes the characteristics of the voice such as pitch, volume, and speaking speed, combined with the emotion recognition model, to evaluate the user's emotional state, such as positive, negative, or neutral. The user's facial emotion data and voice emotion data finally obtained provide the user's emotional fluctuations, which can help capture the user's emotional changes more accurately.

[0037] Step S14: performing user action emotion analysis according to the user action data to obtain user action emotion data, and integrating the user action emotion data, the user facial emotion data, and the user voice emotion data into behavioral features to obtain user behavior perception data; In this embodiment, motion capture technology is used to perform sentiment analysis on the user's motion data to identify the emotional information in the user's motion, for example, whether the user shows tension, anxiety or relaxation. Based on motion features (such as stretching of limbs, rapid movement, etc.) and deep learning models, the motion and emotional state can be associated and analyzed. Then, the facial emotion data, audio emotion data and motion emotion data are integrated, and multimodal learning technology (such as a model that integrates convolutional neural networks and recurrent neural networks) is used to finally obtain the user's comprehensive behavioral perception data. These data provide rich information for evaluating user engagement and emotional state.

[0038] Step S15: construct a user dynamic engagement model based on the user behavior perception data, thereby obtaining a user engagement model.

[0039] In this embodiment, based on the integrated user behavior perception data, a machine learning algorithm, such as random forest or support vector machine (SVM), is used to construct a user dynamic engagement model. The model can evaluate the user's engagement in the learning process by integrating multiple factors such as user behavior, emotional state, and interaction frequency. For example, when a user continues to pay attention to a certain teaching module for a long time, the model will consider that the user's engagement is high. The output of the model can help the teaching platform adjust the course content or push strategy in real time to ensure that the content matches the user's emotional state and engagement, thereby improving the learning effect.

[0040] Optionally, step S13 is specifically: Performing facial feature point alignment on the user's facial expression image to obtain a facial feature point aligned image; In this embodiment, the facial expression image of the user is subjected to facial feature point alignment processing, specifically by using a facial alignment algorithm, such as a 68-point detector in Dlib, to identify and correct the positions of facial feature points in the image, ensuring that the facial area in the image is standardized and aligned. After such processing, the facial area in the image will become consistent, reducing the error caused by angle or position changes, thereby making subsequent facial expression analysis more accurate.

[0041] Perform facial feature point combination extraction on the facial feature point aligned image to obtain facial expression feature data; In this embodiment, facial feature point combination extraction is performed. By combining the positions of facial feature points, a feature extraction algorithm (such as principal component analysis PCA or convolutional neural network CNN) is used to extract facial expression feature data. These features include dynamic changes in areas such as the mouth, eyes, and eyebrows, which can reflect the user's emotional state. For example, by extracting the motion features around the mouth and eyes, it can be determined whether the user is smiling, frowning, or surprised.

[0042] Classify the facial expression and emotional state according to the facial expression feature data, so as to obtain the user's facial emotion data; In this embodiment, based on the extracted facial expression feature data, a deep learning model (such as CNN or LSTM) is used to classify its emotional state. The construction process of the facial emotion classification model includes collecting a large amount of facial expression image data and preprocessing it, such as extracting and aligning facial feature points. Then, a convolutional neural network (CNN) is used to train these image data to classify and identify different emotional states, such as joy, anger, surprise, etc. During the training process, a large amount of labeled data is used to gradually optimize the network parameters to improve the accuracy of the model in the emotion recognition task. Through the pre-trained emotion classification model, the user's facial emotional state, such as happiness, anger, sadness, surprise or neutrality, can be identified. This emotion classification step can provide an important emotional basis for subsequent emotion analysis and user feedback.

[0043] Performing audio data framing on the user audio data to obtain a user sound frame set, and performing Mel-frequency cepstral coefficient sound feature extraction on the user sound frame set to obtain user sound frame feature data; In this embodiment, the audio data of the user is framed. Specifically, the audio signal is divided into frames of fixed length (such as 20ms per frame), and each frame contains the time characteristics of the audio signal. This framing method helps analyze the short-term features in the audio data and capture the emotional fluctuations of the user during the speech. After obtaining each frame of audio data, the Mel-frequency cepstral coefficient (MFCC) feature is further extracted. MFCC is a feature widely used in speech recognition, which can effectively represent the spectral characteristics of the audio signal, thereby helping to capture the emotional clues in the speech. By calculating the MFCC features of each frame, the emotional characteristics of the user in the speech can be analyzed.

[0044] Audio emotion state classification is performed based on the user voice frame feature data to obtain user voice emotion data.

[0045] In this embodiment, the construction process of the audio sentiment analysis model begins with the collection and processing of audio data. First, the audio data is divided into frames and sound features such as Mel-frequency cepstral coefficients (MFCC) are extracted. Next, these features are trained using a long short-term memory network (LSTM) to learn the emotional information hidden in the audio signal. Through training, the model can identify the emotional state (such as happiness, sadness, anger, etc.) in the audio, and optimize the network parameters through back propagation to achieve a high accuracy rate of sentiment classification. Based on the extracted sound frame feature data, the emotional state of each audio frame is classified using a trained sentiment analysis model (such as long short-term memory LSTM or deep neural network DNN). Through these sentiment classification models, the user's voice emotional state, such as pleasure, tension, anxiety, excitement, etc., can be judged. This step can accurately analyze the user's emotions reflected in the audio, thereby providing complete data support for the overall sentiment analysis.

[0046] Optionally, step S14 is specifically: Step S141: estimating the user's limb state based on the user's motion data, thereby obtaining the user's limb state data; In this embodiment, a variety of sensors (such as accelerometers, gyroscopes) and cameras are used to collect the user's motion data in real time, and a deep learning model, such as a convolutional neural network (CNN) or a long short-term memory network (LSTM), is further used to process these data to estimate the user's limb state. The data captured by these devices, such as posture and motion trajectory, are used to estimate whether the user is in a static state, walking, standing, or other dynamic postures. For example, the bracelet worn can estimate whether the user is moving through acceleration data, and the full-body image of the user obtained by the camera can further identify its specific motion posture.

[0047] Step S142: performing limb state time series analysis according to the user's limb state data, thereby obtaining limb state time series data; In this embodiment, a time series analysis algorithm (such as a time series convolutional network (TCN) or LSTM) is applied to perform time series modeling on the limb state data. The dynamic data collected by the user from the sensor is arranged in chronological order, and the changing trend of the limb state is analyzed through the time series model. The model can reveal how the user's limb state changes over time, and identify periodic (such as gait or sitting) and non-periodic (such as irregular hand movements or posture adjustments) limb changes. The result of this process is a set of time series data that can reflect the evolution of the user's movements.

[0048] Step S143: performing time window limb state change statistics on the limb state time series data according to a preset time window, thereby obtaining periodic limb state change data and non-periodic limb state change data; In this embodiment, the continuous limb state data is segmented and analyzed according to the set time window. Each time window will capture the changes in the user's limb state, and then perform statistical analysis. Using, for example, a sliding window algorithm, the data in each time period is classified as periodic and non-periodic. For example, if a user maintains a standing posture for a long time, the limb state change data during this period will be classified as a periodic change; and the user's rapid arm swinging action will be regarded as a non-periodic change. In this way, different types of limb state changes can be clearly distinguished and quantified.

[0049] Step S144: classifying the limb state change frequency according to the non-periodic limb state change data, thereby obtaining non-periodic high-frequency limb state change data and non-periodic low-frequency limb state change data; In this embodiment, when analyzing the non-periodic limb state change data, the data is first classified based on the action frequency analysis tool. The frequency of the action is analyzed using methods such as Fast Fourier Transform (FFT) or Wavelet Transform. The frequency of the action can reflect the tension of the limbs. For example, high-frequency hand waving and trembling usually indicate that the user is in an anxious or nervous state. Low-frequency actions indicate that the user's state is relatively stable, such as a gentle pace or slow body movement.

[0050] Step S145: performing a user gentle motion emotion analysis on the periodic limb state change data and the non-periodic low-frequency limb state change data, thereby obtaining the user gentle motion emotion data; performing a user tense motion emotion analysis on the non-periodic high-frequency limb state change data, thereby obtaining the user tense motion emotion data; In this embodiment, the sentiment analysis of slow movements involves the classification of low-frequency limb state change data, and uses a sentiment recognition model (such as one based on a support vector machine (SVM) or a neural network model) to analyze the emotions of the user's movements. For example, when the user's movements are slow and even, it indicates that the user is in a relaxed or calm state. When processing non-periodic high-frequency limb state changes, a sentiment analysis model (such as one based on a long short-term memory network (LSTM) to analyze the temporal changes of movements) is applied to identify emotional reactions such as tension and anxiety. These emotional reactions are reflected by high-frequency movement features, such as rapid body shaking or leg trembling, which are usually associated with anxiety or tension and irritability.

[0051] Step S146: performing user action emotion pattern recognition based on the user's smooth action emotion data and the user's tense action emotion data, thereby obtaining user action emotion data; In this embodiment, the emotion pattern recognition technology is applied to combine the user's smooth action emotion data and tense action emotion data, using random forest, K-means clustering or more complex deep neural network models (such as multi-layer perceptron (MLP)) for pattern recognition. Through these technologies, the user's action patterns are classified into different emotion patterns. For example, smooth and slow movements will be classified as calm emotion patterns, while fast and violent movements may be labeled as anxious or tense emotion patterns. This step enables the system to identify the user's current emotional state based on the characteristics of the body movements.

[0052] Step S147: Integrate the user action emotion data, the user facial emotion data, and the user voice emotion data into behavioral features to obtain user behavior perception data.

[0053] In this embodiment, a multimodal emotion analysis method is used to integrate the action emotion data, facial expression data and voice emotion data from the user. Through a data fusion algorithm (such as weighted average, principal component analysis (PCA) or a deep learning model, such as a multimodal deep neural network), data from different sources are merged to generate comprehensive user behavior perception data. This comprehensive data reflects the overall emotional state of the user. For example, by analyzing the user's movements, facial expressions and voices, it can be accurately assessed whether the user is in an emotional state such as relaxation, tension, happiness or sadness. This process helps to provide more personalized emotional feedback and teaching content recommendations.

[0054] Optionally, step S2 specifically includes: Step S21: Obtaining a course content library through an online teaching platform; In this embodiment, the course content library in the platform is obtained in real time by connecting the online teaching platform with the teaching content management system. The content library includes a variety of resources such as course outlines, teaching videos, learning materials, and after-class tests, and these data are stored in the cloud database. The platform interacts with the database through an API interface and obtains the required course information through query and indexing. For example, a specific theoretical course can be extracted and displayed to the user, and the course content can be classified by topic.

[0055] Step S22: locating the user's learning course based on the real-time user behavior data to be analyzed according to the course content library, thereby obtaining the user's learning course data; In this embodiment, the information in the course content library is matched with the user's real-time behavior data. Specifically, by analyzing the user's behavior data on the platform (such as click history, viewing time, participation in discussions, etc.), the course module the user is studying can be identified, and his learning progress and interests can be accurately inferred. For example, if a user is frequently accessing content related to "mathematics", other course resources related to "mathematics" will be automatically recommended to the user.

[0056] Step S23: evaluating the user course learning data for user course learning participation through a user participation model, thereby obtaining user course participation data; In this embodiment, the user engagement model is used to evaluate the engagement of each user's learning course. The engagement is measured based on the depth of the user's interaction with the course (such as viewing time, test completion, comment engagement, etc.). The engagement model is trained using machine learning methods (such as support vector machines, decision trees, etc.) to determine which courses can effectively attract users by learning a large amount of user data. For example, if a user completes all chapters in a course and submits homework, the system will assess him as a highly engaged user.

[0057] Step S24: removing low-participation courses from the user's course data according to the user's course participation data, thereby obtaining the user's actual course data; In this embodiment, low-engagement courses are automatically filtered out based on the user's course participation data. Low-engagement courses generally refer to courses that users have not watched in full or have not completed relevant assessments. Specifically, a threshold is set. If the viewing time of a course is lower than a certain proportion or the user has not conducted any interaction (such as comments, tests), the course will be removed from the user's actual learning course data. For example, if a user only stays for 5 minutes during the course viewing process and does not complete any tests, the course will be marked as a low-engagement course and excluded.

[0058] Step S25: Integrate course features according to the user's actual course learning data, so as to obtain the user's actual learning content feature descriptor.

[0059] In this embodiment, based on the user's actual learning course data, these data are integrated with features. This process uses data fusion technology to integrate the user's learning activity data (such as learning time, test scores, interaction frequency, etc.) with the course content features. For example, the user's mastery of a certain module is calculated, and the user's participation in discussions, answering questions, and other behaviors are integrated to generate a "learning content feature descriptor." These descriptors can include the characteristics of the course content, the user's depth of understanding of a certain course module, points of interest, etc., to facilitate personalized recommendations of course content that better meets user needs.

[0060] Optionally, step S22 is specifically: Step S221: obtaining a teaching platform page interaction log through the online teaching platform; In this embodiment, the teaching platform page interaction log is obtained through the online teaching platform. Specifically, the platform records the user's clicks, dwell time, page jumps and other interactive behavior data on each page, and these data are uploaded to the platform's backend in real time in the form of web logs or event logs. The platform uses log analysis tools such as ELK Stack (Elasticsearch, Logstash, Kibana) to collect and manage these logs, thereby ensuring that the acquired interactive data includes information such as browsing of course pages and playback of course videos. For example, after a user clicks on a course page, the system will record his or her click behavior and page dwell time to ensure the accuracy and timeliness of the data.

[0061] Step S222: classifying the user behavior collection type of the real-time user behavior data to be analyzed, thereby obtaining the user behavior continuously collected data and the user behavior instantaneous collected data; In this embodiment, the user behavior collection type is classified for the real-time user behavior data to be analyzed, and the data is divided into continuous collection data and instantaneous collection data. Continuously collected data includes users' long-term browsing and interactive behaviors on the platform, such as watching course videos, participating in online discussions, etc., while instantaneous collection data includes one-time behaviors such as users clicking a specific button and submitting homework. The system automatically identifies and classifies data based on the duration of the behavior and the type of interaction, using algorithms (such as K-means clustering) to ensure the accuracy and real-time nature of data collection. For example, each time a user clicks to enter a course is considered instantaneous collection data, and the length of time they watch videos is counted as continuously collected data.

[0062] Step S223: time matching the continuously collected data of user behavior and the teaching platform page interaction log, thereby obtaining the continuously interactive teaching page data; time matching the instantaneously collected data of user behavior and the teaching platform page interaction log, thereby obtaining the instantaneous interactive teaching page data; In this embodiment, the continuously collected data of user behavior is time-matched with the teaching platform page interaction log to ensure the consistency of the two data sources. The time matching process synchronizes the behavior data between different data sources through timestamps. For example, if a user clicks to enter a course at 10:05, the timestamp of the click event is associated with the user's behavior data to generate continuous interactive teaching page data. This enables the platform to obtain the specific behavior records of each user at different time nodes, and then analyze their learning process.

[0063] Step S224: extracting teaching page teaching course features from the continuous interactive teaching page data, thereby obtaining continuous interactive teaching course data; extracting teaching page teaching course features from the instantaneous interactive teaching page data, thereby obtaining instantaneous interactive teaching course data; In this embodiment, teaching course feature extraction is performed on the continuous interactive teaching page data and the instantaneous interactive teaching page data. For the continuous interactive data, the features such as the user's stay time on each page, the interaction frequency, and the course completion degree are extracted; and for the instantaneous interactive data, the user's single operation, such as the number of clicks, homework submission, etc., is extracted. A feature extraction algorithm (such as PCA principal component analysis) is used to refine the key features of these data. For example, if a user stays on a course page for more than 10 minutes and watches video clips multiple times, high-interaction feature data for the course will be extracted.

[0064] Step S225: performing a teaching course intersection operation on the continuous interactive teaching course data and the instantaneous interactive teaching course data, thereby obtaining user interactive teaching course data; In this embodiment, the teaching course intersection operation is performed on the continuous interactive teaching course data and the instantaneous interactive teaching course data. The intersection operation ensures that only courses in which users show interactive behaviors in both data types are counted as interactive teaching courses. For example, if a user continuously browses a course page in multiple time periods and submits homework in a short period of time, it will be considered that the user has a high frequency of interaction with the course, ensuring that the interactive teaching course data finally obtained can fully reflect the user's learning behavior.

[0065] Step S226: Locate the teaching track of the user interactive teaching course data through the course content library, so as to obtain the user learning course data.

[0066] In this embodiment, the teaching course teaching track is located by the course content library for the user's interactive teaching course data. Specifically, the user's learning track is analyzed based on the user's interactive data, and the user's learning path is modeled in combination with the course sequence and structure in the course content library. For example, the course module that the user has learned is identified based on the course data of the user's interaction, and the next learning content is automatically recommended based on the structure in the course content library. In this way, not only can the user's learning progress be tracked in real time, but also personalized learning suggestions can be provided to improve the learning effect.

[0067] Optionally, step S226 is specifically: The course teaching trajectory is divided into course teaching trajectory datasets based on the course content library. In this embodiment, the course teaching trajectory of the user interaction teaching course data is divided through the course content library. Specifically, during the learning process of the user, the system divides the learning trajectory according to the interaction data (such as behaviors like course clicks, video views, homework submissions, etc.), and by analyzing the structure and sequence of the course, it corresponds the user's behavior with the course content. For example, if the user completes the first four courses when learning a certain module and stays on the page of the last module, the learning trajectory of this user will be defined as "completing the first four modules and currently learning the last module", so that the learning progress of the user can be inferred according to the teaching path of the course.

[0068] Extract the background audio from the user audio data to obtain the background audio data, and perform audio data framing on the background audio data to obtain the background sound frames; In this embodiment, the background audio is extracted from the user audio data. This process involves using efficient background noise cancellation algorithms (such as Wiener filters or Spectral Subtraction) to separate the background audio from the user's recording. For example, if the user uses voice input or listens to the explanation recording during the learning process, the background noise will be automatically removed, and only the audio related to the course content will be extracted. After that, the audio data will be framed, and the audio signal will be divided into multiple frames according to a certain time window (such as a frame length of 25 ms and a frame shift of 10 ms) for subsequent speech processing and analysis.

[0069] Perform speech recognition on the background sound frames to obtain the background sound text data, and perform stop word removal on the background sound text data to obtain the background text data to be matched; In this embodiment, speech recognition is performed on the background sound frames, and existing automatic speech recognition (ASR) technology is used to convert the audio data into text data. Taking the Google Speech-to-Text API as an example, the background sound frames are input into the recognition model one by one, and each word in the audio is recognized and converted into text. These text data contain all the speech information of the user when listening to the explanation, including the content explained by the narrator and other background sounds in the environment.

[0070] Perform keyword matching of the course content on the background text data to be matched through the course content library to obtain the audio-matched teaching content data; In this embodiment, stop word removal is performed on the background sound text data. Specifically, using natural language processing (NLP) technology, through a pre-set stop word list, common meaningless words in the text (such as "de", "shi", etc.) are automatically removed, which can reduce interference and improve the accuracy of subsequent matching. Taking Chinese teaching as an example, after removing the common Chinese stop words, the remaining text content is used for keyword matching of the course content.

[0071] According to the course teaching trajectory dataset, the audio is matched with the teaching content data to identify the user's learning course, so as to obtain the user's learning course data.

[0072] In this embodiment, the course content keyword matching is performed on the background text data to be matched through the course content library. All course materials contained in the course content library are used, and the keywords in the text are identified and matched through algorithms (such as TF-IDF, Word2Vec). For example, when a word such as "Pythagorean Theorem" appears in the background text, the relevant courses in the course content library will be searched, and it will be confirmed whether the course contains relevant content. This process ensures that the audio content heard by the user is consistent with the actual content of the teaching course, and provides a basis for subsequent personalized learning recommendations.

[0073] Optionally, step S3 specifically includes: Step S31: Calculate the course content similarity of the user's actual learning content feature descriptor and the course content library, thereby obtaining course content similarity data; In this embodiment, the course content similarity is calculated for the feature descriptor of the user's actual learning content and the course content library. By using an algorithm based on semantic similarity (such as cosine similarity or BERT model), the similarity between the user's learning content and each course in the course library is calculated. Specifically, if the user has learned the "Linear Algebra" course content, and there is an "Advanced Mathematics" course related to this topic in the course library, the system will compare the semantic similarity of the two and obtain a similarity score. In this way, the relevance of the courses that the user has learned and other course content can be identified, and relevant learning content can be further accurately recommended.

[0074] Step S32: extracting courses with high similarity in course content from the course content library based on the course content similarity data, thereby obtaining course data associated with the user's actual learning content; In this embodiment, courses with high similarity in course content are extracted from the course content library based on the course content similarity data. According to the aforementioned similarity score, courses with high similarity to the user's actual learning content are extracted from the course content library. For example, if it is detected that the user has a high learning participation in "a certain calculation process", and the related "advanced mathematics" course has a high similarity, this course will be recommended to the user to continue to expand its learning content and ensure that the user's learning progress and interests can be met.

[0075] Step S33: classifying the course data associated with the user's actual learning content into course priorities according to the course teaching trajectory data set, thereby obtaining associated course priority data; In this embodiment, the course data associated with the user's actual learning content is prioritized according to the course teaching trajectory data set. Specifically, according to the user's past learning behavior and trajectory, for example, if the user has completed the first two modules of "Economics" and is stuck at the progress of the last module, which courses should be recommended first will be determined based on the progress. For example, if the user has just completed the basic theory part of the study, courses that help the user consolidate and expand existing knowledge will be recommended first, thereby improving the coherence and depth of learning.

[0076] Step S34: performing an associated course complexity evaluation on the associated course data of the user's actual learning content, thereby obtaining associated course complexity data; In this embodiment, the complexity of the associated courses is evaluated for the course data associated with the user's actual learning content. The complexity of the associated courses is evaluated based on the difficulty level of the course content displayed on the platform, the user's historical learning performance, and the preset calibration of the course difficulty. For example, for a user who has completed the "Introduction to Education" course and mastered the basic content, the next course will be evaluated as "Advanced Economics", which is a higher complexity course. Through complexity evaluation, avoid recommending courses that are too simple or too difficult, and ensure that the course recommendation is in line with the user's learning ability.

[0077] Step S35: Plan a course content recommendation strategy based on the associated course priority data and the associated course complexity data, thereby obtaining a user course content recommendation strategy, and upload it to the online teaching platform to perform the content recommendation task.

[0078] In this embodiment, the course content recommendation strategy is planned based on the associated course priority data and the associated course complexity data, and the recommendation strategy is uploaded to the online teaching platform. Through the analysis of comprehensive priority and complexity, a personalized learning path and recommendation strategy will be formulated. If a course content has a high priority for the user's learning goal and is suitable for the current stage in terms of complexity, this course will be recommended to the user. Ultimately, all recommendation strategies will be automatically executed by the platform, and the most suitable learning content will be recommended to help users get the best support and learning experience during the learning process.

[0079] Optionally, step S4 specifically includes: Step S41: acquiring real-time user platform interaction data through the online teaching platform, and performing data preprocessing on the real-time user platform interaction data, thereby obtaining the real-time user platform interaction data to be analyzed; In this embodiment, real-time user platform interaction data is obtained through the online teaching platform, and these data are pre-processed. Specifically, various interactive behavior data of users on the platform are collected, including page clicks, course browsing, comments, likes, etc., and noise removal and missing value filling are performed to ensure the integrity and accuracy of the data. For example, when a user visits a course page, the user's browsing time and click behavior are recorded, and duplicate or irrelevant access records are removed to ensure the data quality of subsequent analysis.

[0080] Step S42: extracting interaction features from the real-time user platform interaction data to be analyzed, thereby obtaining user platform browsing time data and user platform comment content data; In this embodiment, the real-time user platform interaction data to be analyzed is subjected to interaction feature extraction to obtain user platform browsing time data and user platform comment content data. Based on the user's behavior, their stay time on each page is extracted, the user's overall participation in a course is calculated, and the keywords in the comment content are analyzed to mark the sentiment polarity of each comment and other information. For example, if a user stays on the "Linear Algebra" course page for a long time and the comments mention keywords such as "easy to understand" and "deepen understanding", this information will be recorded as the user's interaction feature data.

[0081] Step S43: performing comment text sentiment polarity analysis based on the user platform comment content data, thereby obtaining user platform positive comment content data and user platform negative comment content data; In this embodiment, the sentiment polarity analysis of the comment text is performed based on the user platform comment content data to obtain the positive comment content data and negative comment content data of the user platform. The comment text is analyzed using natural language processing technology (such as sentiment dictionary or deep learning model) to classify the comments as positive or negative. For example, if a user comment contains positive expressions such as "this course is very inspiring" or "the learning effect is significant", it will be classified as a positive comment; if the comment contains words such as "the content is boring" or "not practical", it will be classified as a negative comment.

[0082] Step S44: classifying the platform browsing time according to the user platform browsing time data, thereby obtaining the user's continuous browsing content data and the user's instantaneous browsing content data; In this embodiment, platform browsing time is classified according to the user platform browsing time data, so as to obtain the user's continuous browsing content data and the user's instant browsing content data. By analyzing the user's browsing time, the browsing behavior is divided into continuous browsing and short-time browsing. For example, if the user stays on a course page for more than five minutes and frequently clicks on related content, it is considered that the user is continuously browsing the course; if the browsing time is less than two minutes and only a few pages are browsed, it is considered to be instant browsing content.

[0083] Step S45: performing content-related sentiment analysis on the user platform's positive comment content data and the user's continuous browsing content data, thereby obtaining the user's positive feedback sentiment data; performing content-related sentiment analysis on the user platform's negative comment content data and the user's instant browsing content data, thereby obtaining the user's negative feedback sentiment data; In this embodiment, the content-related sentiment analysis is performed on the user platform's positive comment content data and the user's continuous browsing content data to obtain the user's positive feedback sentiment data; at the same time, the content-related sentiment analysis is performed on the user platform's negative comment content data and the user's instant browsing content data to obtain the user's negative feedback sentiment data. By analyzing the user's comments and browsing behavior on the course, the user's emotional tendency towards a certain course can be judged. For example, if the user's comments are positive and he browses a certain course for a long time, the user's emotional feedback on the course will be judged as positive; on the contrary, if the comments are negative and only browsed for a short time, the system will evaluate it as negative feedback sentiment.

[0084] Step S46: merging the user's positive feedback emotion data and the user's negative feedback emotion data to obtain the user's feedback emotion data; In this embodiment, the user's positive feedback sentiment data and negative feedback sentiment data are merged to obtain user feedback sentiment data. At this time, the sentiment data from different sources, including user comments and browsing behaviors, are integrated, and a comprehensive sentiment score is obtained by weighted average. For example, if the user has both positive comments and long browsing behaviors, a high sentiment score will be obtained; conversely, if there are only negative comments and short browsing behaviors, a lower sentiment score will be evaluated.

[0085] Step S47: Evaluate user course acceptance based on user feedback sentiment data, thereby obtaining user course acceptance data.

[0086] In this embodiment, the user course acceptance evaluation is performed based on the user feedback sentiment data, thereby obtaining the user course acceptance data. Using the constructed sentiment model, the feedback sentiment data of each user is comprehensively analyzed to evaluate the user's acceptance of a certain course. For example, if the user's emotional feedback on a certain course is relatively positive, the course acceptance will be evaluated as high, otherwise it is considered that the user's acceptance is low.

[0087] Optionally, step S47 is specifically: Step S471: calculating the emotion intensity of the user feedback emotion data, thereby obtaining the user feedback emotion intensity data; In this embodiment, the sentiment intensity of the user feedback sentiment data is calculated. Specifically, sentiment analysis is performed on the user's comment text and behavior data to extract the sentiment intensity of each feedback, for example, by calculating the intensity value of the positive or negative sentiment in each feedback through a sentiment dictionary or a deep learning model. For example, when a user comments that "the course content is very detailed and I gained a lot after learning", the sentiment intensity value is 0.85, which means it is very positive; while "the course is too boring and I didn't gain anything after listening to it" gets a sentiment intensity value of -0.75, which means it is relatively negative. In this way, the sentiment intensity of each feedback is quantified to provide data support for subsequent steps.

[0088] Step S472: Calculate the emotional polarity ratio according to the emotional intensity data fed back by the user, thereby obtaining emotional polarity ratio data; In this embodiment, the emotional polarity ratio is calculated based on the emotional intensity data of user feedback. Specifically, the emotional intensity values ​​in the user feedback are divided into positive and negative categories, and the proportion of each type of emotion is calculated. For example, if there are 50 positive feedbacks, 30 negative feedbacks, and 20 neutral feedbacks among 100 user feedbacks, the positive emotion ratio will be calculated to be 50%, and the negative emotion ratio will be 30%. This ratio data will help further evaluate the emotional tendency and provide a basis for scoring the subsequent recommended content.

[0089] Step S473: marking the user feedback emotion data with emotion polarity proportions based on the emotion polarity proportions data, thereby obtaining positive feedback emotion polarity proportions data and negative feedback emotion polarity proportions data; In this embodiment, the emotion polarity ratio of user feedback emotion data is marked based on the emotion polarity ratio data, and the positive feedback emotion polarity ratio data and the negative feedback emotion polarity ratio data are obtained. Specifically, the emotion classification is performed on each user's feedback, and the proportion of positive and negative emotions in each feedback is marked. For example, if the positive emotion accounts for 60% and the negative emotion accounts for 40% in a user's comments, the emotion polarity ratio data of the user will be marked as 60% positive and 40% negative. Through this marking, the system can understand the user's emotional attitude more accurately and make more refined recommendations.

[0090] Step S474: Perform sentiment weighted scoring on the recommended content based on the positive feedback sentiment polarity ratio data and the negative feedback sentiment polarity ratio data, thereby obtaining user course acceptance data.

[0091] In this embodiment, the emotional weighted scoring of the recommended content is performed according to the positive feedback emotional polarity ratio data and the negative feedback emotional polarity ratio data to obtain the user course acceptance data. Specifically, the emotional weighting of the recommended content of different courses is performed in combination with the user's emotional polarity ratio data. For example, if the positive emotion ratio of a course is 70% and the negative emotion ratio is 30%, the recommendation score of the course will be adjusted according to this ratio. If the positive emotion ratio of the course is high, the recommendation score is high, indicating that the course is more attractive to users; conversely, if the negative feedback ratio is high, the course recommendation score will be low, indicating that the user acceptance of the course is low.

[0092] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0093] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An information sharing method for an online teaching platform based on the Internet of Things, characterized in that: The following steps are involved: Step S1: using the Internet of Things platform to collect real-time user behavior data, and quantifying user behavior perception of the real-time user behavior data, thereby obtaining user behavior perception data; Build a user dynamic engagement model based on user behavior perception data to obtain a user engagement model; Step S2: Obtain the course content library through the online teaching platform, and locate the user's learning course in the course content library based on the user participation model, so as to obtain the user's actual learning course data; Integrate course features based on the user's actual course data to obtain the user's actual learning content feature descriptor; Step S3: analyzing the courses associated with the actual learning content in the course content library according to the user's actual learning content feature descriptor, thereby obtaining the user's actual learning content associated course data; Based on the course data associated with the user's actual learning content, the course content recommendation strategy is analyzed to obtain the user's course content recommendation strategy, which is then uploaded to the online teaching platform to perform the content recommendation task; Step S4: obtaining real-time user platform interaction data through the online teaching platform, and performing user feedback sentiment analysis on the real-time user platform interaction data through a user engagement model, thereby obtaining user feedback sentiment data; Conduct user course acceptance evaluation based on user feedback sentiment data to obtain user course acceptance data; Step S5: Iteratively optimize the subsequent recommendation strategy for user course content according to the user course acceptance data, so as to obtain the optimized user course content recommendation strategy, and upload it to the online teaching platform to perform the content recommendation task.

2. The information sharing method of the online teaching platform based on the Internet of Things according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: using the Internet of Things to collect real-time user behavior data, and performing data preprocessing on the real-time user behavior data, thereby obtaining real-time user behavior data to be analyzed; Step S12: extracting behavior perception features from the real-time user behavior data to be analyzed, thereby obtaining user facial expression images, user action data, and user audio data; Step S13: performing emotion fluctuation sentiment analysis according to the user's facial expression image and the user's audio data, thereby obtaining the user's facial emotion data and the user's voice emotion data; Step S14: performing user action emotion analysis according to the user action data to obtain user action emotion data, and integrating the user action emotion data, the user facial emotion data, and the user voice emotion data into behavioral features to obtain user behavior perception data; Step S15: construct a user dynamic engagement model based on the user behavior perception data, thereby obtaining a user engagement model.

3. The information sharing method of the online teaching platform based on the Internet of Things according to claim 2 is characterized in that: Step S13 is specifically as follows: Performing facial feature point alignment on the user's facial expression image to obtain a facial feature point aligned image; Perform facial feature point combination extraction on the facial feature point aligned image to obtain facial expression feature data; Classify the facial expression and emotional state according to the facial expression feature data, so as to obtain the user's facial emotion data; Performing audio data framing on the user audio data to obtain a user sound frame set, and performing Mel-frequency cepstral coefficient sound feature extraction on the user sound frame set to obtain user sound frame feature data; Audio emotion state classification is performed based on the user voice frame feature data to obtain user voice emotion data.

4. The information sharing method of the online teaching platform based on the Internet of Things according to claim 2 is characterized in that: Step S14 is specifically as follows: Step S141: estimating the user's limb state based on the user's motion data, thereby obtaining the user's limb state data; Step S142: performing limb state time series analysis according to the user's limb state data, thereby obtaining limb state time series data; Step S143: performing time window limb state change statistics on the limb state time series data according to a preset time window, thereby obtaining periodic limb state change data and non-periodic limb state change data; Step S144: classifying the limb state change frequency according to the non-periodic limb state change data, thereby obtaining non-periodic high-frequency limb state change data and non-periodic low-frequency limb state change data; Step S145: performing user smooth motion emotion analysis on the periodic limb state change data and the non-periodic low-frequency limb state change data, thereby obtaining user smooth motion emotion data; Performing user nervous action emotion analysis on non-periodic high-frequency limb state change data, thereby obtaining user nervous action emotion data; Step S146: performing user action emotion pattern recognition based on the user's smooth action emotion data and the user's tense action emotion data, thereby obtaining user action emotion data; Step S147: Integrate the user action emotion data, the user facial emotion data, and the user voice emotion data into behavioral features to obtain user behavior perception data.

5. The information sharing method of the online teaching platform based on the Internet of Things according to claim 1 is characterized in that: Step S2 is specifically as follows: Step S21: Obtaining a course content library through an online teaching platform; Step S22: locating the user's learning course based on the real-time user behavior data to be analyzed according to the course content library, thereby obtaining the user's learning course data; Step S23: evaluating the user course learning data for user course learning participation through a user participation model, thereby obtaining user course participation data; Step S24: removing low-participation courses from the user's course data according to the user's course participation data, thereby obtaining the user's actual course data; Step S25: Integrate course features according to the user's actual course learning data, so as to obtain the user's actual learning content feature descriptor.

6. The information sharing method of the online teaching platform based on the Internet of Things according to claim 5 is characterized in that: Step S22 is specifically as follows: Step S221: obtaining a teaching platform page interaction log through the online teaching platform; Step S222: classifying the user behavior collection type of the real-time user behavior data to be analyzed, thereby obtaining the user behavior continuously collected data and the user behavior instantaneous collected data; Step S223: time matching the continuously collected data of user behavior and the teaching platform page interaction log, thereby obtaining the continuously interactive teaching page data; time matching the instantaneously collected data of user behavior and the teaching platform page interaction log, thereby obtaining the instantaneous interactive teaching page data; Step S224: extracting teaching page teaching course features from the continuous interactive teaching page data, thereby obtaining continuous interactive teaching course data; extracting teaching page teaching course features from the instantaneous interactive teaching page data, thereby obtaining instantaneous interactive teaching course data; Step S225: performing a teaching course intersection operation on the continuous interactive teaching course data and the instantaneous interactive teaching course data, thereby obtaining user interactive teaching course data; Step S226: Locate the teaching track of the user interactive teaching course data through the course content library, so as to obtain the user learning course data.

7. The information sharing method of the online teaching platform based on the Internet of Things according to claim 6 is characterized in that: Step S226 is specifically as follows: The course teaching trajectory is divided into course teaching trajectory datasets based on the course content library. Performing background audio extraction on the user audio data to obtain background audio data, and performing audio data framing on the background audio data to obtain background sound frames; Performing speech recognition on the background sound frame to obtain background sound text data, and removing stop words from the background sound text data to obtain background text data to be matched; Matching the course content keywords with the background text data to be matched through the course content library, thereby obtaining audio matching teaching content data; According to the course teaching trajectory dataset, the audio is matched with the teaching content data to identify the user's learning course, so as to obtain the user's learning course data.

8. The information sharing method of the online teaching platform based on the Internet of Things according to claim 1 is characterized in that: Step S3 is specifically as follows: Step S31: Calculate the course content similarity of the user's actual learning content feature descriptor and the course content library, thereby obtaining course content similarity data; Step S32: extracting courses with high similarity in course content from the course content library based on the course content similarity data, thereby obtaining course data associated with the user's actual learning content; Step S33: classifying the course data associated with the user's actual learning content into course priorities according to the course teaching trajectory data set, thereby obtaining associated course priority data; Step S34: performing an associated course complexity evaluation on the associated course data of the user's actual learning content, thereby obtaining associated course complexity data; Step S35: Plan a course content recommendation strategy based on the associated course priority data and the associated course complexity data, thereby obtaining a user course content recommendation strategy, and upload it to the online teaching platform to perform the content recommendation task.

9. The information sharing method of the online teaching platform based on the Internet of Things according to claim 1 is characterized in that: Step S4 is specifically as follows: Step S41: acquiring real-time user platform interaction data through the online teaching platform, and performing data preprocessing on the real-time user platform interaction data, thereby obtaining the real-time user platform interaction data to be analyzed; Step S42: extracting interaction features from the real-time user platform interaction data to be analyzed, thereby obtaining user platform browsing time data and user platform comment content data; Step S43: performing comment text sentiment polarity analysis based on the user platform comment content data, thereby obtaining user platform positive comment content data and user platform negative comment content data; Step S44: classifying the platform browsing time according to the user platform browsing time data, thereby obtaining the user's continuous browsing content data and the user's instantaneous browsing content data; Step S45: performing content-related sentiment analysis on the user platform's positive comment content data and the user's continuous browsing content data, thereby obtaining user's positive feedback sentiment data; Conduct content-related sentiment analysis on the user platform's negative comment content data and the user's instant browsing content data to obtain the user's negative feedback sentiment data; Step S46: merging the user's positive feedback emotion data and the user's negative feedback emotion data to obtain the user's feedback emotion data; Step S47: Evaluate user course acceptance based on user feedback sentiment data, thereby obtaining user course acceptance data.

10. The information sharing method of the online teaching platform based on the Internet of Things according to claim 9 is characterized in that: Step S47 is specifically as follows: Step S471: calculating the emotion intensity of the user feedback emotion data, thereby obtaining the user feedback emotion intensity data; Step S472: Calculate the emotional polarity ratio according to the emotional intensity data fed back by the user, thereby obtaining emotional polarity ratio data; Step S473: marking the user feedback emotion data with emotion polarity proportions based on the emotion polarity proportions data, thereby obtaining positive feedback emotion polarity proportions data and negative feedback emotion polarity proportions data; Step S474: Perform sentiment weighted scoring on the recommended content based on the positive feedback sentiment polarity ratio data and the negative feedback sentiment polarity ratio data, thereby obtaining user course acceptance data.

Images