Intelligent teaching analysis system and method based on big data

Abstract

The present application provides an intelligent teaching analysis system and method based on big data, wherein the method includes: obtaining students' real-time physiological monitoring data and behavioral monitoring data during the class; performing instant pattern recognition on the real-time physiological monitoring data and behavioral monitoring data respectively to generate physiological state indicator sequences and action feature vector sets; performing multimodal time series correlation analysis on the physiological state indicator sequences and action feature vector sets to generate dynamic learning state maps; based on the dynamic learning state maps, generating intelligent teaching analysis results, the intelligent teaching analysis results include content adjustment instructions and interactive guidance strategies for the current teaching stage. The present application improves the dynamic adaptability of classroom teaching strategies and student participation.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to an intelligent teaching analysis system and method based on big data. Background Art

[0002] With the development of smart education, classroom teaching requires real-time monitoring of students' attention and mood changes, allowing for dynamic adjustments to teaching strategies to improve learning outcomes. Traditional methods, which rely on subjective teacher observation, struggle to accurately grasp the real-time status of all students. Multimodal data fusion and analysis technologies are urgently needed to enable automated and refined assessment of student learning status and generate personalized teaching intervention plans.

[0003] Existing solutions use video analysis technology to collect students' facial expressions and body movements through classroom cameras. These technologies then use pre-trained behavioral recognition models to determine their concentration levels. Combined with a pre-set teaching strategy library, these solutions generate simple classroom interaction suggestions. This solution achieves basic attention monitoring through the analysis of behavioral data from a single modality.

[0004] This solution relies solely on visual behavioral data for analysis, ignoring the reflection of physiological indicators such as heart rate and blood oxygen on emotions and cognitive states, resulting in a single dimension of state assessment; at the same time, due to the lack of temporal correlation analysis of multimodal data, it is difficult to capture the intrinsic correlation between attention fluctuations and emotional changes, and the output teaching suggestions deviate from the actual learning state. Summary of the Invention

[0005] This application provides an intelligent teaching analysis system and method based on big data to solve the problems of poor dynamic adaptability of classroom teaching strategies and low student participation in the existing technology.

[0006] In a first aspect, the present application provides an intelligent teaching analysis method based on big data, comprising:

[0007] During the class, obtain students' real-time physiological and behavioral monitoring data;

[0008] Performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data to generate a physiological state indicator sequence and an action feature vector set;

[0009] Performing multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state map;

[0010] Based on the dynamic learning state map, an intelligent teaching analysis result is generated, and the intelligent teaching analysis result includes content adjustment instructions and interactive guidance strategies for the current teaching stage.

[0011] Optionally, performing multimodal temporal correlation analysis on the physiological state indicator sequence and the motion feature vector set to generate a dynamic learning state map includes:

[0012] Establishing a cross-modal mapping relationship in a time dimension according to the physiological state indicator sequence and the motion feature vector set;

[0013] A dynamic learning state map is generated according to the correlation strength between the changes in the physiological state indicators and the changes in the motion feature vectors in the cross-modal mapping relationship.

[0014] Optionally, generating a dynamic learning state map according to the correlation strength between the change of the physiological state indicator and the change of the motion feature vector in the cross-modal mapping relationship includes:

[0015] Calculate the correlation strength between the amplitude change rate used to reflect the change of the physiological state indicator and the direction dimension change amount used to reflect the change of the motion feature vector in the direction dimension to generate a correlation strength value;

[0016] marking the first physiological signal segment whose association strength value exceeds the set association strength threshold as an attention fluctuation node;

[0017] The second physiological signal segment in which the amplitude dimension variation exceeds the preset fluctuation range is marked as an emotion change node, and the amplitude dimension variation is used to reflect the degree of variation of the action feature vector in the amplitude dimension;

[0018] A dynamic learning state graph is constructed based on the attention fluctuation nodes and the emotion change nodes.

[0019] Optionally, constructing a dynamic learning state graph according to the attention fluctuation node and the emotion change node includes:

[0020] Performing overlapping area detection on the start time of the attention fluctuation node and the duration of the emotion change node to generate a spatiotemporal interaction coefficient between the attention fluctuation node and the emotion change node;

[0021] generating, based on the node pairs whose spatiotemporal interaction coefficients exceed a preset coupling threshold, a graph connection line reflecting the correlation between attention fluctuations and emotion changes;

[0022] Counting the number of interactions of the graph connection lines within a preset time window to generate a topological structure;

[0023] The attention fluctuation nodes, the emotion change nodes, the graph connection lines and the topological structure are integrated to construct a dynamic learning state graph.

[0024] Optionally, generating intelligent teaching analysis results based on the dynamic learning state graph includes:

[0025] Determining a pattern evolution segment corresponding to a current teaching stage according to the dynamic learning state map;

[0026] Dynamically matching the pattern evolution fragments with the teaching scenario templates in the preset teaching strategy library;

[0027] When the matching result is multiple teaching scene templates, content adjustment instructions and interactive guidance strategies are generated according to the priority weights of different teaching scene templates in the matching result;

[0028] The content adjustment instruction and the interactive guidance strategy are combined to generate an intelligent teaching analysis result.

[0029] Optionally, generating content adjustment instructions and interactive guidance strategies according to the priority weights of different teaching scenario templates in the matching results includes:

[0030] The first priority weight is given to the teaching scenario templates whose attention fluctuation frequency exceeds the preset attention fluctuation threshold range and whose emotion change amplitude is within the preset emotion fluctuation tolerance range in the matching results;

[0031] The second priority weight is assigned to the teaching scenario templates whose emotion change amplitude exceeds the preset emotion fluctuation tolerance range and whose attention fluctuation frequency is within the preset attention fluctuation threshold range in the matching results;

[0032] The third priority weight is assigned to the teaching scenario templates whose emotion change amplitude exceeds the preset emotion fluctuation tolerance range and whose attention fluctuation frequency exceeds the preset attention fluctuation threshold range in the matching results;

[0033] A content adjustment instruction and an interaction guidance strategy are generated according to the ranking results of the first priority weight, the second priority weight, and the third priority weight.

[0034] Optionally, the performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data to generate a physiological state indicator sequence and an action feature vector set includes:

[0035] Segmenting the real-time physiological monitoring data into a plurality of physiological signal segments according to a preset time window;

[0036] identifying physiological state indicators associated with attention fluctuations from each physiological signal segment;

[0037] Arranging all the physiological state indicators in chronological order into a physiological state indicator sequence;

[0038] Splitting the limb movements captured in the behavior monitoring data into multiple action units;

[0039] Identifying action feature vectors associated with emotion changes from each action unit;

[0040] All the action feature vectors are combined into an action feature vector set.

[0041] In a second aspect, the present application provides an intelligent teaching analysis system based on big data, comprising:

[0042] The acquisition module is used to obtain students' real-time physiological monitoring data and behavioral monitoring data during the class;

[0043] an identification module for performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data, respectively, to generate a physiological state indicator sequence and an action feature vector set;

[0044] An analysis module, configured to perform multimodal temporal correlation analysis on the physiological state indicator sequence and the motion feature vector set to generate a dynamic learning state map;

[0045] A generation module is used to generate intelligent teaching analysis results based on the dynamic learning state map, and the intelligent teaching analysis results include content adjustment instructions and interactive guidance strategies for the current teaching stage.

[0046] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the big data-based intelligent teaching analysis methods described in the first aspect.

[0047] In a fourth aspect, the present application provides a computer storage medium having computer program instructions stored thereon, which, when executed by a processor, implements an intelligent teaching analysis method based on big data as described in any one of the first aspects.

[0048] In the present application, a big data-based intelligent teaching analysis method is provided, which includes: obtaining students' real-time physiological monitoring data and behavioral monitoring data during class; performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data, respectively, to generate a physiological state indicator sequence and a motion feature vector set; performing multimodal time series correlation analysis on the physiological state indicator sequence and the motion feature vector set to generate a dynamic learning state map; based on the dynamic learning state map, generating intelligent teaching analysis results, the intelligent teaching analysis results including content adjustment instructions and interactive guidance strategies for the current teaching stage.

[0049] The technical solution provided by this application has the following beneficial effects:

[0050] This application realizes the multi-dimensional data collection of students' learning status, providing a comprehensive and real-time data foundation for subsequent analysis. It converts raw data into structured features and extracts key indicators related to attention and emotion to facilitate subsequent correlation analysis. Through cross-modal data fusion, a dynamic model of learning status is established to accurately reflect students' attention fluctuations and emotional trends. It outputs executable strategies for the current teaching stage to achieve dynamic adjustment of teaching content and precise guidance of teacher-student interaction.

[0051] Furthermore, the present application also realizes deep coupling analysis of multimodal data by establishing a cross-modal mapping relationship between the physiological state indicator sequence and the action feature vector set in the time dimension, and generating a dynamic learning state map based on the changing correlation strength between the two.

[0052] Moreover, it breaks through the limitations of single-modal data analysis and reveals the intrinsic connection between physiological indicators and behavioral characteristics through cross-modal time series correlation, making the generated learning state map more accurate and dynamic, and providing a reliable basis for the optimization of teaching strategies.

[0053] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0055] Figure 1 A flowchart of an intelligent teaching analysis method based on big data provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the structure of an intelligent teaching analysis system based on big data provided in an embodiment of the present application;

[0057] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0059] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0060] Existing teaching status monitoring solutions primarily rely on single-image behavioral data analysis, identifying student focus solely through facial expressions and body movements. This approach presents limitations: First, they ignore the direct reflection of physiological indicators such as heart rate and blood oxygen levels on cognitive load and emotional state, resulting in a lack of physiological support for status assessment. Second, due to the lack of a dynamic correlation model between behavioral characteristics and physiological indicators, it is difficult to capture the co-evolution of attention fluctuations and emotional changes. Consequently, the output teaching suggestions are poorly aligned with the actual learning state, limiting the effectiveness of personalized teaching interventions. The essence of this problem lies in the fact that existing technologies process multimodal data in isolation, failing to meet the technical requirements for cross-modal dynamic correlation analysis for precise teaching decisions.

[0061] In response to the above-mentioned defects, this application proposes an intelligent teaching analysis method based on big data. The core of its innovation lies in the simultaneous collection of physiological monitoring data and behavioral monitoring data, and after generating a physiological state indicator sequence and a set of action feature vectors through real-time pattern recognition, a cross-modal time series correlation model of the two is established. Specifically, in the stage of constructing the dynamic learning state map, based on the spatiotemporal coupling analysis of the change rate of physiological indicators and the change amount of behavioral characteristics, the correlation strength between attention fluctuation nodes and emotion change nodes is quantified, and finally an intelligent decision is generated including teaching rhythm adjustment and interaction strategy optimization. This method breaks through the limitations of single modal analysis. Through the deep spatiotemporal alignment and collaborative calculation of physiological and behavioral data, it not only solves the state misjudgment problem caused by the lack of data dimensions in the existing technology, but also overcomes the decision lag caused by the isolated processing of multimodal data, and realizes the dynamic and accurate matching of teaching strategies with students' real learning states, thereby improving the timeliness of classroom interaction and the scientific nature of teaching adjustments.

[0062] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0063] Figure 1 A flowchart of an intelligent teaching analysis method based on big data provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0064] Step 101: During the class, obtain students' real-time physiological monitoring data and behavioral monitoring data.

[0065] In step 101, physiological monitoring data refers to time series signals reflecting students' physiological status, such as heart rate and blood oxygen saturation, collected by wearable devices. Behavioral monitoring data refers to time series data reflecting classroom behavioral characteristics, such as changes in sitting posture and body movement frequency, collected through video.

[0066] In this embodiment, after class begins, a smart bracelet continuously collects students' heart rate and blood oxygen data, generating a physiological signal stream measured in seconds. Simultaneously, a classroom camera captures a video stream of students' body movements at a fixed frame rate. Both types of data are uploaded to an edge computing node in real time via a wireless transmission module, where data format standardization and timestamp alignment are performed to ensure timing consistency for subsequent analysis.

[0067] For example, in math class, the smart bracelet worn by Student A collects heart rate data every 2 seconds (for example, the heart rate value rises from 78 beats / minute to 92 beats / minute). At the same time, the camera records his body leaning forward, turning his head, and other movements at 15 frames per second. Both types of data are marked with a unified timestamp and transmitted to the server.

[0068] Step 102: Performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data respectively to generate a physiological state indicator sequence and an action feature vector set.

[0069] In step 102, the physiological state indicator sequence represents a time-ordered sequence of attention-related indicators (e.g., heart rate variability) calculated from segmented heart rate data. The motion feature vector set represents a set of vectors encoding limb motion parameters (e.g., head deflection angle) extracted from video frames.

[0070] In this embodiment, a sliding window is used to segment heart rate data. The ratio of the difference between peaks and troughs within each window is calculated to generate a sequence of physiological indicators reflecting attentional concentration. A posture estimation algorithm is used to extract joint coordinates from video data. The angle and amplitude of joint displacement between adjacent frames are calculated to form a set of feature vectors describing the direction and force of the movement. Both types of data are stored in time-aligned windows.

[0071] For example, student A's heart rate data over 10 minutes is divided into 30-second windows, and the heart rate change rate of each window is calculated (for example, the change rate in window 1 is 0.6, and in window 2 is 0.9), generating the sequence [0.6, 0.9, ...]; synchronously, his head turning action is decomposed into the direction angle change (for example, turning left 15 degrees) and amplitude (lasting 2 seconds), encoded as a vector [(15,2), ...].

[0072] Step 103: Perform multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state map.

[0073] In step 103, the dynamic learning state graph represents a topological network with time as the horizontal axis and nodes as state markers (such as attention fluctuation nodes), and edges represent state transition relationships.

[0074] In an embodiment of the present application, the heart rate change rate in the physiological indicator sequence and the directional change of the motion feature vector are aligned according to the time window, and the product of the two is calculated as the association strength; when the association strength exceeds the threshold, the attention fluctuation node is marked at the corresponding time point; the emotion change node is marked for the period when the movement amplitude suddenly increases and is opposite to the heart rate change; finally, weighted connection edges are generated according to the temporal relationship between the nodes to construct a graph.

[0075] For example, during the 20-25 minute period, the correlation strength between Student A's heart rate change rate of 0.8 and the head turning angle change of 20 degrees was 16 (0.8×20), which exceeded the threshold of 12, so it was marked as an attention fluctuation node; during the same period, his sudden arm-crossing action (sudden increase in amplitude) was accompanied by a decrease in heart rate, which was marked as an emotion change node, and finally a graph containing two types of nodes and connection relationships was generated.

[0076] Step 104: Generate intelligent teaching analysis results based on the dynamic learning state map, where the intelligent teaching analysis results include content adjustment instructions and interactive guidance strategies for the current teaching stage.

[0077] In step 104, the intelligent teaching analysis results refer to a set of executable teaching instructions generated based on the analysis conclusions of the dynamic learning state map. Specifically, these include two types of strategies: content adjustment instructions dynamically optimize the teaching rhythm and knowledge presentation based on changes in student attention (such as adjusting the explanation speed, increasing repetitions, or inserting examples); and interactive guidance strategies design personalized interaction plans based on student emotional fluctuations (such as reducing the difficulty of questions, changing the interaction format, or adjusting the distance between teacher and student). These two strategies work together to accurately match teaching strategies with students' real-time status. The current teaching stage refers to the teaching progress or time period corresponding to the real-time teaching progress during the class.

[0078] In an embodiment of the present application, a dense area of attention nodes in three consecutive time windows is extracted from the graph, and the "key reinforcement" template in the teaching strategy library is matched to generate instructions to slow down the speech speed and increase example questions; the "emotional guidance" template is matched to isolated emotional nodes to trigger a strategy of approaching students and reducing the difficulty of asking questions.

[0079] For example, for Student A's three consecutive attention nodes at 20-30 minutes, the system prompts the teacher to "slow down the progress"; for his emotional node at 25 minutes, the system pushes the "encouraging questioning" strategy (such as "Can you tell me your solution ideas?").

[0080] This method achieves refined perception of students' attention and emotional states through the simultaneous collection of physiological and behavioral data and cross-modal dynamic correlation modeling, and outputs teaching strategies that adapt to the classroom progress, effectively improving the accuracy and timeliness of teaching interventions and solving the problems of misjudgment and lag caused by single modality analysis.

[0081] In order to solve the problem of inaccurate evaluation caused by isolated analysis of multimodal data in existing teaching status monitoring, in some embodiments, step 103: performing multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state map includes:

[0082] Step 201: establishing a cross-modal mapping relationship in the time dimension based on the physiological state indicator sequence and the motion feature vector set.

[0083] In step 201, the cross-modal mapping relationship refers to establishing a corresponding relationship between the heart rate change trend in the physiological state indicator sequence and the limb movement direction change in the action feature vector set within the same time segment, which is used to reflect the coordinated change characteristics of physiological reactions and behavioral performance.

[0084] In an embodiment of the present application, the physiological state indicator sequence is first divided into time segments of fixed duration, and the upward or downward trend of the heart rate change is extracted in each segment; synchronously, the deflection angle change trend of the limb movement direction in the action feature vector is extracted in the same time segment; finally, the two types of trend changes are aligned in time and a one-to-one mapping relationship table is established to form a basic framework for cross-modal data association.

[0085] Step 202: Generate a dynamic learning state map based on the correlation strength between the changes in the physiological state indicators and the changes in the motion feature vectors in the cross-modal mapping relationship.

[0086] In step 202, the correlation strength refers to the coupling coefficient obtained by quantifying the degree of matching between the amplitude of changes in physiological indicators and the direction of changes in movement characteristics, and is used to determine the change nodes of students' attention or emotional state.

[0087] In an embodiment of the present application, based on the established cross-modal mapping relationship table, the product of the heart rate change amplitude and the limb movement direction change angle in each time segment is calculated as the initial association strength; the association strength of continuous time segments is smoothed, and the peak intervals exceeding the preset threshold are screened out; attention fluctuation nodes are marked in these intervals, and emotion change nodes are marked for time segments with a sudden increase in limb movement amplitude and in the opposite direction of heart rate change; finally, all nodes are connected in chronological order to construct a dynamic learning state map.

[0088] Here's a specific example:

[0089] In a math classroom scenario, the smart bracelet worn by student A collects heart rate data every 2 seconds. His heart rate value rises from 78 beats per minute to 92 beats per minute. At the same time, the classroom camera records the student's body leaning forward and turning his head at a frequency of 15 frames per second. After the two types of data are synchronously transmitted to the server with a unified timestamp, the system divides the 10-minute heart rate data into 30-second windows and calculates the heart rate change rate in each window using the formula heart rate change rate equals the maximum heart rate in the window minus the minimum heart rate divided by the time window length. For example, the heart rate measured in the first window rises from 80 beats per minute to 92 beats per minute, and the change rate is 92 minus 80 divided by 30, which is 0.4. The change rate in the second window is 0.6, generating Sequence 0.4, 0.6; when processing behavioral data synchronously, the head turning action is decomposed into a direction angle change of 15 degrees to the left and a duration of 2 seconds, which is encoded as a vector 15, 2; when analyzing the 20 to 25 minute period, the system calculates that the correlation strength value between the heart rate change rate of 0.8 and the head turning angle change of 20 degrees in this period is 0.8 multiplied by 20, which is equal to 16. Since it exceeds the preset threshold of 12, it is marked as an attention fluctuation node. At the same time, it is detected that the student suddenly makes an arm-crossing action and the amplitude of the action reaches 1.5 times the preset threshold, accompanied by a heart rate drop of 10 beats per minute. The system marks it as an emotion change node, and finally generates a dynamic learning state map containing these two types of nodes and their time correlation relationship.

[0090] In an embodiment of the present application, by establishing a cross-modal dynamic correlation model of physiological and behavioral data, a three-dimensional assessment of students' learning status is achieved, so that the generated teaching strategy takes into account both attention persistence and emotional changes, thereby improving the accuracy and timeliness of classroom intervention.

[0091] To further improve the accuracy of learning state analysis, in some embodiments, step 202: generating a dynamic learning state map based on the correlation strength between the changes in physiological state indicators and the changes in motion feature vectors in the cross-modal mapping relationship, includes:

[0092] Step 301: performing correlation strength calculation on the amplitude change rate used to reflect the change of the physiological state indicator and the direction dimension change amount used to reflect the change of the motion feature vector in the direction dimension to generate a correlation strength value.

[0093] In step 301, the amplitude change rate is calculated by calculating the difference between adjacent peaks and troughs of the heart rate data in the physiological signal segment and dividing it by the duration of the signal segment (in seconds). The calculation formula is: Amplitude change rate = (current peak heart rate value - previous trough heart rate value) / peak-to-trough time interval. The directional dimension change is calculated by comparing the spatial orientation angle differences of consecutive action units in the action feature vector set, taking the absolute value, and dividing it by the action duration (in seconds). The calculation formula is: Directional dimension change = |current action unit orientation angle - previous action unit orientation angle| / action duration. The correlation strength value is the product of the amplitude change rate and the directional dimension change, reflecting the degree of coordination between physiological responses and behavioral changes.

[0094] In this embodiment, the heart rate fluctuation amplitude for each time segment is first extracted from the physiological state indicator sequence, and its ratio to the time window length is calculated to obtain the amplitude change rate. Simultaneously, the difference in the motion direction angles of adjacent motion units is extracted from the action feature vector set as the directional dimension change. Finally, the amplitude change rate within the same time segment is multiplied by the directional dimension change to generate a correlation strength value representing the degree of association between the two. For example, in a math classroom scenario, if Student A is detected to have a heart rate amplitude change rate of 0.8 beats / second (calculated as (95 beats / minute - 79 beats / minute) ÷ 20 seconds = 0.8) and a head rotation direction change of 15 degrees / second (calculated as |45 degrees - 30 degrees| ÷ 1 second = 15) during the 10:00-10:05 period, the correlation strength value is calculated as the amplitude change rate × the directional dimension change × the time window weight coefficient (0.5) = 0.8 × 15 × 0.5 = 6.0 (an attention fluctuation node is generated when the correlation strength value exceeds the preset threshold of 5).

[0095] Step 302: Mark the first physiological signal segment whose correlation strength value exceeds a set correlation strength threshold as an attention fluctuation node.

[0096] In step 302, the first physiological signal segment refers to a segment in the physiological state indicator sequence that contains heart rate changes and blood oxygen saturation data after segmenting the real-time physiological monitoring data into preset time windows, and the correlation strength value of this segment exceeds the set correlation strength threshold after correlation strength calculation. Attention fluctuation nodes are special nodes in the graph that mark these time segments and are used to indicate abnormal changes in the student's attention state.

[0097] In an embodiment of the present application, a sliding window detection is performed on the calculated association strength value sequence. When the association strength values of three consecutive time segments exceed the threshold, the period is marked as the first physiological signal segment, and an attention fluctuation node is created at the corresponding time position of the dynamic learning state map. The node size is proportional to the association strength value.

[0098] Step 303: The second physiological signal segment in which the amplitude dimension variation exceeds the preset fluctuation range is marked as an emotion change node. The amplitude dimension variation is used to reflect the degree of variation of the action feature vector in the amplitude dimension.

[0099] In step 303, the second physiological signal segment refers to a segment in the physiological state indicator sequence that contains heart rate changes and blood oxygen saturation data after segmenting the real-time physiological monitoring data into preset time windows, and the change in the amplitude dimension of the motion feature vector corresponding to this segment exceeds the preset fluctuation range. Emotion change nodes are nodes in the graph that mark these special time periods and are used to reflect sudden changes in students' emotions.

[0100] In an embodiment of the present application, the change in the amplitude dimension in the action feature vector is monitored. When it is detected that the amplitude value suddenly increases and exceeds twice the average fluctuation range, combined with the characteristic of the heart rate decrease during this period, the corresponding time segment is marked as a second physiological signal segment, and an emotion change node is created in the graph. The node color depth reflects the intensity of the emotion fluctuation.

[0101] Step 304: Construct a dynamic learning state graph based on the attention fluctuation node and the emotion change node.

[0102] In an embodiment of the present application, the marked attention fluctuation nodes and emotion change nodes are arranged in chronological order, the time intervals and attribute similarities of adjacent nodes are calculated, weighted connection edges are generated, and finally a dynamic learning state graph containing a complete state evolution path is constructed.

[0103] Here is a specific example:

[0104] In a math classroom scenario, Student A's smart bracelet continuously monitored his heart rate, which fluctuated from 82 beats per minute to 98 beats per minute within a period of 20 to 25 minutes. The system used a 30-second time window to calculate the heart rate change rate. The specific formula is that the heart rate change rate is equal to the maximum heart rate in the window minus the minimum heart rate divided by the window duration. The heart rate change rate during this period was measured to be 98 minus 82 divided by 30, which is equal to 0.53. At the same time, the camera captured an increase in the frequency of the student's head turning. After analysis and calculation, the average directional dimension change was 22 degrees every 30 seconds. The system multiplied the heart rate change rate of 0.53 by the directional dimension change of 22 to obtain a correlation strength value of 11.66. When this value exceeded the preset threshold of 10, it automatically The system marked the period as an attention fluctuation node; then at 26 minutes, the system detected that the student suddenly crossed his arms and the amplitude of the movement reached 35 cm, which was 1.75 times the baseline amplitude of 20 cm. At the same time, the heart rate dropped sharply from 95 beats per minute to 83 beats per minute, satisfying the conditions that the amplitude dimension change exceeded 1.5 times the preset fluctuation range and the heart rate dropped by more than 10 beats per minute, so it was marked as an emotion change node; finally, the system connected these nodes in chronological order to construct a dynamic learning state map showing that attention fluctuations preceded emotion changes, where the node spacing reflects the state transition speed and the line thickness represents the transition intensity, providing teachers with an intuitive basis for visual analysis of state evolution.

[0105] In the embodiment of the present application, by quantifying the dynamic correlation characteristics of physiological and behavioral data, the changes in learning status are captured in a refined manner. The constructed map can intuitively display the complete process of students from attention fluctuations to emotional changes, providing teachers with accurate state evolution clues and a basis for judging the timing of intervention.

[0106] To further improve the accuracy and interpretability of the learning state map, in some embodiments, step 304: constructing a dynamic learning state map based on the attention fluctuation nodes and the emotion change nodes includes:

[0107] Step 401: performing overlapping area detection on the start time of the attention fluctuation node and the duration of the emotion change node to generate a spatiotemporal interaction coefficient between the attention fluctuation node and the emotion change node.

[0108] In step 401, the starting time of the attention fluctuation node refers to the specific moment when the student's attention begins to fluctuate, as identified by the correlation strength calculation. This moment marks the critical point when the change rate of the physiological indicator and the change amount of the behavioral characteristics first reach the preset correlation threshold. The duration of the emotion change node refers to the total time from the first time the change amount of the action feature vector amplitude dimension exceeds the preset fluctuation range to the time when the change amount falls back to the normal range, reflecting the duration of the student's abnormal emotional state. The spatiotemporal interaction coefficient refers to the weighted value of the overlapping part of the time interval of the attention fluctuation node and the time interval of the emotion change node. It is used to quantify the degree of correlation between the two types of nodes in the spatiotemporal dimension. The calculation takes into account the two factors of overlapping duration and node strength.

[0109] In an embodiment of the present application, the start time and duration of each attention fluctuation node, as well as the appearance time and duration of the emotion change node, are first obtained, and the length of the overlapping part of the two time intervals is calculated; then the overlapping duration is multiplied by the strength value of the two nodes to obtain an interaction coefficient that reflects the degree of spatiotemporal correlation. The larger the coefficient, the stronger the correlation between the two types of nodes.

[0110] Step 402: Generate a graph connection line reflecting the correlation between attention fluctuation and emotion change based on the node pairs whose spatiotemporal interaction coefficient exceeds a preset coupling threshold.

[0111] In step 402, the graph connection line is a directed edge connecting nodes of different categories in the dynamic learning state graph. The direction of the edge represents the state evolution path from attention fluctuation to emotion change, and the thickness of the edge reflects the strength of the association between the two types of nodes.

[0112] In an embodiment of the present application, node pairs whose spatiotemporal interaction coefficients exceed a preset threshold are screened out, and the initial weight of the connecting line is determined according to the size of the coefficient; then, in chronological order, the attention fluctuation node is pointed to the emotion change node to generate a directed connecting line, and the thickness of the connecting line is proportional to the spatiotemporal interaction coefficient.

[0113] Step 403: Count the number of interactions of the graph connection lines within a preset time window to generate a topological structure.

[0114] In step 403, the topological structure refers to a network relationship diagram consisting of multiple nodes and connection lines. The densely connected areas in the structure represent critical periods where the learning state fluctuates frequently, and the sparse areas represent stable state periods.

[0115] In an embodiment of the present application, the number of interactions of all connection lines within a preset time window is counted, and cluster analysis is performed on areas with a high number of interactions to identify hot spots of state transitions. Based on the distribution characteristics of the hot spots and the density of connection lines, a topological structure framework reflecting the overall state evolution law is generated. For example, in a 45-minute classroom scenario, every 5 minutes is set as an analysis time window. When it is detected that a student's attention fluctuation node and emotion change node overlap three times in the 10-15 minute window (the number of overlaps is calculated as follows: when there are both attention fluctuation nodes and emotion change nodes in the same time segment, it is counted as one overlap), and the spatiotemporal interaction coefficient of each overlap exceeds the preset coupling threshold of 0.7 (this threshold is obtained through historical classroom data statistics, and the calculation formula is: coupling threshold = number of effective interactions / total number of interactions × 0.9, where the effective number of interactions refers to the node interaction events that actually affect the teaching strategy), then a graph connection line with a connection strength of 3 is generated in this time window; when the connection strength of three consecutive time windows (i.e., 15 minutes) is greater than 2, an evolutionary path from "focused state" to "emotional fluctuation state" is generated, and the weight value of this path is 2.3 (weight value = Σwindow connection strength / number of windows × 0.8, where 0.8 is the classroom scenario correction coefficient), and finally the topological structure containing this evolutionary path is integrated into the dynamic learning state map.

[0116] Step 404: Integrate the attention fluctuation nodes, the emotion change nodes, the graph connection lines, and the topological structure to construct a dynamic learning state graph.

[0117] In an embodiment of the present application, the marked attention fluctuation nodes and emotion change nodes are arranged on the timeline in chronological order, and the vertical coordinate position is determined according to the node strength; then, the connection lines between the nodes are drawn according to the connection relationship of the topological structure; finally, all elements are integrated to generate a complete dynamic learning state map, in which the node size represents the state fluctuation intensity, and the color depth of the connection line reflects the transformation speed.

[0118] Here's a specific example:

[0119] In a math classroom scenario, the system identified a node for Student A experiencing an attention fluctuation between 20:30 and 23:45, with an onset time of 20:30 and a duration of 3:15. Subsequently, an emotion fluctuation node was detected between 24:10 and 26:20, lasting 2:10. The system calculated the overlap between the two nodes to be 1:35, from 24:10 to 23:45. The spatiotemporal interaction coefficient was calculated as the overlap duration multiplied by the mean node strength. Taking the mean of the attention node strength of 12 and the emotion node strength of 8, the interaction coefficient was 10, resulting in an interaction coefficient of 1.58 minutes multiplied by 10, equaling 15.8. Because this value exceeded the preset coupling threshold of 12, the system generated a graph link from the attention node to the emotion node, with a line width set to 3.16 units: the interaction coefficient of 15.8 divided by the baseline value of 5. During the 30-minute teaching period, a total of four interactions between similar node pairs were counted, forming a dense topological region centered at the 25-minute mark. During the final integration, the attention node was marked with a blue circle at 20 minutes and 30 seconds, and the emotion node was marked with a red triangle at 24 minutes and 10 seconds. The two nodes were connected with an orange connecting line with a width of 3.16 units. A topological heat map reflecting the four state transitions was generated on the right side of the map, fully displaying the student's dynamic evolution from distracted attention to low mood. The heat value of the topological structure is proportional to the number of interactions, providing teachers with a clear reminder that "focus on observation before and after 25 minutes."

[0120] In the embodiment of the present application, the node correlation is accurately quantified by the spatiotemporal interaction coefficient, and the state evolution law is revealed by the topological structure. The constructed dynamic learning state map can not only intuitively display the state change details of a single student, but also reflect the typical state transition pattern in classroom teaching, providing a reliable basis for teachers to implement precise teaching intervention.

[0121] In order to further improve the accuracy and practicality of teaching strategy generation, in some embodiments, step 104: generating intelligent teaching analysis results based on the dynamic learning state map includes:

[0122] Step 501: Determine the pattern evolution segment corresponding to the current teaching stage according to the dynamic learning state map.

[0123] In step 501, the pattern evolution segment refers to a typical state transition period extracted from the dynamic learning state map, which contains continuous attention fluctuation nodes and emotion change nodes, and is used to reflect the changing rules of students' learning states in a specific teaching stage.

[0124] In an embodiment of the present application, the time distribution characteristics of the nodes in the graph are first analyzed to identify continuous time periods in which the nodes appear intensively; then all the nodes and their connection relationships within these time periods are extracted to form pattern fragments that reflect the complete state evolution process; finally, according to the current teaching progress, the pattern evolution fragment closest in time is selected as the basis for analysis.

[0125] Step 502: Dynamically match the pattern evolution segment with the teaching scenario template in the preset teaching strategy library.

[0126] In step 502, the teaching scenario template is a typical teaching scenario description stored in a preset strategy database, including the corresponding relationship between attention fluctuation characteristics, emotion change characteristics and recommended strategies. Dynamic matching is the process of calculating the similarity between the pattern features extracted in real time and the features in the template library.

[0127] In an embodiment of the present application, characteristic parameters such as node type, strength, and connection line density are extracted from the pattern evolution fragment; these parameters are compared item by item with the characteristic dimensions in the teaching scene template; the matching score of each template is calculated, and a set of candidate templates with scores higher than the threshold are screened out.

[0128] Step 503: When the matching result is a plurality of teaching scene templates, content adjustment instructions and interactive guidance strategies are generated according to the priority weights of different teaching scene templates in the matching result.

[0129] In step 503, the priority weight is a strategy recommendation index calculated based on the matching score and the importance of the teaching stage, and is used to determine the order of strategy execution during multi-template matching.

[0130] In an embodiment of the present application, the product of the matching score of each candidate template and the weight of the current teaching stage is calculated to obtain the priority weight; the templates are sorted from high to low according to the weight value, and the top three templates with the highest weight are selected; the content adjustment methods and interaction strategies recommended by these templates are extracted respectively for fusion optimization.

[0131] Step 504: Combine the content adjustment instruction and the interactive guidance strategy to generate an intelligent teaching analysis result.

[0132] In an embodiment of the present application, the selected content adjustment instructions are arranged and combined according to the logical order of knowledge points; at the same time, the interactive guidance strategies are graded according to the implementation difficulty and expected effect; and finally a complete teaching analysis report including specific implementation steps and time arrangements is generated.

[0133] Here's a specific example:

[0134] In a math classroom scenario, the system extracted a pattern evolution segment from Student A's dynamic learning state graph for a 20-30 minute teaching period. This segment contained three consecutive attention fluctuation nodes and one emotion change node. The average correlation strength of the attention node was 15, and the amplitude change of the emotion node was 1.8 times the baseline value. When matching this segment's features with the teaching strategy library, the calculated match with the "difficulty understanding knowledge points" template was 80 points (the matching formula is the number of nodes multiplied by the average strength coefficient of 15 divided by 10, which equals 45, plus the amplitude change coefficient of 1.8 multiplied by 20, which equals 36, rounded to the nearest whole number of 81 points). The match with the "low class participation" template was 75 points. The system assumed the current teaching stage to be "new knowledge explanation" and the importance coefficient to be 1.1, and calculated the priority weights as 80 times 1.1 (88) and 75 times 1.1 (82.5), respectively. The "difficulty in understanding knowledge points" template with the highest weight was selected to generate the content adjustment instruction of "increasing the number of example demonstrations" and the interactive guidance strategy of "using heuristic questions". Finally, the intelligent teaching analysis result of "first supplement two transitional examples and then ask 'What do you think about this problem-solving step'" was combined.

[0135] In the embodiments of the present application, dynamic adaptation of teaching strategies to students' actual status is achieved through precise extraction of pattern evolution fragments and multi-dimensional strategy matching. The generated intelligent analysis results take into account both the systematic nature of knowledge transfer and the targeted nature of teacher-student interaction, effectively improving the flexibility and effectiveness of classroom teaching.

[0136] To further improve the accuracy and applicability of teaching strategy generation, in some embodiments, step 503: generating content adjustment instructions and interactive guidance strategies based on the priority weights of different teaching scenario templates in the matching results, includes:

[0137] Step 601: Assign a first priority weight to the teaching scenario templates in the matching results whose attention fluctuation frequency exceeds the preset attention fluctuation threshold range and whose emotion change amplitude is within the preset emotion fluctuation tolerance range.

[0138] In step 601, the first priority weight refers to the weight value of the teaching strategy that is preferentially adopted when the student's attention fluctuates but the emotion is relatively stable. This weight focuses on solving the problem of knowledge comprehension.

[0139] In an embodiment of the present application, the number of attention fluctuation nodes in the pattern evolution segment is first counted, and the fluctuation frequency per unit time is calculated; then, it is detected whether the amplitude change of the emotion change node is within the preset normal fluctuation range; finally, the teaching template that meets both the excessive attention frequency and normal emotion amplitude is given a priority weight calculated based on the fluctuation frequency value.

[0140] Step 602: Assign a second priority weight to the teaching scenario templates in the matching results whose emotion change amplitude exceeds the preset emotion fluctuation tolerance range and whose attention fluctuation frequency is within the preset attention fluctuation threshold range.

[0141] In step 602, the second priority weight refers to the weight value of the teaching strategy that is preferentially adopted when the student's emotions fluctuate but his attention is relatively stable. This weight focuses on solving psychological state problems.

[0142] In an embodiment of the present application, it is detected whether the amplitude of the emotion change node exceeds the upper tolerance limit; at the same time, it is verified whether the attention fluctuation frequency is within the normal threshold; for the teaching template that meets this condition, a priority weight calculated based on the amplitude of the emotion change is given, and the larger the amplitude, the higher the weight.

[0143] Step 603: Assign a third priority weight to the teaching scenario templates in the matching results whose emotion change amplitude exceeds the preset emotion fluctuation tolerance range and whose attention fluctuation frequency exceeds the preset attention fluctuation threshold range.

[0144] In step 603, the third priority weight refers to the weight value of the teaching strategy adopted when the student has both attention fluctuations and emotional fluctuations. The weight needs to take into account both cognitive and emotional factors.

[0145] In an embodiment of the present application, abnormal time periods that exceed both the attention frequency threshold and the emotion amplitude tolerance are identified; a weighted sum of the two abnormal indicators is calculated; and a priority weight that comprehensively considers the two abnormal degrees is given to the matching composite teaching template.

[0146] Step 604: Generate content adjustment instructions and interaction guidance strategies according to the ranking results of the first priority weight, the second priority weight, and the third priority weight.

[0147] In an embodiment of the present application, teaching templates are selected in the order of first, second, and third priority; the content adjustment methods and interactive strategies recommended by each template are extracted; and strategies are integrated according to the characteristics of the current teaching stage to generate an executable instruction combination.

[0148] Here's a specific example:

[0149] In a math classroom scenario, the system analyzed Student A's pattern evolution over the 20-30 minute period and detected that the student's attention fluctuation frequency reached 0.7 times per minute, exceeding the preset threshold of 0.5 times. The amplitude of his emotional fluctuations was 1.8 times the baseline value, which was within the preset tolerance range of 1.5-2.0 times. This met the criteria for assigning a first-priority weight. According to the formula, the first-priority weight is equal to the matching degree multiplied by the teaching stage coefficient. The matching degree of the knowledge point comprehension difficulty template of 80 points multiplied by the new knowledge explanation stage coefficient of 1.1 resulted in a weight of 88 points. Furthermore, it was found that the student's emotional fluctuations increased sharply to the baseline value of 2. 3 times, which exceeds the upper limit of tolerance by 2.0 times, but the frequency of attention fluctuations of 0.3 times is within the normal range of less than 0.5 times, which meets the second priority weight condition. The low classroom participation template matching degree of 75 points is multiplied by 1.1 to obtain a weight of 82.5 points; the system gives priority to the knowledge point understanding difficulty template with a weight of 88 points, and generates a content adjustment instruction to add 2 transition examples. The specific number of examples is obtained by dividing the duration of attention fluctuations of 3 minutes by the preset benchmark interval of 1.5 minutes. In conjunction with the interactive strategy of heuristic questioning, the question interval is set to the time required for emotions to calm down, which is 5 minutes, divided by the preset question density coefficient of 2, which is equal to 2.5 minutes.

[0150] In the embodiment of the present application, the core problem types of students are accurately identified through a hierarchical weight mechanism, so that the generated teaching strategies can not only solve the main contradictions in a targeted manner, but also take into account the secondary problems, thereby achieving a precise transformation from state identification to strategy implementation, and effectively improving the timeliness and effectiveness of teaching intervention.

[0151] To further improve the accuracy and timeliness of physiological and behavioral data analysis, in some embodiments, step 102: performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data to generate a physiological state indicator sequence and a motion feature vector set includes:

[0152] Step 701: Segment the real-time physiological monitoring data into a plurality of physiological signal segments according to a preset time window.

[0153] In step 701, the physiological signal segment refers to a time period data unit obtained by dividing continuous physiological monitoring data into fixed time periods, and each segment contains a complete heart rate fluctuation cycle feature.

[0154] In an embodiment of the present application, a time window length suitable for a classroom scenario is first set, and then the physiological data collected in real time is divided into equal lengths according to the window to ensure that each segment can reflect a complete state change cycle. Finally, the segmented segments are quality checked to eliminate invalid segments with incomplete signals.

[0155] Step 702: Identify physiological state indicators associated with attention fluctuations from each physiological signal segment.

[0156] In step 702 , the physiological state indicator associated with attention fluctuation refers to a quantitative feature extracted from physiological parameters such as heart rate variability and blood oxygen saturation change, which can characterize the change in the degree of attention concentration.

[0157] In an embodiment of the present application, waveform feature analysis is performed on each valid physiological signal segment, the rhythm change pattern therein is detected, characteristic parameters that can reflect the activity of the autonomic nervous system are calculated, and a correspondence is established between these parameters and the attention state to form a standardized state indicator.

[0158] Step 703: Arrange all the physiological status indicators in chronological order to form a physiological status indicator sequence.

[0159] In step 703, the time sequence and the preset time window are not the same concept: the preset time window is a fixed interval for data segmentation, and the time sequence is the sequential relationship of the segmented segments.

[0160] In an embodiment of the present application, the attention-related indicators calculated in each time window are arranged in chronological order of acquisition time, while retaining the timestamp information, to form a time series data sequence that can reflect the continuous changes in the attention state.

[0161] Step 704: Split the body movements captured in the behavior monitoring data into multiple movement units.

[0162] In step 704, action units are semantically complete segments of physical movements segmented from continuous behavioral data. Each unit contains a complete movement process with a start, duration, and end. The structure of an action unit contains encoded information in the direction and amplitude dimensions, which are used to describe the spatial variation characteristics of the physical movement.

[0163] In an embodiment of the present application, by analyzing the motion trajectory and posture change characteristics in the behavior monitoring data, the start and end key frames of the action are identified, the continuous behavior stream is divided into independent units with clear action intentions, and the action type and spatiotemporal characteristics are labeled for each unit.

[0164] Step 705: Identify action feature vectors associated with emotion changes from each action unit.

[0165] In step 705 , the action feature vector associated with the emotion change refers to a combination of motion features extracted from the action unit that can reflect the change in the emotional state.

[0166] In an embodiment of the present application, the basic characteristics of each action unit, such as movement amplitude, speed, and direction, are analyzed, the statistics and change patterns of these characteristics are calculated, and feature combinations with stable correlation with typical emotional states are screened out and encoded into multidimensional feature vectors.

[0167] Step 706: Combine all the action feature vectors into an action feature vector set.

[0168] In an embodiment of the present application, the emotion feature vectors obtained by analyzing each action unit are arranged in the chronological order of the action, while retaining the time correspondence with the physiological data, to form a feature set that can be used for subsequent multimodal analysis.

[0169] Here's a specific example:

[0170] In a math classroom scenario, the system segments Student A's real-time physiological monitoring data into preset 30-second time windows, obtaining 10 physiological signal segments over a 20-25 minute period. For each segment, the heart rate variability is calculated using the formula: the maximum heart rate minus the minimum heart rate within the window divided by the time window length. For example, in the first window, the measured heart rate increases from 82 to 98 beats per minute, resulting in a variability of 98 minus 82 divided by 30, which equals 0.53. In the second window, the variability is 0.67, generating the sequence 0.53, 0.67. To simultaneously process behavioral data, the 20-25 minute video stream is broken down into 32 action units, including 15 head turns and 5 forward leans. For each head turn, the directional angle change is calculated. For example, a 20-degree left turn lasting 3 seconds is encoded as the vector 20, 3. For forward leans, the forward lean angle change is calculated, such as a 30-degree forward lean lasting 2 seconds is encoded as the vector 30, 2. Specifically, an arm-crossing motion was detected at 24 minutes and 30 seconds, with its amplitude calculated to be 2.1 times the baseline value, lasting for 4 seconds, and encoded as a special emotion feature vector 2.1, 4. Ultimately, the system generated a physiological state indicator sequence consisting of 10 heart rate variability indicators and a motion feature vector set of 32 motion feature vectors, providing a data foundation for the subsequent construction of a dynamic learning state map.

[0171] In the embodiments of the present application, a unified representation of multi-source heterogeneous data is achieved through standardized data segmentation and feature extraction processes, providing a reliable feature basis for accurately identifying students' learning status while ensuring the real-time and interpretability of data analysis.

[0172] Figure 2 A schematic diagram of the structure of an intelligent teaching analysis system based on big data provided in the embodiment of the present application is shown as follows: Figure 2 As shown, the system includes:

[0173] The acquisition module 21 is used to acquire students' real-time physiological monitoring data and behavioral monitoring data during class.

[0174] The recognition module 22 is used to perform instant pattern recognition on the real-time physiological monitoring data and the behavior monitoring data, and generate a physiological state indicator sequence and an action feature vector set.

[0175] The analysis module 23 is used to perform multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state map.

[0176] The generation module 24 is used to generate intelligent teaching analysis results based on the dynamic learning state map, and the intelligent teaching analysis results include content adjustment instructions and interactive guidance strategies for the current teaching stage.

[0177] Figure 2 The intelligent teaching analysis system based on big data can be executed Figure 1 The implementation principles and technical effects of the big data-based intelligent teaching analysis method described in the illustrated embodiment will not be elaborated on here. The specific manner in which each module and unit performs operations in the big data-based intelligent teaching analysis system in the above embodiment has been described in detail in the embodiment of the method and will not be elaborated on here.

[0178] In one possible design, Figure 2 An intelligent teaching analysis system based on big data in the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0179] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0180] The processing component 32 is used for the above Figure 1 The embodiment provides an intelligent teaching analysis method based on big data.

[0181] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0182] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0183] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0184] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0185] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0186] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0187] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is an intelligent teaching analysis method based on big data.

[0188] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0189] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0190] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0191] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent teaching analysis method based on big data, characterized in that: include: During the class, obtain students' real-time physiological and behavioral monitoring data; Performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data to generate a physiological state indicator sequence and an action feature vector set; Performing multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state map; Based on the dynamic learning state map, generating intelligent teaching analysis results, the intelligent teaching analysis results including content adjustment instructions and interactive guidance strategies for the current teaching stage; The step of performing multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state graph includes: Establishing a cross-modal mapping relationship in a time dimension according to the physiological state indicator sequence and the motion feature vector set; generating a dynamic learning state map according to the correlation strength between the changes in the physiological state indicators and the changes in the motion feature vectors in the cross-modal mapping relationship; Generating a dynamic learning state map according to the correlation strength between the changes in the physiological state indicators and the changes in the motion feature vectors in the cross-modal mapping relationship includes: Calculate the correlation strength between the amplitude change rate used to reflect the change of the physiological state indicator and the direction dimension change amount used to reflect the change of the motion feature vector in the direction dimension to generate a correlation strength value; marking the first physiological signal segment whose association strength value exceeds the set association strength threshold as an attention fluctuation node; The second physiological signal segment in which the amplitude dimension variation exceeds the preset fluctuation range is marked as an emotion change node, and the amplitude dimension variation is used to reflect the degree of variation of the action feature vector in the amplitude dimension; Constructing a dynamic learning state graph based on the attention fluctuation node and the emotion change node; The step of constructing a dynamic learning state graph based on the attention fluctuation node and the emotion change node includes: Perform overlapping area detection on the start time of the attention fluctuation node and the duration of the emotion change node to generate a spatiotemporal interaction coefficient between the attention fluctuation node and the emotion change node. The spatiotemporal interaction coefficient is a weighted value of the overlapping portion of the time interval of the attention fluctuation node and the time interval of the emotion change node, and is used to quantify the degree of correlation between the attention fluctuation node and the emotion change node in the spatiotemporal dimension; generating, based on the node pairs whose spatiotemporal interaction coefficients exceed a preset coupling threshold, a graph connection line reflecting the correlation between attention fluctuations and emotion changes; Counting the number of interactions of the graph connection lines within a preset time window to generate a topological structure; The attention fluctuation nodes, the emotion change nodes, the graph connection lines and the topological structure are integrated to construct a dynamic learning state graph.

2. The method according to claim 1, characterized in that Generating intelligent teaching analysis results based on the dynamic learning state graph includes: Determining a pattern evolution segment corresponding to a current teaching stage according to the dynamic learning state map; Dynamically matching the pattern evolution fragments with the teaching scenario templates in the preset teaching strategy library; When the matching result is multiple teaching scene templates, content adjustment instructions and interactive guidance strategies are generated according to the priority weights of different teaching scene templates in the matching result; The content adjustment instruction and the interactive guidance strategy are combined to generate an intelligent teaching analysis result.

3. The method according to claim 2, characterized in that Generating content adjustment instructions and interactive guidance strategies according to the priority weights of different teaching scenario templates in the matching results includes: The first priority weight is given to the teaching scenario templates whose attention fluctuation frequency exceeds the preset attention fluctuation threshold range and whose emotion change amplitude is within the preset emotion fluctuation tolerance range in the matching results; The second priority weight is assigned to the teaching scenario templates whose emotion change amplitude exceeds the preset emotion fluctuation tolerance range and whose attention fluctuation frequency is within the preset attention fluctuation threshold range in the matching results; The third priority weight is assigned to the teaching scenario templates whose emotion change amplitude exceeds the preset emotion fluctuation tolerance range and whose attention fluctuation frequency exceeds the preset attention fluctuation threshold range in the matching results; A content adjustment instruction and an interaction guidance strategy are generated according to the ranking results of the first priority weight, the second priority weight, and the third priority weight.

4. The method according to claim 1, wherein The performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data to generate a physiological state indicator sequence and an action feature vector set includes: Segmenting the real-time physiological monitoring data into a plurality of physiological signal segments according to a preset time window; identifying physiological state indicators associated with attention fluctuations from each physiological signal segment; Arranging all the physiological state indicators in chronological order into a physiological state indicator sequence; Splitting the limb movements captured in the behavior monitoring data into multiple action units; Identifying action feature vectors associated with emotion changes from each action unit; All the action feature vectors are combined into an action feature vector set.

5. An intelligent teaching analysis system based on big data, characterized in that: include: The acquisition module is used to obtain students' real-time physiological monitoring data and behavioral monitoring data during the class; an identification module for performing instant pattern recognition on the real-time physiological monitoring data and the behavioral monitoring data, respectively, to generate a physiological state indicator sequence and an action feature vector set; An analysis module, configured to perform multimodal temporal correlation analysis on the physiological state indicator sequence and the motion feature vector set to generate a dynamic learning state map; A generation module, configured to generate intelligent teaching analysis results based on the dynamic learning state map, wherein the intelligent teaching analysis results include content adjustment instructions and interactive guidance strategies for the current teaching stage; The step of performing multimodal temporal correlation analysis on the physiological state indicator sequence and the action feature vector set to generate a dynamic learning state graph includes: Establishing a cross-modal mapping relationship in a time dimension according to the physiological state indicator sequence and the motion feature vector set; generating a dynamic learning state map according to the correlation strength between the changes in the physiological state indicators and the changes in the motion feature vectors in the cross-modal mapping relationship; Generating a dynamic learning state map according to the correlation strength between the changes in the physiological state indicators and the changes in the motion feature vectors in the cross-modal mapping relationship includes: Calculate the correlation strength between the amplitude change rate used to reflect the change of the physiological state indicator and the direction dimension change amount used to reflect the change of the motion feature vector in the direction dimension to generate a correlation strength value; marking the first physiological signal segment whose association strength value exceeds the set association strength threshold as an attention fluctuation node; The second physiological signal segment in which the amplitude dimension variation exceeds the preset fluctuation range is marked as an emotion change node, and the amplitude dimension variation is used to reflect the degree of variation of the action feature vector in the amplitude dimension; Constructing a dynamic learning state graph based on the attention fluctuation node and the emotion change node; The step of constructing a dynamic learning state graph based on the attention fluctuation node and the emotion change node includes: Perform overlapping area detection on the start time of the attention fluctuation node and the duration of the emotion change node to generate a spatiotemporal interaction coefficient between the attention fluctuation node and the emotion change node. The spatiotemporal interaction coefficient is a weighted value of the overlapping portion of the time interval of the attention fluctuation node and the time interval of the emotion change node, and is used to quantify the degree of correlation between the attention fluctuation node and the emotion change node in the spatiotemporal dimension; generating, based on the node pairs whose spatiotemporal interaction coefficients exceed a preset coupling threshold, a graph connection line reflecting the correlation between attention fluctuations and emotion changes; Counting the number of interactions of the graph connection lines within a preset time window to generate a topological structure; The attention fluctuation nodes, the emotion change nodes, the graph connection lines and the topological structure are integrated to construct a dynamic learning state graph.

6. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an intelligent teaching analysis method based on big data as described in any one of claims 1 to 4.

7. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, an intelligent teaching analysis method based on big data as described in any one of claims 1 to 4 is implemented.

Images