A preschool smart classroom management system based on big data

Abstract

The present invention relates to the field of learning management technology, specifically a preschool smart classroom management system based on big data, the system includes: a classroom behavior analysis module, which collects preschool classroom video stream data, extracts image sequences to analyze image features, calculates and obtains time series of features and image frames, combines and classifies feature sequences to determine student behavior labels, and combines labels and image frames to generate classroom behavior feature data. In the present invention, based on the spatial distribution parameters generated by location points and activity frequencies, the heat distribution of classroom activities in space is accurately quantified through frequency classification and location parameter processing, and spatial visualization of classroom activities is achieved, which facilitates teachers to make scene decisions based on classroom dynamics. By comparing time series and activity heat maps, the participation distribution is refined, so that teachers can obtain student participation data with more time period and location details, thereby adjusting teaching strategies and achieving personalized teaching.

Description

Technical Field

[0001] The present invention relates to the technical field of learning management, and in particular to a preschool smart classroom management system based on big data. Background Art

[0002] The field of learning management technology mainly focuses on combining information technology with education management to improve the efficiency and effectiveness of the teaching process. The technologies covered in this field include data analysis, intelligent recognition, behavior tracking, learning path optimization, etc. Through the careful management of the learning process, it helps teachers better understand students' learning situation, behavior performance and learning outcomes.

[0003] The existing technology of data collection and analysis in learning management is mostly focused on behavior tracking and general learning path management. It lacks dynamic spatial analysis of real-time classroom behavior and detailed feedback on participation data, which makes it difficult for teachers to obtain the activity heat in the classroom and the accurate distribution of student behavior. This limitation reduces the pertinence of decision-making basis in teachers' classroom management. For example, conventional data analysis methods are difficult to refine to the behavioral status and participation in different time periods and specific areas, and it is impossible to fully grasp the students' concentration at each time point, especially in a classroom environment with frequent position changes. It is difficult for teachers to obtain the behavioral dynamics and participation changes of each student. Due to the lack of participation data analysis combined with time and space, it is difficult for teachers to adjust teaching content or activity layout in a timely manner. The lack of real-time feedback in the classroom affects the flexibility of teaching and the refinement of classroom management. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose a preschool smart classroom management system based on big data.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: A preschool smart classroom management system based on big data includes:

[0006] The classroom behavior analysis module collects preschool classroom video stream data, extracts image sequences to analyze image features, calculates and obtains the time series of features and image frames, combines and classifies feature sequences to determine student behavior labels, combines labels with image frames to generate classroom behavior feature data, and obtains behavior feature distribution through parameter screening, classification and judgment of the classroom behavior feature data;

[0007] A classroom activity heat map generation module, based on the distribution of the behavioral characteristics, extracts the position points and activity frequencies of the image frames, generates spatial distribution parameters by combining the position points and frequency values, obtains the classroom activity distribution by parameter classification and comparison, and generates a classroom activity heat map by combining the classroom activity distribution and the image area frequency;

[0008] A class participation evaluation module, based on the class activity heat map, obtains time series and activity heat map frequency comparison behavior feature data, combines heat map position and time series classification to obtain class participation comparison data, extracts the class participation comparison data and classifies it according to the activity time series to obtain participation distribution;

[0009] The teaching feedback mechanism module extracts the participation changes and time differences of the behavior parameters based on the participation distribution, classifies the changes and behavior parameter sequences into feedback sequences, and obtains a feedback suggestion set by combining the feedback sequence and historical behavior data.

[0010] Preferably, the steps of acquiring the behavior feature distribution are:

[0011] According to the image features of each frame in the image sequence, combined with the corresponding timestamp in the frame sequence, the feature information and time parameters of each image frame are combined to form an image feature time series;

[0012] Calling the image feature time series, performing association analysis between image features and behavior labels, matching and judging the features of each frame of the image with a predefined behavior label set, selecting and classifying image frames that meet the label standard, and generating a feature label classification sequence;

[0013] Calling the feature label classification sequence, classifying each behavior label and feature sequence according to the image frame interval, and generating a behavior feature classification sequence;

[0014] The behavior feature classification sequence is called, and the feature similarity between the tag groups is compared. The parameter screening is performed in combination with the distribution of the behavior classification, using the formula:

[0015]

[0016] Get the behavior feature distribution; where, represents the distribution value of behavioral characteristics, Indicates The image feature value of the frame, Indicates The label feature value of the frame, is the inter-frame feature weight parameter.

[0017] Preferably, the steps of obtaining the spatial distribution parameters are:

[0018] The result of the classroom behavior distribution is called, the position point and activity frequency of each image frame are analyzed, each position point is weighted according to the frequency of occurrence, the weighted frequency value of each position point is calculated, and a weighted position frequency sequence is generated;

[0019] Calling the weighted position frequency sequence, normalizing the weighted frequency value of each position point, and generating a normalized position frequency sequence;

[0020] The normalized position frequency sequence is called, and the spatial distribution parameters of the position frequency of each activity area are calculated in combination with the size and layout of each activity area in the classroom, using the formula:

[0021]

[0022] Generate spatial distribution parameters; among them, represents the spatial distribution parameter, Indicates The normalized frequency value of the position, Indicates the visual impact value of the position in the spatial layout, Represents the maximum value of the normalized frequency values ​​of all positions.

[0023] Preferably, the steps for obtaining the classroom activity heat map are:

[0024] Calling the spatial distribution parameters, analyzing the location points and activity frequencies of each activity area in the classroom, mapping the activity level of each location point in the classroom, calculating the activity heat of each location point, and generating an activity heat sequence;

[0025] The activity heat sequence is called, and the activity heat value is adjusted by normalization processing in combination with the spatial size of each position to generate a normalized activity heat sequence;

[0026] Call the normalized activity heat sequence, through the formula:

[0027]

[0028] The heat distribution of each location point is converted into color changes on the image to generate a classroom activity heat map; Indicates at location The heat map intensity, Indicates location With location point The distance between For location Normalized activity heat.

[0029] Preferably, the steps for obtaining the classroom participation comparison data are:

[0030] Calling the classroom activity heat map, combining the image feature time series, comparing the activity heat map frequency at each time point with the historical activity data, analyzing the change in activity intensity within each time period, and generating a time point activity intensity comparison sequence;

[0031] Calling the time point activity intensity comparison sequence, determining the frequency change of the activity heat map at each time point, quantifying the activity frequency at each time point, analyzing the frequency change using the mean and standard deviation, and generating a time point frequency statistical sequence;

[0032] Call the time point frequency statistics sequence, using the formula:

[0033]

[0034] Compare the participation in each time period to obtain the classroom participation comparison data; among them, represents the classroom participation comparison data, Indicates at time The frequency of activity, represents the average activity frequency at a time point, is the total time points.

[0035] Preferably, the steps of obtaining the participation distribution are:

[0036] Calling the class participation comparison data, classifying the participation data at each time point according to the time series, analyzing the participation intensity of each time period, obtaining the participation fluctuation of the time series, and obtaining the time series participation fluctuation value;

[0037] Calling the time series participation fluctuation value, combining the participation frequency of each time period and the activity heat map data of the location point, calculating the correlation between the participation frequency and the spatial distribution of each time period, correlating the frequency with the spatial data, and generating a time and space participation correlation sequence;

[0038] The time and space participation correlation sequence is called to perform cluster analysis on the participation frequency of each time period, using the formula:

[0039]

[0040] Calculate the participation distribution of each time period to obtain the participation distribution; where, represents the participation distribution value, For the The participation frequency value of the time period, is the average participation frequency in all time periods, is the weight factor.

[0041] Preferably, the steps of obtaining the feedback suggestion set are:

[0042] Calling the engagement distribution, analyzing the engagement fluctuation in each time period, identifying key time periods that affect engagement, and generating a key time period engagement sequence;

[0043] The engagement sequence of the key time period is called, combined with historical behavior data, the engagement change pattern of the time period is compared with similar patterns in history, the differences in behavioral characteristics under each pattern are analyzed, the key behavioral characteristics that affect classroom effectiveness are identified, and a behavioral characteristic difference sequence is generated.

[0044] Preferably, the step of obtaining the feedback suggestion set further includes:

[0045] The behavior feature difference sequence is called, and according to each behavior feature combination, the formula is used:

[0046]

[0047] Generate a feedback suggestion set; where, Indicates the weight value of the suggestion in the feedback suggestion set, Behavioral characteristics The importance factor, Behavioral characteristics The frequency of occurrence, Indicates the highest frequency feature among all features.

[0048] Compared with the prior art, the advantages and positive effects of the present invention are:

[0049] In the present invention, by real-time acquisition of classroom video stream data and time series analysis of image sequence features, image features and frame information are combined to generate classified behavior labels and feature distributions, and students' classroom behaviors are captured and classified. This data screening and classification process identifies students' behavioral characteristics in the classroom and improves data real-time and responsiveness. Based on the spatial distribution parameters generated by location points and activity frequencies, the heat distribution of classroom activities in space is accurately quantified through frequency classification and location parameter processing, and the spatial visualization of classroom activities is realized, which facilitates teachers to make scene decisions according to classroom dynamics. By comparing time series and activity heat maps, the participation distribution is refined, so that teachers can obtain student participation data with more time period and location details, thereby adjusting teaching strategies and realizing personalized teaching. The analysis of participation changes is combined with feedback suggestions, and by analyzing student behavior parameters and time differences, the teaching plan design is optimized and the teaching quality is enhanced. The innovation of overall logic and participation items uses multi-level behavior analysis and dynamic data generation to provide accurate feedback, improve the intelligence and responsiveness of classroom management, ensure comprehensive monitoring and guidance of student behavior in the preschool education stage, and effectively support teachers' management and targeted teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0052] See also Figure 1 The present invention provides a technical solution: a preschool smart classroom management system based on big data includes:

[0053] The classroom behavior analysis module collects preschool classroom video stream data, extracts image sequences to analyze image features, calculates the time series of features and image frames, combines and classifies feature sequences to determine student behavior labels, combines labels with image frames to generate classroom behavior feature data, and obtains behavior feature distribution through parameter screening and classification of classroom behavior feature data;

[0054] The classroom activity heat map generation module extracts the location points and activity frequencies of the image frame based on the distribution of behavioral characteristics, generates spatial distribution parameters by combining the location points and frequency values, obtains the classroom activity distribution by parameter classification and comparison, and generates the classroom activity heat map by combining the classroom activity distribution and the image area frequency;

[0055] The classroom participation evaluation module obtains the behavioral feature data of the time series and activity heat map frequency comparison based on the classroom activity heat map, combines the heat map position and time series classification to obtain the classroom participation comparison data, extracts the comparison data and classifies it according to the activity time series to obtain the participation distribution;

[0056] The teaching feedback mechanism module extracts the participation changes and time differences of behavioral parameters based on the participation distribution, classifies the changes and behavioral parameter sequences into feedback sequences, and combines the feedback sequences with historical behavioral data to obtain a feedback suggestion set.

[0057] The steps to obtain the behavior feature distribution are:

[0058] According to the image features of each frame in the image sequence, combined with the corresponding timestamp in the frame sequence, the feature information and time parameters of each image frame are combined to form an image feature time series;

[0059] Call the image feature time series to perform correlation analysis between image features and behavior labels, match and judge the features of each frame of the image with the predefined behavior label set, select and classify image frames that meet the label standards, and generate a feature label classification sequence;

[0060] Calling the feature label classification sequence, classifying each behavior label and feature sequence according to the image frame interval, and generating a behavior feature classification sequence;

[0061] Call the behavior feature classification sequence, compare the feature similarity between the tag groups, and select parameters based on the distribution of behavior classification. The formula is:

[0062]

[0063] Get the behavior feature distribution; where, represents the distribution value of behavioral characteristics, Indicates The image feature value of the frame, Indicates The label feature value of the frame, is the inter-frame feature weight parameter.

[0064] Specifically, the features of each frame in the image sequence are combined with the time parameters of the frame sequence. By real-time monitoring of the image stream data, the associated data of the image features and the time series are obtained. The specific feature values ​​of each image frame are calculated by a real-time data processing module. Combined with the timestamp, this process includes the extraction of image features such as color, texture, and shape to ensure the time accuracy and feature integrity of each frame data, corresponding to the operation of generating the image feature time series, providing a data basis for subsequent behavior label matching and feature classification, and generating the image feature time series.

[0065] In the image sequence, the specific operation is to capture the classroom situation in real time through a high-definition camera, and standardize each frame of the image, including grayscale conversion and noise reduction, in order to extract features. The extracted features include edges, corners and texture information. These features are combined with the time series through the frame timestamp to form an independent data set. The image features of each frame correspond to the corresponding time label one by one and are stored in the image feature time series database. This database provides an accurate input data source for subsequent behavioral analysis and generates image feature time series.

[0066] When performing the association analysis between features and behavior labels, we first call the image feature time series in the form of feature vectors, use the support vector machine model to perform feature classification, and match the features of each image frame with the preset behavior labels. The matching process takes into account the multiple dimensions of the features, such as angle, size, and shape, to ensure the accuracy of the classification. Through multiple iterations, the classification boundaries are optimized to eventually form a feature label classification sequence, laying the foundation for further behavior analysis and generating a feature label classification sequence.

[0067] Further processing of the feature label classification sequence includes a more detailed division of each type of behavioral features. This process is achieved through a clustering algorithm. The specific method is to calculate the feature similarity within each type of label, and by setting a similarity threshold, classify the features with a similarity higher than this threshold, so that the features in the same category are more concentrated. The parameters considered in the classification process include but are not limited to the geometric distance of the features and the color histogram comparison. By gradually adjusting the threshold, the classification results are optimized. These operations are performed based on the classification data generated in the previous steps to generate a behavioral feature classification sequence.

[0068] formula The calculation process:

[0069] in, represents the distribution value of behavioral characteristics, Indicates The image feature value of the frame, Indicates The label feature value of the frame, is the inter-frame feature weight parameter. The weight parameter is set as an adjustment coefficient for the behavior stability according to the inter-frame behavior change.

[0070] There are 3 frames, , , .

[0071] First calculate the difference of each frame and multiply it by the weight:

[0072]

[0073] Then sum:

[0074] The denominator is the square root of the sum of the squares of the features in each frame:

[0075] ,The result shows that after considering the weight adjustment, the ,distribution value of the behavior feature is 0.372, indicating that the ,difference between the behavior features in different frames is small and the ,stability is high. This numerical result shows that if the distribution value of the behavior ,feature is greater than a certain threshold, it may indicate that the classroom ,behavior features have obvious clustering or abnormality.

[0076] The steps to obtain spatial distribution parameters are:

[0077] Call the results of classroom behavior distribution, analyze the position points and activity frequency of each image frame, assign weights to each position point according to the frequency of occurrence, calculate the weighted frequency value of each position point, and generate a weighted position frequency sequence;

[0078] Call the weighted position frequency sequence, normalize the weighted frequency value of each position point, and generate a normalized position frequency sequence;

[0079] Call the normalized position frequency sequence, combine the size and layout of each activity area in the classroom, and calculate the spatial distribution parameters of the position frequency of each activity area using the formula:

[0080]

[0081] Generate spatial distribution parameters; among them, represents the spatial distribution parameter, Indicates The normalized frequency value of the position, Indicates the visual impact value of the position in the spatial layout, Represents the maximum value of the normalized frequency values ​​of all positions.

[0082] Specifically, the process of generating classroom behavior distribution results begins with a detailed analysis of the location points and their activity frequencies in each image frame. The data of the location points are acquired through real-time monitoring by cameras, while the activity frequencies are determined by analyzing the frequency of students' movements and interactions. Each location point is weighted according to its activity frequency, and the specific weight coefficient is adjusted based on the activity concentration of past data statistics and the actual application effect of the classroom layout. The weight calculation formula is determined based on the strategic importance of the location and the statistical data of the frequency of use. For example, areas that are more frequently used for presentations or group activities will be assigned higher weights. This weight adjustment coefficient is dynamically adjusted with changes in the type of teaching activities and student participation to ensure dynamic reflection and real-time nature of the data.

[0083] The process of generating the normalized position frequency series involves standardizing the weighted frequency values ​​through mathematical formulas, so that the data expression of all location points is more uniform in spatial distribution. The normalized frequency value of each location is calculated based on the actual monitored data, rather than a preset value, ensuring the practicality and accuracy of the data.

[0084] formula The calculation process:

[0085] By normalizing the frequency of each position The visual impact value of the location Multiply, then sum the results of all positions to get the spatial distribution parameters , where the visual impact value It is determined based on the visual importance and physical location of the location, for example, the value of the front of the classroom will be higher than the back;

[0086] A classroom has three location points, whose normalized frequency values ​​are 0.2, 0.5, and 0.3, and whose visual impact values ​​are 1, 2, and 1.5. For example, if the maximum normalized frequency value is 0.5, the calculation process is:

[0087]

[0088] The value reflects the total visual impact of classroom space activities. A value greater than a certain threshold, such as 3, indicates that high activity and high visibility areas are well distributed. If it is lower than this value, the classroom layout may need to be adjusted to enhance visual effects and activity efficiency.

[0089] The steps to obtain the classroom activity heat map are:

[0090] Call the spatial distribution parameters, analyze the location points and activity frequency of each activity area in the classroom, map the activity level of each location point in the classroom, calculate the activity heat of each location point, and generate an activity heat sequence;

[0091] Call the activity heat sequence, combine the space size of each position, use normalization processing to adjust the activity heat value, and generate a normalized activity heat sequence;

[0092] Call the normalized activity heat sequence through the formula:

[0093]

[0094] The heat distribution of each location point is converted into color changes on the image to generate a classroom activity heat map; Indicates at location The heat map intensity, Indicates location With location point The distance between For location Normalized activity heat.

[0095] Call the spatial distribution parameters, analyze the location points and activity frequency of each activity area in the classroom, use these data to map the activity level of each location point in the classroom, calculate the activity heat of each location point, generate an activity heat sequence, adjust the heat value according to the actual spatial size of each activity area, perform normalization to ensure that the heat of each area matches the actual activity intensity, and apply the adjusted heat sequence to the next step of generating a classroom activity heat map.

[0096] The activity heat sequence is called, and the activity heat value is adjusted through normalization processing in combination with the actual space size of each location to ensure that each part of the heat map can truly reflect the actual intensity of classroom activities and generate a normalized activity heat sequence. This sequence reflects the activity heat of each location after adjustment according to its actual space size, providing an accurate data basis for the final generation of the heat map.

[0097] formula The calculation process:

[0098] In the generation of classroom activity heat maps, Representative in position The heat map intensity, It's location With location point The distance between them can be calculated using the Euclidean distance formula. It's location The normalized activity heat is obtained by combining the activity frequency with the spatial size of the area. The size of a classroom is 10m x 10m, and the calculation of activity frequency may involve the number of activities captured in the area within one minute. By normalizing the weighted sum of activity heat and distance, a heat map reflecting classroom dynamics can be generated;

[0099] Location and The activity heats are 0.3 and 0.7 respectively, and the distance is 2.83 meters. The heat map intensity at this location can be calculated as follows:

[0100]

[0101] If the heatmap intensity If it is greater than 0.5, it means that the area is very active, and if it is less than 0.5, it means that the activity is average.

[0102] The steps to obtain classroom participation comparison data are as follows:

[0103] Call the classroom activity heat map, combine it with the image feature time series, compare the activity heat map frequency at each time point with the historical activity data, analyze the change of activity intensity in each time period, and generate a time point activity intensity comparison sequence;

[0104] Call the time point activity intensity comparison sequence, determine the frequency change of the activity heat map at each time point, quantify the activity frequency at each time point, analyze the frequency change using the mean and standard deviation, and generate a time point frequency statistical sequence;

[0105] Call the time point frequency statistics sequence and use the formula:

[0106]

[0107] Compare the participation in each time period to obtain the classroom participation comparison data; among them, represents the classroom participation comparison data, Indicates at time The frequency of activity, represents the average activity frequency at a time point, is the total time points.

[0108] Specifically, the results of classroom behavior distribution are called to perform activity area analysis. The location points and activity frequencies of each activity area in the classroom are mapped to an activity heat calculation. The heat of each location point is weighted by the specific activity frequency data. This weighting process involves multiplying the frequency of occurrence of each location point by a preset activity weight. This weight is calculated based on the average value of the activity frequency of the location point in historical data. In this way, the activity heat not only reflects the frequency, but also takes into account the actual use of the location. Such calculations ensure the practical applicability and operability of the data, calculate the activity heat of each location point, and generate an activity heat sequence.

[0109] Normalization is performed based on the activity heat sequence. In actual implementation, the activity heat sequence of each location is adjusted to reflect its actual proportion in the classroom space. The normalization process includes dividing the activity heat of each location by the maximum heat value in the sequence, which ensures that the activity heat of all locations is between 0 and 1, making the heat map more visually uniform. This processing not only improves the visualization of the data, but also enhances the data interpretability. Through this series of operations, a normalized activity heat sequence is generated, which further lays the foundation for the creation of a heat map.

[0110] formula The calculation process:

[0111] Indicates location The heat map intensity is calculated by exponentially decaying the activity heat of the location point and the distance weight between the locations. It's location With location point The Euclidean distance between For location The normalized activity heat is adjusted based on the maximum activity heat. Consider three locations, the normalized heat of location 1 is 0.5, location 2 is 0.3, and location 3 is 0.2, and the distances between these locations and the reference location are 1, 2, and 3, respectively. The calculation formula can be expressed as:

[0112]

[0113] Calculated The value reflects the intensity of activity at a given location. The result shows the actual activity intensity at a given location. If the value is greater than 0.5, it means that the activity is more frequent, and if it is less than 0.5, it means that the activity is less frequent.

[0114] The steps to obtain the participation distribution are:

[0115] Call the classroom participation comparison data, classify the participation data at each time point according to the time series, analyze the participation intensity of each period, obtain the participation fluctuation of the time series, and obtain the time series participation fluctuation value;

[0116] Call the time series participation fluctuation value, combine the participation frequency of each time period and the activity heat map data of the location point, calculate the correlation between the participation frequency and spatial distribution of each time period, associate the frequency with the spatial data, and generate a time and space participation correlation sequence;

[0117] Call the time and space participation correlation sequence, and perform cluster analysis on the participation frequency in each time period, using the formula:

[0118]

[0119] Calculate the participation distribution of each time period to obtain the participation distribution; where, represents the participation distribution value, For the The participation frequency value of the time period, is the average participation frequency in all time periods, is the weight factor.

[0120] Specifically, call the classroom participation comparison data, classify the participation data at each time point according to the time series, organize the participation data at each time point according to the participation frequency distribution of the time period, and gradually construct the participation curves of different time periods. The difference in participation intensity of each time period is extracted through the trend changes of the curve data and a time series participation fluctuation value is formed. The fluctuation value is then used to analyze the difference in participation intensity at each time point, specifically including using these intensity differences to identify high and low participation frequency points and locate the participation trend distribution in active areas in the classroom, and generate a time series participation fluctuation value.

[0121] The time series participation fluctuation value is called, and the spatial frequency characteristics of the classroom are correlated with the participation frequency of each time period according to the participation frequency of each time period and the activity heat map data of the location point. The correlation degree between the participation frequency and the spatial distribution data of each time period is calculated to identify the location participation hotspots. The activity frequency and location heat map are used for joint calculation, and the activity frequency of different spatial locations is refined in the time dimension, so as to gradually complete the data sorting of the time-space participation correlation sequence. The correlation sequence data can represent the spatiotemporal distribution of the participation frequency.

[0122] formula The calculation process:

[0123] represents the participation distribution value in the entire time period, Indicates The participation frequency value of a time period can be obtained by real-time monitoring of the class participation frequency. The average value is calculated by counting the frequency values ​​of participation activities in each time period. It represents the average participation frequency of all time periods. To obtain this average, you need to combine the frequency values ​​of each time period and take the sum and divide it by the number of time periods to get the average frequency value of all time periods. is the weight factor. The weight is set according to the standard deviation of the importance of the class period and the frequency of the activity. The variance fluctuation of the class participation in the time period is used as the basis for weight adjustment. The derivation process of the formula is as follows: First, the frequency deviation value of each time period is calculated, that is, , multiply the deviation value of each period by its weight and sum them up. The denominator is the square root of the sum of the squares of the frequency values ​​of each period. By calculation, we can get The distribution of participation in each time period represented;

[0124] The class is divided into five time periods, with participation frequency values and weight The specific assignments are as follows:

[0125] Time period 1: ,

[0126] Time period 2: ,

[0127] Time period 3: ,

[0128] Time period 4: ,

[0129] Time period 5: ,

[0130] Average participation frequency The calculation process is as follows:

[0131]

[0132] Calculation process:

[0133] Calculate the absolute value of the frequency value deviation in each time period and multiply it by the weight:

[0134]

[0135]

[0136]

[0137]

[0138]

[0139] Add these values ​​to get the sum of the numerator:

[0140]

[0141] Calculate the denominator, that is, the square root of the sum of the squares of the frequency values ​​in each time period:

[0142]

[0143]

[0144]

[0145] Substitute the formula into the calculation value:

[0146]

[0147] Calculated , which indicates the degree of distribution deviation of classroom participation in different time periods. This result shows that the difference in participation frequency in each time period is small and close to the average level. If the value is close to 0, it means that the participation is uniform in each period. If it is greater than 0.5, it means there is a significant difference in participation.

[0148] The steps for obtaining the feedback suggestion set are:

[0149] Call the engagement distribution, analyze the engagement fluctuations in each time period, identify the key time periods that affect engagement, and generate engagement sequences for key time periods;

[0150] Call the key time period participation sequence, combine it with historical behavior data, compare the participation change pattern of the time period with the same type in history, analyze the differences in behavioral characteristics under each mode, identify the key behavioral characteristics that affect the classroom effect, and generate a behavioral characteristic difference sequence;

[0151] Call the behavior feature difference sequence, according to each behavior feature combination, use the formula:

[0152]

[0153] Generate a feedback suggestion set; where, Indicates the weight value of the suggestion in the feedback suggestion set, Behavioral characteristics The importance factor, Behavioral characteristics The frequency of occurrence, Indicates the highest frequency feature among all features.

[0154] Specifically, call up the participation distribution, analyze the fluctuation of participation in each time period, compare the changes in participation in each time period with the time frequency data, extract the time intervals with lower participation, mark the time periods with low frequency participation as the focus of attention, and identify the key time periods that have a greater impact on the overall class participation. Classify the importance of the time periods by referring to the frequency values ​​of the key time periods, record the participation frequency and fluctuation range of each time period, separately classify the low-frequency time period data that affects class participation, and record the corresponding specific time points to generate a key time period participation sequence.

[0155] Call the key time period participation sequence, combine it with the time series pattern in the historical behavior data, compare the low-frequency participation time period data in the current class with the historical data item by item, analyze the changes in each behavior feature according to the time distribution pattern of the behavior feature, select the historical behavior feature pattern in the same time period, and extract the main behavior feature combination of the key time period with low participation and the historical data in the same time period by statistically comparing the frequency differences of each feature, mark it as the behavioral feature that affects the effect in the class, and generate a behavior feature difference sequence.

[0156] formula The calculation process:

[0157] Indicates the suggestion weight value of the feedback suggestion set; Behavioral characteristics The importance factor depends on the influence of the behavior characteristics on the overall effect of the class. The influence coefficient is obtained through classroom observation and data analysis. Indicates behavioral characteristics The frequency of occurrence is calculated by statistically calculating the number of occurrences of each behavioral feature in different time periods; It represents the maximum value of all behavioral feature frequencies and is used for standardization so that the frequency value of each behavioral feature is proportionally adjusted relative to the highest frequency feature frequency.

[0158] In the characteristic frequency data of a certain classroom, (Indicates that this behavior characteristic has a greater impact on the overall effect), , and the maximum frequency , the calculation formula is:

[0159]

[0160] The calculation results show that the recommended weight value of behavior feature 1 is 0.4, which means that it occupies a certain proportion in the generated feedback suggestion set, which is used to identify the importance and improvement needs of this behavior feature.

Claims

1. A preschool smart classroom management system based on big data, characterized in that: The system comprises: The classroom behavior analysis module collects preschool classroom video stream data, extracts image sequences to analyze image features, calculates and obtains the time series of features and image frames, combines and classifies feature sequences to determine student behavior labels, combines labels with image frames to generate classroom behavior feature data, and obtains behavior feature distribution through parameter screening, classification and judgment of the classroom behavior feature data; A classroom activity heat map generation module, based on the distribution of the behavioral characteristics, extracts the position points and activity frequencies of the image frames, generates spatial distribution parameters by combining the position points and frequency values, obtains the classroom activity distribution by parameter classification and comparison, and generates a classroom activity heat map by combining the classroom activity distribution and the image area frequency; A class participation evaluation module, based on the class activity heat map, obtains time series and activity heat map frequency comparison behavior feature data, combines heat map position and time series classification to obtain class participation comparison data, extracts the class participation comparison data and classifies it according to the activity time series to obtain participation distribution; A teaching feedback mechanism module extracts the participation changes and time differences of the behavior parameters based on the participation distribution, classifies the changes and behavior parameter sequences into feedback sequences, and combines the feedback sequences with historical behavior data to obtain a feedback suggestion set; The steps of obtaining the behavior feature distribution are: According to the image features of each frame in the image sequence, combined with the corresponding timestamp in the frame sequence, the feature information and time parameters of each image frame are combined to form an image feature time series; Calling the image feature time series, performing association analysis between image features and behavior labels, matching and judging the features of each frame of the image with a predefined behavior label set, selecting and classifying image frames that meet the label standard, and generating a feature label classification sequence; Calling the feature label classification sequence, classifying each behavior label and feature sequence according to the image frame interval, and generating a behavior feature classification sequence; The behavior feature classification sequence is called, and the feature similarity between the tag groups is compared. The parameter screening is performed in combination with the distribution of the behavior classification, using the formula: Get the behavior feature distribution; where, represents the distribution value of behavioral characteristics, Indicates The image feature value of the frame, Indicates The label feature value of the frame, is the inter-frame feature weight parameter; The steps for obtaining the spatial distribution parameters are: Call the results of classroom behavior distribution, analyze the position points and activity frequency of each image frame, assign weights to each position point according to the frequency of occurrence, calculate the weighted frequency value of each position point, and generate a weighted position frequency sequence; Calling the weighted position frequency sequence, normalizing the weighted frequency value of each position point, and generating a normalized position frequency sequence; The normalized position frequency sequence is called, and the spatial distribution parameters of the position frequency of each activity area are calculated in combination with the size and layout of each activity area in the classroom, using the formula: Generate spatial distribution parameters; among them, represents the spatial distribution parameter, Indicates The normalized frequency value of the position, Indicates the visual impact value of the position in the spatial layout, Represents the maximum value of the normalized frequency values ​​of all positions; The steps for obtaining the classroom activity heat map are: Calling the spatial distribution parameters, analyzing the location points and activity frequencies of each activity area in the classroom, mapping the activity level of each location point in the classroom, calculating the activity heat of each location point, and generating an activity heat sequence; The activity heat sequence is called, and the activity heat value is adjusted by normalization processing in combination with the spatial size of each position to generate a normalized activity heat sequence; Call the normalized activity heat sequence, through the formula: The heat distribution of each location point is converted into color changes on the image to generate a classroom activity heat map; Indicates at location The heat map intensity, Indicates location With location point The distance between For location Normalized activity heat of The steps for obtaining the classroom participation comparison data are as follows: Calling the classroom activity heat map, combining the image feature time series, comparing the activity heat map frequency at each time point with the historical activity data, analyzing the change in activity intensity within each time period, and generating a time point activity intensity comparison sequence; Calling the time point activity intensity comparison sequence, determining the frequency change of the activity heat map at each time point, quantifying the activity frequency at each time point, analyzing the frequency change using the mean and standard deviation, and generating a time point frequency statistical sequence; Call the time point frequency statistics sequence, using the formula: Compare the participation in each time period to obtain the classroom participation comparison data; among them, represents the classroom participation comparison data, Indicates at time The frequency of activity, represents the average activity frequency at a time point, is the total time points; The steps for obtaining the participation distribution are: Calling the class participation comparison data, classifying the participation data at each time point according to the time series, analyzing the participation intensity of each time period, obtaining the participation fluctuation of the time series, and obtaining the time series participation fluctuation value; Calling the time series participation fluctuation value, combining the participation frequency of each time period and the activity heat map data of the location point, calculating the correlation between the participation frequency and the spatial distribution of each time period, correlating the frequency with the spatial data, and generating a time and space participation correlation sequence; The time and space participation correlation sequence is called to perform cluster analysis on the participation frequency of each time period, using the formula: Calculate the participation distribution of each time period to obtain the participation distribution; where, represents the participation distribution value, For the The participation frequency value of the time period, is the average participation frequency in all time periods, is the weight factor; The steps of obtaining the feedback suggestion set are: Calling the engagement distribution, analyzing the engagement fluctuation in each time period, identifying key time periods that affect engagement, and generating a key time period engagement sequence; Calling the key time period participation sequence, combining historical behavior data, comparing the participation change pattern of the time period with similar patterns in history, analyzing the differences in behavioral characteristics under each pattern, identifying key behavioral characteristics that affect classroom effects, and generating a behavioral characteristic difference sequence; The step of obtaining the feedback suggestion set further includes: The behavior feature difference sequence is called, and according to each behavior feature combination, the formula is used: Generate a feedback suggestion set; where, Indicates the weight value of the suggestion in the feedback suggestion set, Behavioral characteristics The importance factor, Behavioral characteristics The frequency of occurrence, Indicates the highest frequency feature among all features.

Images