Education data processing method and verification system based on privacy calculation

Abstract

The application provides an education data processing method and verification system based on privacy calculation, relates to the technical field of digital data processing, and can accurately capture the student's line of sight and fixed time length by adopting a common camera combined with an embedded eye movement tracking algorithm, overcoming the recognition error problem of traditional expression recognition in a complex environment, and realizing accurate evaluation of the student's real attention state. By constructing a comprehensive attention evaluation index, the student's attention level is effectively quantified, and based on the mathematical model of normalization processing and deviation degree coefficient, a more scientific and reasonable attention score is provided, which can provide real-time feedback and dynamically adjust the teaching strategy and optimize the teaching effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital data processing, in particular to an education data processing method and verification system based on privacy computing. BACKGROUND

[0002] As an important part of video analysis technology, by taking classroom video data, the behavior, attention distribution and interaction of students are monitored and analyzed in real time, aiming to improve the teaching quality and learning effect. At the same time, as an important means to protect data security and privacy, privacy computing has been widely concerned and applied in data processing in recent years. Privacy computing ensures that sensitive information is not leaked in the process of data analysis and processing through encryption algorithms, multi-party secure computing and other technical means, and realizes the effective use of data under the premise of protecting privacy.

[0003] In the prior art, a university personal information anonymization protection system with perfect structure, simple and safe operation and reliable implementation method is disclosed, which is named CN118036075A. The system aims to solve the key problems of information leakage and fuzzy authority division in the process of student personal information management in current universities. It is developed by using Python and PHP programming languages, which realizes the effective protection of student personal information and the accurate control of authority. The system mainly includes student personal information portal, university department work platform, student personal information database and off-campus database, and is equipped with an efficient anonymization processing module. It can desensitize sensitive information according to the "department authority setting of university personal information system", and protect personal privacy to the maximum extent without hindering data analysis and use.

[0004] The existing deficiencies are:

[0005] Insufficient attention monitoring accuracy: traditional expression recognition and voice analysis methods are difficult to accurately reflect the real attention state of students, and are greatly affected by environmental light, camera angle and noise, with significant error;

[0006] Lack of privacy protection: the existing data processing technology does not fully consider the privacy protection of students in the education scene, and personal behavior data is easy to be leaked or misused;

[0007] Insufficient real-time and scientific nature: there is a lack of effective real-time feedback mechanism and scientific quantitative index in the attention evaluation process, which makes it difficult to dynamically adjust the teaching strategy and optimize the teaching effect.

[0008] The above information disclosed in the above background art section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0009] The present application aims to provide a privacy computing-based educational data processing method and verification system to solve the problems raised in the background.

[0010] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0011] The privacy computing-based educational data processing method includes the following specific steps:

[0012] Step S1: Collect video data of the target classroom, and divide the student area in the video data into regions through a preset segmentation algorithm to generate a corresponding target student region sequence set;

[0013] Step S2: Determine whether the pupil center position of a student in each target student region is within a preset range according to a preset line-of-sight stable area of a teaching screen, thereby extracting the line-of-sight fixation duration data of each student within the line-of-sight stable area;

[0014] The line-of-sight fixation duration data includes a line-of-sight deviation degree coefficient and a total line-of-sight fixation duration;

[0015] Step S3: Analyze the line-of-sight fixation duration data of each target student region to obtain line-of-sight deviation frequency and line-of-sight deviation degree distribution results, respectively;

[0016] The line-of-sight deviation degree distribution results are used to divide the target student region sequence set into a plurality of deviation region subsets with different deviation degrees;

[0017] Step S4: Receive the line-of-sight deviation frequency and total line-of-sight fixation duration of each deviation region subset for comprehensive analysis to construct a comprehensive attention evaluation index of each deviation region subset;

[0018] Step S5: Combine the teaching quality feedback data and the comprehensive attention evaluation index of each deviation region subset for comprehensive analysis to generate a verification and optimization model, which is used to optimize the preset range of the line-of-sight stable area in the teaching screen and develop a strategy for verifying the quality and applicability of related teaching content.

[0019] A privacy computing-based educational data processing verification system, which is used to execute the privacy computing-based educational data processing method and specifically includes:

[0020] The region division module is used to collect video data of the target classroom, and divide the student area in the video data into regions through a preset segmentation algorithm to generate a corresponding target student region sequence set;

[0021] Judgment and extraction module: It is used to determine whether the pupil center position of each student in each target student area is within the preset range based on the preset stable gaze area of ​​the teaching screen, so as to extract the gaze fixation time data of each student in the stable gaze area.

[0022] The data on the duration of fixed gaze includes the coefficient of gaze deviation and the total duration of fixed gaze.

[0023] Deviation Zone Division Module: This module analyzes the fixed gaze duration data for each target student area, and obtains the distribution results of gaze deviation frequency and gaze deviation degree.

[0024] The results of the line-of-sight deviation distribution are used to divide the target student region sequence set into multiple deviation region subsets with different degrees of deviation;

[0025] Index construction module: used to receive the gaze deviation frequency and total gaze fixed duration of each deviation region subset for comprehensive analysis in order to construct a comprehensive attention evaluation index for each deviation region subset;

[0026] Strategy Validation Module: This module combines teaching quality feedback data from each deviation region subset with the comprehensive attention evaluation index for comprehensive analysis, generating a validation and optimization model to optimize the preset range of the stable viewing area on the teaching screen and formulate strategies to validate the quality and applicability of related teaching content.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. By using a regular camera combined with an embedded eye-tracking algorithm, it can accurately capture the student's gaze direction and fixed duration, overcoming the recognition error problem of traditional facial expression recognition in complex environments, and realizing accurate assessment of the student's true attention state;

[0029] 2. By constructing a comprehensive attention evaluation index, students' attention levels are effectively quantified. Based on a mathematical model of normalization and deviation coefficient, a more scientific and reasonable attention score is provided, which can provide real-time feedback and dynamically adjust teaching strategies and optimize teaching effectiveness.

[0030] 3. By utilizing the privacy computing technology of the Trusted Execution Environment (TEE), sensitive information is encrypted and protected during data collection and processing, ensuring a high level of security for student privacy and complying with relevant laws and regulations. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0032] Figure 2 This is a block diagram of the overall system modules of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0034] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0035] Example 1:

[0036] Please see Figure 1 The present invention provides a technical solution:

[0037] The method for processing educational data based on privacy-preserving computation includes the following steps:

[0038] Step S1: The edge server collects video data from the target classroom and divides the area where the students are located in the video data into regions using a preset segmentation algorithm to generate a corresponding target student region sequence set;

[0039] To further explain, the specific implementation steps are as follows:

[0040] 1.1) A high-definition camera is fixedly installed in the target classroom. The high-definition camera is required to have a resolution of not less than 1920×1080 pixels and a frame rate of not less than 30 frames / second. The installation position of the camera is adjusted to ensure panoramic coverage of the target student area.

[0041] 1.2) The image processing module is used to preprocess the acquired video data; specifically, the Gaussian mixture model of the background segmentation algorithm is used to separate the areas that are not related to students from the video data, and the seating distribution map of the target classroom is delineated based on the segmentation results.

[0042] 1.3) Based on the pre-defined seating distribution map, each seat is uniquely identified by a digital identifier according to the actual position of the students in the target classroom. The physical position of the seat is then mapped to the target student area in the video data through the digital identifier, so as to clarify the area range corresponding to each student in the video data.

[0043] The target student region sequence set is denoted as {1,2,…,i,…,n}, where i represents the index label of the target student region and n is the total number of target student regions;

[0044] 1.4) The video data of the target student area is transmitted to the edge server through the real-time monitoring system. The video data is then encrypted using an encryption module on the edge server and transmitted to the Trusted Execution Environment (TEE) through a secure data channel for data privacy processing and analysis.

[0045] Encrypted data transmission and processing within a Trusted Execution Environment (TEE) ensure privacy protection and data security, meeting the requirements of privacy-based computation-based de-identification technology.

[0046] Step S2: The edge server determines whether the pupil center position of each student in each target student area is within the preset range based on the preset stable gaze area of ​​the teaching screen, thereby extracting the gaze fixation time data of each student in the stable gaze area.

[0047] The data on the duration of fixed gaze includes the coefficient of gaze deviation and the total duration of fixed gaze.

[0048] To further explain, the specific implementation steps are as follows:

[0049] The pupil center position was detected using an embedded eye-tracking algorithm to detect the center position of the student's face and pupil in each target student area;

[0050] The embedded eye-tracking algorithm includes a face detection module and a pupil detection algorithm;

[0051] 2.1) Obtain video data of the target student region isolated by the Trusted Execution Environment (TEE), call the face detection module within the TEE to locate the face of the target student region i, and use the pupil detection algorithm to determine the center position of the pupil of the target student region i.

[0052] In this embodiment, the face detection module uses a trained convolutional neural network model MTCNN; the pupil detection algorithm uses a Haar cascade or deep learning eye-tracking detection model.

[0053] Description of the face detection module: It is used to locate student faces in the target student area as the basis for further pupil detection; in this embodiment, a trained convolutional neural network model MTCNN (Multi-task Cascaded Convolutional Networks) is used, which is good at simultaneously detecting the coordinates of face bounding boxes and key points (such as eyes, nose, mouth, etc.).

[0054] The MTCNN face detection model consists of three cascaded networks, each progressively narrowing the search range and improving detection accuracy:

[0055] P-Net (Proposal Network): Performs a fast, coarse-grained scan of the input image to generate a set of candidate face regions.

[0056] The output candidate region represents the possible locations where a face might appear.

[0057] R-Net (Refinement Network): Further filters and precisely locates the candidate regions generated by P-Net, removes noise, and adjusts the bounding boxes.

[0058] Outputs the face boundaries after preliminary optimization.

[0059] O-Net (Output Network): Performs a final refinement on the results generated by R-Net, while also predicting key facial features such as the eyes;

[0060] The final output includes accurate face bounding boxes and key point coordinates.

[0061] The operation steps are as follows:

[0062] Input video data: Obtain video frame data of the target student region from the Trusted Execution Environment (TEE) and provide it to the MTCNN model.

[0063] Model processing: Each input frame of image is processed step by step through the three networks of MTCNN to obtain the face bounding box and key point coordinates.

[0064] Output: The face bounding box and key points provide candidate locations for the next step of pupil detection.

[0065] For pupil detection algorithms, Haar cascades or deep learning eye-tracking detection models are used. The specific logic includes:

[0066] Haar cascade detection: Haar cascade is a method for detecting target regions by training a cascade classifier. It is suitable for lightweight devices, is fast, but has limited accuracy.

[0067] Based on the coordinates of key eye points output by the face detection module, the detection area is limited to the local area of ​​the eyes;

[0068] Haar cascade detection is employed: a pre-trained eye classifier is used to scan local candidate regions, detect the bounding box of the small region where the pupil is located, and calculate the coordinates of the center point as the pupil center;

[0069] 2.2) Based on optical detection using an infrared camera, a pupil tracking algorithm is used to extract the pupil center position of each frame in the video data corresponding to the target student region i;

[0070] The pupil center position of the target student region i is represented by coordinates (x...). p ,y p );

[0071] The coordinates of the detected pupil center position are calculated and recorded in real time; in this embodiment, the pixel values ​​on the teaching screen are used to represent the pupil center position.

[0072] It should be noted that: "Pixel values ​​on the screen" refers to designing a coordinate system to locate the teaching screen in pixels, dividing the display area of ​​the teaching screen into a two-dimensional pixel grid, and then mapping the coordinates of the pupil center position to the grid coordinate system of the teaching screen.

[0073] 2.3) Based on the location and functional settings of the teaching screen, the central area of ​​the teaching screen is pre-set as the "stable line of sight area" for the students corresponding to the target student area;

[0074] The preset range of the stable line of sight area is defined as a circular area with a radius of d centered on the center of the teaching screen;

[0075] The radius of the stable line-of-sight region is defined as d ≤ 50 pixels; and d is the maximum allowable deviation radius of the stable line-of-sight region.

[0076] 2.4) In the continuous frame sequence of video data corresponding to the target student region i, the pupil center position of each frame is detected in real time, and it is determined whether the pupil center position falls within the stable line of sight area.

[0077] If the center of the pupil is within the stable line of sight area, then the line of sight is considered fixed;

[0078] If the center of the pupil is not within the stable visual field, it is considered that the visual field has deviated; and the degree of visual deviation in the target student area i is analyzed.

[0079] Set a preset gaze fixation time window for the students corresponding to the target student area i. Within the preset gaze fixation time window, record the total number of gaze fixations for the target student area i as J.

[0080] When the center of the pupil is outside the stable visual area, a deviation event is recorded, and the total number of deviation events is counted as N1, where N1≤J;

[0081] In this embodiment, the result of the visual deviation degree of the target student area i is represented as the visual deviation degree coefficient α. i ;

[0082] The line-of-sight deviation coefficient is used to quantitatively describe the degree of deviation of the pupil center position from the stable line-of-sight area;

[0083] Determine (x) p ,y p (x) represents the coordinates of the center of the pupil; c ,y c () represents the center coordinates of the stable line-of-sight region;

[0084] The formula for calculating the line-of-sight deviation coefficient is defined as follows:

[0085]

[0086] Where, α i It is the coefficient of visual deviation from the target student area i;

[0087] Molecular part (x) p,i -x c,i ) 2 +(y p -y c ) 2 It is the Euclidean distance, used to calculate the actual deviation distance between the center of the pupil and the center of the stable visual area;

[0088] The denominator represents the maximum permissible deviation radius of the stable line-of-sight region. By dividing by d, the actual deviation distance can be standardized so that different degrees of deviation can be directly compared.

[0089] α i =1 indicates that the pupil center of the target student region i is located on the boundary deviating from the radius;

[0090] α i <1 indicates that the pupil center of the target student region i is located within or near the boundary of the stable line of sight area;

[0091] α i >1 indicates that the pupil center of the target student region i is outside the stable visual area, and the degree of deviation increases with the increase of the value;

[0092] When the pupil center of the student corresponding to target student region i enters the stable gaze region, timing begins immediately. When the pupil center leaves the stable gaze region, timing stops. The time interval from start to stop is recorded and is called the gaze fixation duration T of the student corresponding to target student region i. fix,i This embodiment uses seconds as the unit for cumulative timing.

[0093] The above timing process uses frame timestamps for calculation to ensure accurate time accumulation;

[0094] 2.5) The fixed duration T of the line of sight obtained from all timings within the preset fixed time window. fix,i The results are summarized to calculate the total fixed viewing time T for students corresponding to target student region i. fix,i,raw ;T fix,i,raw The specific calculation formula is as follows:

[0095]

[0096] Where T fix,i,j This represents the duration of gaze fixation for the student corresponding to target student area i under the j-th gaze fixation, where J is the total number of gaze fixations within the preset gaze fixation time window; in this embodiment, the preset gaze fixation time window is 2 minutes.

[0097] The total line of sight to the target student area i is fixed for a duration T. fix,i,raw and the coefficient of visual deviation α i This is represented by data on the duration of fixed gaze.

[0098] It should be noted that: d≤50 pixels and T fix,i,j The design of the timing unit parameters needs to be adjusted by an expert group based on actual conditions such as classroom screen size, resolution, and student distance range to ensure the rationality of the judgment criteria.

[0099] Step S3: The edge server analyzes the fixed gaze duration data for each target student area to obtain the distribution results of gaze deviation frequency and gaze deviation degree.

[0100] The results of the line-of-sight deviation distribution are used to divide the target student region sequence set into multiple deviation region subsets with different degrees of deviation;

[0101] Further explanation: The target student regions within the off-region subset are set to exhibit a clustered pattern;

[0102] 3.1) The formula for calculating the frequency of line-of-sight deviation is defined as follows:

[0103] Among them, Fp iN1 is the frequency of gaze deviation of the student corresponding to target student region i within a preset fixed gaze time window; i J is the total number of deviation events from the target student region i; i It is the total number of times the target student's gaze is fixed in area i;

[0104] 3.2) Based on the target student region sequence set {1,2,…,i,…,n}, the edge server divides the target classroom seating distribution map into independent parts of the left region set, the middle region set, and the right region set, and analyzes the line-of-sight deviation coefficient α of the target student region within each region set. i Value distribution; Specific operation steps:

[0105] Obtain the seating distribution map of the target classroom, including the spatial coordinates of each target student area i;

[0106] Based on the spatial layout of the classroom seating chart, the classroom is divided into left, middle, and right zones; the dividing lines are determined according to the geometric center of the classroom or a pre-defined seating grouping standard. In this embodiment:

[0107] x <x center -Δ

[0108] x center -Δ≤x≤x center +Δ

[0109] x>x center +Δ

[0110] Where, x center Here is the x-coordinate of the center of the classroom seating distribution map, and Δ is the offset of the area division;

[0111] For each target student region i in the target student region sequence set {1,2,…,i,…,n}, read its spatial coordinates (x...y). i ,y i ).

[0112] If x i <x center -Δ, then the target student region i will be assigned to the left region set;

[0113] If x center -Δ≤x i ≤x center If +Δ, then the target student region i will be assigned to the middle region set;

[0114] If x i >x center If +Δ, then the target student region i will be assigned to the right region set;

[0115] Each target student region i is added to the corresponding region set according to the above determination results;

[0116] In the database of the edge server, create independent data tables or sub-databases for the left region set, the middle region set, and the right region set;

[0117] The divided target student region i and its related data α i The values ​​are stored in their respective region sets for independent analysis later.

[0118] 3.3) Calculate the line-of-sight deviation coefficient α corresponding to the left, middle, and right region sets. i The statistical characteristics include the mean and standard deviation.

[0119] Detailed operation steps:

[0120] Extract all line deviation coefficients α from the left region set, the middle region set, and the right region set. i value;

[0121] Calculate the following statistical indicators for each region set:

[0122] Mean:

[0123]

[0124]

[0125] Standard deviation:

[0126]

[0127] Among them, S L ∈{1, 2, ..., N L}, S M ∈{1, 2, ..., N M}, S R ∈{1, 2, ..., N R};N L N M and N R These represent the total number of target student regions in the left region set, the middle region set, and the right region set, respectively.

[0128] Based on the calculated mean and standard deviation, determine the line-of-sight deviation coefficient α within each region set. i The degree of concentration of values;

[0129] Generate a line-of-sight deviation coefficient α for each of the three regions: left, center, and right. i A statistical report on the distribution of values, including the mean and standard deviation;

[0130] The statistical results are stored in the distributed analysis database on the edge server for easy access and reference later.

[0131] 3.4) Apply preset clustering rules to filter and delete target student regions that do not conform to the clustering rules; to ensure that target student regions within each region set present a preset clustering pattern; specific operation steps:

[0132] Based on the statistical characteristics, the following predefined clustering rules are defined for the left region set, the middle region set, and the right region set:

[0133] Set a first preset threshold and a second preset threshold for the mean and standard deviation in the statistical characteristics, respectively;

[0134] The fuzzy hierarchical analysis method (FAHP) is used to handle the fuzziness and multi-criteria weight relationship in the decision-making process. Combining the mean and standard deviation of the data, the first preset threshold of the mean of each region set and the second preset threshold of the standard deviation of each region set are calculated and evaluated.

[0135] If the mean of each region set is less than the corresponding first preset threshold and the standard deviation is lower than the corresponding second preset threshold, then the region set is considered to have clustering.

[0136] It should be noted that the first and second preset threshold values ​​for each region set are set differently;

[0137] If the visual deviation coefficient α is located in the target student area i, i If the value is higher than the preset aggregation threshold, the target student region i is deemed not to meet the aggregation rules, and the target student region i that does not meet the aggregation rules is deleted from the corresponding region set.

[0138] The preset aggregation threshold, specifically the determination criteria, are expressed as follows:

[0139] α i >μ X +k·σ X Then delete the target student region i; μ X +k·σ X This is represented as a preset aggregation threshold;

[0140] Where X represents the left region set, the middle region set, or the right region set, and μ X σ represents the mean of the corresponding set of regions. X The standard deviation of the corresponding region set is represented by k, which is a constant controlling the threshold.

[0141] For the target student region i that does not meet the aggregation rules, perform the following operations:

[0142] From the set of regions S L ∈{1, 2, ..., N L}, S M ∈{1, 2, ..., N M}, S R ∈{1, 2, ..., N R Remove the target student region i from the list.

[0143] Update the target student region sequence set in each region set to ensure data continuity and accuracy.

[0144] 3.6) Set S of regions after the deletion operation L ′∈{1,2,…,N′ L}, S M′ ∈{1, 2, ..., N′ M}, S R ′∈{1,2,…,N' R Update the database on the edge server; S L ', S M 'and S R ′ represent the regions after deletion operations on the left, middle, and right regions, respectively; and N′ L N' M and N' R These represent the total number of target student regions in the left, middle, and right region sets after the deletion operation, respectively.

[0145] This technical solution divides the classroom seating chart into three independent areas: left, center, and right. It then analyzes the distribution of the visual deviation coefficients of target students within each area, and further applies preset clustering rules for filtering and deletion. This not only improves the detail and accuracy of data analysis but also enhances the system's understanding and management capabilities of student visual behavior in different areas. The reasons for choosing this technical solution include:

[0146] Regionalization analysis: By dividing the classroom into three independent areas—left, center, and right—it is possible to analyze students' gaze deviation behavior in more detail within each area and adapt to the possible differences in gaze behavior between different areas.

[0147] Clustering rule application: Based on the distribution of the line-of-sight deviation coefficient, preset clustering rules are applied to ensure that the target student area within the deviation region subset presents a reasonable clustering pattern, thereby improving the reliability of data analysis.

[0148] By employing quantitative clustering rules, precise screening and deletion of target student regions can be achieved, avoiding errors from subjective judgment and ensuring the objectivity and consistency of the screening process.

[0149] 3.7) Calculate SL ', S M 'and S R The coefficient of visual deviation α corresponding to the target student area i The mean of each is denoted as μ. L ',μ M 'and μ R ';

[0150] Different deviation classification criteria are set to divide the left, middle, and right region sets into low, medium, and high deviation levels. The specific logic includes:

[0151] Based on the actual teaching environment and needs, S is set L ′, S M ′ and S R The mean μ of the left, middle, and right regions represented by ′ is... L or μ M ′ or μ R The classification thresholds for ' are q1 and q2;

[0152] μ L or μ M ′ or μ R When the value of ' is in the interval (0, q1], the degree of line deviation is low deviation level;

[0153] μ L or μ M ′ or μ R When the value of ' is in the interval (q1, q2], the degree of line of sight deviation is at the medium deviation level;

[0154] μ L or μ M 'or μ R When the value of ' is greater than q2, the degree of visual deviation is classified as high deviation.

[0155] The set of regions that meet the low deviation level is denoted as the low deviation region subset, the set of regions that meet the medium deviation level is denoted as the medium deviation region subset, and the set of regions that meet the high deviation level is denoted as the high deviation region subset.

[0156] Step S4: The edge server receives the gaze deviation frequency and total gaze duration of each deviation region subset and performs comprehensive analysis to construct a comprehensive attention evaluation index for each deviation region subset;

[0157] Further explanation: The formula for calculating the Comprehensive Attention Evaluation Index is as follows:

[0158]

[0159] Among them, A ris the comprehensive attention evaluation index for the region set r. w1 and w2 are weighting coefficients, representing the importance of gaze deviation frequency and total gaze fixation time in the comprehensive evaluation, respectively, and satisfying w1 + w2 = 1; the values ​​of w1 and w2 are both within the interval (0,1); k1 is the exponential decay coefficient, used to adjust the influence of gaze deviation frequency on the comprehensive evaluation index; e is the base of the natural logarithm; ln represents the natural logarithm function; used to mitigate the contribution of gaze fixation time to the comprehensive evaluation index, avoiding excessively large evaluation index due to linear growth;

[0160] It is the average frequency of line-of-sight deviations from the region set r; R is the average total fixed duration of sight lines in the region set r; r i1 represents the total number of target student regions in the region set r; i1 represents the index of the target student region.

[0161] Frequency of gaze deviation: This reflects the frequency with which students' gazes deviate from the target student area. The higher the frequency, the easier it is for students to lose focus.

[0162] Total fixation time: This indicates the total time that students in the target student area fixate their gaze. The longer the total fixation time, the higher the degree of student concentration.

[0163] The comprehensive attention evaluation index A of the set of regions r is set. r The range of its value is within the interval (0,1);

[0164] When A r The closer the value is to 0, the higher the average frequency of students' gaze deviation within the region set r, and the shorter the average total gaze fixation time; this indicates that students' attention is more scattered within the region set r; the system can trigger corresponding teaching intervention measures, such as adjusting classroom activity content or teacher guidance methods, to improve students' attention concentration in this region.

[0165] When A r The closer the value is to 1, the lower the average frequency of students' gaze deviation and the longer the average total duration of gaze fixation within region set r; indicating a higher level of concentration among students within region set r. The system can record this region as a high-attention-concentration area, serving as an important reference for future teaching optimization and resource allocation.

[0166] In this embodiment, A is initially set. r The low, middle, and high intervals are (0, 0.3), [0.3, 0.65), and [0.65, 1), respectively.

[0167] When A rWhen the value is in the low range (0, 0.3), it indicates that the students' attention is more than 70% distracted within the region set r.

[0168] When A r When the value is in the middle interval [0.3, 0.65), it indicates that the students' attentional distraction level in the region set r is between 30% and 70%.

[0169] When A r When the value is in the high interval [0.65, 1), it indicates that the students' attentional distraction level in the region set r is below 30%.

[0170] Add: Exponential decay term Decrease, leading to A r The value decreased;

[0171] This indicates that students within region set r are becoming increasingly distracted, and the system needs to take measures to improve their concentration.

[0172] Add: Corresponding to A r An increase in the value indicates that the students' attention span within the region set r has improved, and the system can consider the teaching effect in that region to be good, thus maintaining the current teaching strategy.

[0173] Increasing the weighting coefficient w1: improves Increase the weight in the overall evaluation. For A r The impact;

[0174] Increasing the exponential decay coefficient k1 will accelerate the process. For A r The negative impact of this made A at a higher level r The problem needs to be addressed quickly. The penalties for frequent eye deviations need to be strengthened to encourage the system to pay more attention to areas where students' attention is frequently diverted.

[0175] A r The formula achieves non-linear data fusion by combining exponential and logarithmic functions. This effectively suppresses extreme negative evaluations caused by high-frequency deviation events while moderately amplifying the positive impact of fixed viewing time, thus achieving a balance in the overall evaluation.

[0176] In this embodiment, the frequency of gaze deviation and the total fixed gaze duration correspond to the student's inattentive and focused states in the classroom, respectively. The formula design reflects the interaction between the two in the actual teaching environment and quantifies them through mathematical means, enabling the system to accurately capture and evaluate the student's attention state.

[0177] A r The exponential decay function in the formula Similar to the decay process in physics, this means that as the frequency of gaze deviation increases, the attention evaluation index decreases exponentially. And the logarithmic function... This reflects the trend of a slow increase in the attention evaluation index as the duration of fixed gaze increases, allowing the comprehensive index to respond smoothly to changes in different parameters.

[0178] Step S5: The edge server combines the teaching quality feedback data of each deviation region subset with the comprehensive attention evaluation index to perform a comprehensive analysis, generate a validation and optimization model, and use it to optimize the preset range of the stable viewing area in the teaching screen, and formulate strategies to validate the quality and applicability of the relevant teaching content.

[0179] Further explanation: 4.1) The edge server collects teaching quality feedback data Tq related to each off-region subset. r Teaching quality feedback data Tq r It consists of evaluation indicators based on students' understanding of the teaching content, their participation in teaching activities, and the effectiveness of teaching.

[0180]

[0181] Among them, Q i1 This represents the specific feedback indicators of teaching quality from students within the target student region i1; in this embodiment, Q i1 The degree of students' understanding of the teaching content is used as a representation.

[0182] It should be noted that Q is defined as follows: i1 The calculation formula is as follows:

[0183] Q i1 =w u1 ·U i1 +w p1 ·P i1 +w s1 ·S i1 ;

[0184] Q i1 The specific feedback indicators on teaching quality from students within the target student region i1 are composed of a weighted combination of the following three factors:

[0185] Students' level of understanding of the teaching content (U) i1 ;

[0186] Student participation in teaching activities P i1 ;

[0187] Student rating of teaching effectiveness (S) i1 ;

[0188] U i1It is calculated by combining data such as classroom test scores and completion of after-class exercises;

[0189]

[0190] Test i1 It is the classroom test score of students in the target student area i1, with a value range of (0,1);

[0191] At i1 It is the homework score of students in the target student region i1, and the value range is (0,1);

[0192] If there is no homework data, then only the classroom test score U will be used. i1 =Test i1 ;

[0193] P i1 The results were obtained by combining statistics on the number of times students answered questions, participated in discussions, and their interactive behavior indicators.

[0194]

[0195] IC i1 TI is the actual number of interactions between students in the target student area i1 during the course; TI is the total number of interactions that can be participated in as planned in the teaching activities.

[0196] S i1 This data is obtained through questionnaires, rating scales (e.g., 1 to 5 points), or course satisfaction feedback.

[0197]

[0198] RS i1 It is the actual rating of the teaching effectiveness by students within the target student area i1;

[0199] Se min and Se max This represents the minimum and maximum score; for example, if the score range is 1 to 5 points, then Se... min and Se max They are 1 and 5 respectively; after standardization, S i1 The range of values ​​is normalized to the interval (0,1);

[0200] Weighting coefficient w u1 w p1 w s1 These are the weights of each feedback sub-indicator, used to adjust the impact of each indicator on Q. i1 The degree of influence satisfies the following constraint: w u1 +w p1 +w s1 =1;

[0201] The weights are set according to the specific application and are determined by the expert group using the fuzzy hierarchical analysis method.

[0202] In this embodiment, if more emphasis is placed on students' understanding of the teaching content, w is set... u1 w p1 w s1 The values ​​are 0.5, 0.3, and 0.2 respectively.

[0203] When Q i1 As U approaches 1, i1 or P i1 or S i1 The higher; U i1 or P i1 or S i1 They are represented as follows:

[0204] The higher the students' understanding of the teaching content, the higher their enthusiasm for classroom participation, and the higher their subjective satisfaction with the teaching effect.

[0205] Edge server for Tq r Standardization is performed using min-max normalization to reduce Tq r The value range is limited to the same interval (0,1);

[0206] This embodiment initially sets up teaching quality feedback data. The low, middle, and high intervals are (0, 0.3), [0.3, 0.65), and [0.65, 1), respectively.

[0207] when When the value is in the low interval (0, 0.3), it means that the students in the region set r rate the teaching quality within 30% of the standard value;

[0208] when When the value is in the middle interval [0.3, 0.65), it indicates that the teaching quality score of students in the region set r is between 30% and 70% of the standard value;

[0209] when When the value is in the high interval [0.65, 1), it means that the teaching quality score of students in the region set r is above 70% of the standard value;

[0210] Further explanation: The edge server uses a weighted linear regression model to process the standardized teaching quality feedback data. As the dependent variable, the comprehensive attention evaluation index A r The following validation and optimization model is established using the independent variable:

[0211]

[0212] Where β0 is the intercept term of the model; β1 is A r The regression coefficient represents A r The degree of impact on teaching quality; η r This is the error term for the region set r; it reflects the deviation between the model prediction and the actual feedback data.

[0213] Edge servers assign different weights w based on the importance or data reliability of each region set r. r Specifically, η r ~N(0,σ 2 / w r );

[0214] w r σ represents the weighting coefficient of the region set r. A higher weight indicates more reliable data for that region set and a greater impact on the model. 2 The variance of the error term;

[0215] The edge server uses the least squares method to estimate the regression coefficients β0 and β1 to ensure the model fits the training data well.

[0216] Further explanation: The edge server optimizes the preset radius d of the stable viewing area in the teaching screen based on the output of the validation and optimization model; that is, it adjusts the radius d of the circular area with the center of the teaching screen as the center and a radius of d to improve students' attention concentration.

[0217] Select A from the region set r. r The set of regions with the smallest values ​​is selected as the optimization target within a preset range. The specific optimization logic is as follows:

[0218] The edge server is based on the regression coefficient β1 in the validation and tuning model. The radius d is dynamically adjusted as follows;

[0219]

[0220] Where, d new The radius of the optimized stable line-of-sight region; α is an adjustment coefficient that controls the magnitude of radius adjustment; To verify the teaching quality feedback data values ​​predicted by the optimized model;

[0221] Setting the adjustment coefficient α:

[0222] α is set to a positive number less than 1 to ensure the gradual adjustment of the radius and the stability of the system; the specific adjustment is determined by the expert group system using fuzzy hierarchical analysis.

[0223] To avoid the stable line-of-sight zone radius being too small or too large, a radius d is set. new The range of values ​​for d is min ≤d new ≤d max ;

[0224] d min In this embodiment, d represents the minimum radius of the stable line-of-sight region. min It is 0.5 meters;

[0225] d max In this embodiment, d represents the maximum radius of the stable line-of-sight region. max It is 2 meters;

[0226] After each adjustment, check d. new Is it in d? min and d max If it exceeds the range, it will be limited to the corresponding boundary value;

[0227] Further explanation: 4.2) Based on the output of the validation and optimization model, select region A from the region set r. r The strategy for validating the quality and applicability of relevant teaching content is based on the set of regions with the smallest numerical values. The specific logic includes:

[0228] The edge server obtains the teaching content quality index Cq by weighting the difficulty, interactivity, and information content. The standardized value of the teaching content quality index Cq is then denoted as... and The range of values ​​is the interval (0,1);

[0229] The formula for calculating the teaching content quality index Cq is defined as follows:

[0230] Cq = w d ·D1+w l ·I1+w f F1;

[0231] Among them, w d It is the weight of content difficulty D1; w l It is the weight of interactivity I1; w f It is the weight of information content F1; w d +w l +w f =1;

[0232] Formula D1 is

[0233] Where C2 is the complexity score, with a value range of (0,1);

[0234] B2 is the breadth score, with a value range of (0,1);

[0235] P2 is the knowledge point depth score, with a value range of (0,1);

[0236] The formula for I1 is: A2 represents the actual number of activities, indicating the frequency of student interaction with the teaching content.

[0237] A max It represents the maximum number of activities, indicating the total number of possible interactive events in the design of the instructional content.

[0238] The F1 formula is Where V2 represents the actual information volume, such as the number of words and the number of knowledge points; V max It represents the maximum information capacity of the teaching content.

[0239] This embodiment initially sets the quality indicators of the teaching content. The low, middle, and high intervals are (0, 0.3), [0.3, 0.65), and [0.65, 1), respectively.

[0240] when When the value is in the low range (0, 0.3), it indicates that the quality score of the teaching content is within 30% of the standard value;

[0241] when When the value is in the middle range [0.3, 0.65), it indicates that the quality score of the teaching content is between 30% and 70% of the standard value;

[0242] when When the value is in the high range [0.65, 1), it indicates that the quality score of the teaching content is above 70% of the standard value;

[0243] Furthermore, combined with the comprehensive attention evaluation index A r and teaching quality feedback data Conduct a comprehensive assessment;

[0244] The edge server adopts a multivariate weighted decision model, comprehensively considering A r , Based on three indicators—Cq, teaching content adjustment strategy—the following formula is used in this embodiment:

[0245]

[0246] Among them, S qScore the quality and applicability strategy of teaching content. γ1, γ2, and γ3 are the weight coefficients of each index, satisfying γ1 + γ2 + γ3 = 1; Is the standardized value of the teaching content quality index;

[0247] Set S q The value range is the interval (0, 1), and different strategy classification criteria are set:

[0248]

[0249] In this embodiment, it is set that the value ranges of h1 and h2 are both within the range (0.2, 0.85);

[0250] When A r When the importance level is high, in this embodiment, γ1 = 0.4, γ2 = 0.3, γ3 = 0.3;

[0251] When When the importance level is high, in this embodiment, γ1 = 0.25, γ2 = 0.5, γ3 = 0.25;

[0252] When When the importance level is high, in this embodiment, γ1 = 0.2, γ2 = 0.35, γ3 = 0.45;

[0253] It should be noted that the above settings of γ1, γ2, and γ3 are initial settings, and are specifically adjusted by the expert group using the fuzzy analytic hierarchy process according to the application situation;

[0254] When 0 < S q <h1, execute the first-level teaching content adjustment strategy; A in this interval r or or In the low interval, the first-level teaching content adjustment strategy includes increasing interaction links, reducing content difficulty, and adding visual auxiliary materials;

[0255] When h1 ≤ S q <h2, execute the second-level teaching content adjustment strategy; A in this interval r or or In the middle interval, the second-level teaching content adjustment strategy includes adjusting the teaching rhythm, increasing case analysis, and introducing more practical applications;

[0256] When h2 ≤ S q <1, execute the teaching content maintenance strategy. A in this interval r or or In the high-level range, the three-level teaching content adjustment strategy includes recording the teaching content as high-quality content, continuously monitoring its effectiveness, maintaining the current teaching strategy, and considering its reuse in similar teaching situations.

[0257] Among them, the adjustment strategy for Level 1 teaching content is more complex than that for Level 2 teaching content.

[0258] The strategy is applied as follows: the edge server sends the established teaching content quality and applicability strategy to the teaching management system or the teacher's end to guide the specific adjustment and optimization of the teaching content.

[0259] After the teaching content is adjusted, the system will continue to monitor A. r , Together with the Cq index, new teaching quality feedback data is generated, which then enters the next round of verification and optimization model generation to achieve closed-loop feedback optimization.

[0260] The weighted regression model links the attention evaluation index with teaching quality feedback data through a linear relationship, enabling the system to rationally adjust the preset range of the stable gaze area using mathematical methods, thus ensuring the applicability and effectiveness of the teaching content.

[0261] The radius *d* of the stable visual field on the teaching screen directly affects the students' visual focus range, and thus their attention distribution. Guided by mathematical models, optimizing the radius *d* ensures that the teaching content is displayed within the students' core visual field, thereby improving the concentration of visual attention.

[0262] Example 2:

[0263] Please provide parameters Figure 2 A privacy-preserving computation-based educational data processing verification system, the system being used to execute the privacy-preserving computation-based educational data processing method, specifically including:

[0264] Region segmentation module: used to collect video data of the target classroom and divide the area where the students are located in the video data into regions using a preset segmentation algorithm to generate a corresponding target student region sequence set;

[0265] Judgment and extraction module: It is used to determine whether the pupil center position of each student in each target student area is within the preset range based on the preset stable gaze area of ​​the teaching screen, so as to extract the gaze fixation time data of each student in the stable gaze area.

[0266] The data on the duration of fixed gaze includes the coefficient of gaze deviation and the total duration of fixed gaze.

[0267] Deviation Zone Division Module: This module analyzes the fixed gaze duration data for each target student area, and obtains the distribution results of gaze deviation frequency and gaze deviation degree.

[0268] The results of the line-of-sight deviation distribution are used to divide the target student region sequence set into multiple deviation region subsets with different degrees of deviation;

[0269] Index construction module: used to receive the gaze deviation frequency and total gaze fixed duration of each deviation region subset for comprehensive analysis in order to construct a comprehensive attention evaluation index for each deviation region subset;

[0270] Strategy Validation Module: This module combines teaching quality feedback data from each deviation region subset with the comprehensive attention evaluation index for comprehensive analysis, generating a validation and optimization model to optimize the preset range of the stable viewing area on the teaching screen and formulate strategies to validate the quality and applicability of related teaching content.

[0271] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, Min-Max Normalization and Z-Score standardization.

[0272] The technical solution of this invention, or the part 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 a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of this invention.

[0273] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0274] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0275] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for processing educational data based on privacy-preserving computation, characterized in that, The specific steps include: Step S1: Collect video data of the target classroom, and divide the area where the students are located in the video data into regions using a preset segmentation algorithm to generate a corresponding target student region sequence set; The edge server divides the target classroom seating distribution map into independent parts: a left region set, a middle region set, and a right region set. Step S2: Determine whether the pupil center position of each student in each target student area is within the preset range based on the preset stable gaze area of ​​the teaching screen, and then extract the gaze fixation time data of each student in the stable gaze area. The data on the duration of fixed gaze includes the coefficient of gaze deviation and the total duration of fixed gaze. Step S3: Analyze the data on the duration of fixed gaze in each target student area to obtain the distribution results of gaze deviation frequency and gaze deviation degree; The results of the line-of-sight deviation distribution are used to divide the target student region sequence set into multiple deviation region subsets with different degrees of deviation; Step S4: Receive the gaze deviation frequency and total gaze duration of each deviation region subset and perform comprehensive analysis to construct a comprehensive attention evaluation index for each deviation region subset; The formula for calculating the Comprehensive Attention Evaluation Index is defined as follows: ； in, It is the comprehensive attention evaluation index of the region set r. and Let be the weight coefficient, and satisfy... ; and The values ​​all fall within the interval (0,1); is the exponential decay coefficient; e is the base of the natural logarithm; ln represents the natural logarithm function; It is the frequency of gaze deviation in the target student area i1; It is the total fixed duration of the student's gaze in the target student area i1; , and These represent the regions after deletion operations on the left, middle, and right regions, respectively; and , and These represent the total number of target student regions in the left, middle, and right region sets after the deletion operation, respectively. It is the average frequency of line-of-sight deviations from the region set r; It is the average total fixed duration of sight lines in the region set r; It is the total number of target student regions in the region set r; Indicates the index of the target student region; Define the comprehensive attention evaluation index for the set of regions r. The range of its value is within the interval (0,1); when The closer the value is to 0, the higher the average frequency of students' gaze deviation and / or the shorter the average total gaze fixation time within region set r; indicating that students' attention is more scattered within region set r. when The closer the value is to 1, the lower the average frequency of students' gaze deviation and / or the longer the average total duration of gaze fixation within region set r; indicating that students within region set r have a higher level of concentration. Step S5: Combine the teaching quality feedback data of each deviation region subset with the comprehensive attention evaluation index for comprehensive analysis, generate a validation and optimization model, and use it to optimize the preset range of the stable gaze area in the teaching screen, and formulate strategies to validate the quality and applicability of the relevant teaching content.

2. The method for processing educational data based on privacy computing according to claim 1, characterized in that: Based on a pre-defined seating distribution map, each seat is uniquely identified by a digital identifier according to the actual location of the students in the target classroom, and the physical location of the seat is mapped to the target student area in the video data through the digital identifier. The target student region sequence set is denoted as {1,2,…,i,…,n}, where i represents the index label of the target student region and n is the total number of target student regions; Video data within the target student area is transmitted to an edge server via a real-time monitoring system. The video data is then encrypted using an encryption module on the edge server and transmitted to a Trusted Execution Environment (TEE) via a secure data channel for data privacy processing and analysis.

3. The method for processing educational data based on privacy computing according to claim 2, characterized in that: The pupil center position was detected using an embedded eye-tracking algorithm to detect the center position of the student's face and pupil in each target student area; The embedded eye-tracking algorithm includes a face detection module and a pupil detection algorithm; Within the Trusted Execution Environment (TEE), the face detection module is invoked to locate the face in the target student region i, and the pupil detection algorithm is used to determine the center position of the pupil in the target student region i. The center position of the pupil is represented by the pixel values ​​on the teaching screen; The center area of ​​the teaching screen is pre-defined as the "stable gaze area" of the students corresponding to the target student area; The preset range of the stable line of sight area is defined as a circular area with a radius of d centered on the center of the teaching screen; In the continuous frame sequence of video data corresponding to the target student region i, the pupil center position of each frame is detected in real time, and it is determined whether the pupil center position falls within the stable line of sight area. If the center of the pupil is within the stable line of sight area, then the line of sight is considered fixed; If the center of the pupil is not located within the stable visual field, then the visual field is considered to be deviated. And analyze the results of the degree of visual deviation in target student area i; Set a preset gaze fixation time window for the students corresponding to the target student area i. Within the preset gaze fixation time window, record the total number of gaze fixations for the target student area i as J. When the center of the pupil is outside the stable visual area, a deviation event is recorded, and the total number of deviation events is counted as N1, where N1≤J; The result of the degree of visual deviation in the target student area i is represented as a visual deviation degree coefficient. ; Fixed duration of all sight sights obtained within the preset fixed sight sight time window The results are aggregated to calculate the total fixed viewing time of students corresponding to target student region i. ; The total viewing time of the target student area i is fixed. and the coefficient of visual deviation This is represented by data on the duration of fixed gaze.

4. The method for processing educational data based on privacy computing according to claim 3, characterized in that: The target student regions within the off-region subset are set to exhibit a clustered pattern. The formula for calculating the frequency of gaze deviation is defined as follows: ; in, It is the frequency of gaze deviation of the student corresponding to the target student area i within a preset fixed gaze time window; It represents the total number of deviation events from the target student region i; It is the total number of times the target student's gaze is fixed in area i; Based on the target student region sequence set {1,2,…,i,…,n}, analyze the visual deviation coefficient of the target student region within each region set. Value distribution; Calculate the line-of-sight deviation coefficients corresponding to the left, middle, and right region sets. The statistical characteristics include the mean and standard deviation. Apply preset clustering rules to filter and delete target student regions that do not conform to the clustering rules; the preset clustering rules are as follows: Set a first preset threshold and a second preset threshold for the mean and standard deviation in the statistical characteristics, respectively; If the mean of each region set is less than the corresponding first preset threshold and the standard deviation is lower than the corresponding second preset threshold, then the region set is considered to have clustering. If the visual deviation coefficient of the target student area i is... If the value is higher than the preset aggregation threshold, the target student region i is deemed not to meet the aggregation rules, and the target student region i that does not meet the aggregation rules is deleted from the corresponding region set.

5. The educational data processing method based on privacy computing according to claim 4, characterized in that: Collect the regions after the deletion operation , , Update the database on the edge server; calculate , and The degree of visual deviation corresponding to the target student area The mean of each is denoted as . , and ; based on , and The range of values ​​is determined by setting different deviation classification standards to divide the left region set, the middle region set, and the right region set into low deviation level, medium deviation level, and high deviation level. The set of regions that meet the low deviation level is denoted as the low deviation region subset, the set of regions that meet the medium deviation level is denoted as the medium deviation region subset, and the set of regions that meet the high deviation level is denoted as the high deviation region subset.

6. The method for processing educational data based on privacy computing according to claim 5, characterized in that: Collect teaching quality feedback data related to each subset of deviation regions. Teaching quality feedback data It consists of evaluation indicators based on students' understanding of the teaching content, their participation in teaching activities, and the effectiveness of teaching. The edge server uses a weighted linear regression model to process the standardized teaching quality feedback data. As the dependent variable, the comprehensive attention evaluation index The following validation and optimization model is established using the independent variable: ； in, This is the intercept term of the model; and The regression coefficient represents The degree of impact on teaching quality; For the error term of the region set r; Select from region set r The set of regions with the smallest values ​​is selected as the optimization target within a preset range. The specific optimization logic is as follows: Based on the regression coefficients in the validation and optimization model and The radius d is dynamically adjusted as follows; ； in, The radius of the optimized line-of-sight stable region; For adjustment coefficients; To verify the teaching quality feedback data values ​​predicted by the optimized model.

7. The method for processing educational data based on privacy computing according to claim 6, characterized in that: Based on the output of the validation and optimization model, select from the region set r The strategy for validating the quality and applicability of relevant teaching content is based on the set of regions with the smallest numerical values. The specific logic includes: The teaching content quality index Cq is obtained by weighting the difficulty, interactivity, and information content. The standardized value of the teaching content quality index Cq is denoted as... ; A multivariate weighted decision model is adopted to comprehensively consider... , and Three indicators to formulate strategies for adjusting teaching content; ； in, Scoring of teaching content quality and applicability strategies , , The weighting coefficients for each indicator satisfy the following conditions: ; set up The value range is the interval (0,1), and different strategy classification criteria are set: ； when At that time, implement the first-level teaching content adjustment strategy; when At that time, implement the strategy of adjusting the teaching content at the secondary level; when At that time, implement strategies to maintain the teaching content.

8. A privacy-preserving computation-based educational data processing and verification system, characterized in that: The system is used to execute the privacy-based computation-based educational data processing method according to any one of claims 1-7, specifically including: Region segmentation module: used to collect video data of the target classroom and divide the area where the students are located in the video data into regions using a preset segmentation algorithm to generate a corresponding target student region sequence set; Judgment and extraction module: It is used to determine whether the pupil center position of each student in each target student area is within the preset range based on the preset stable gaze area of ​​the teaching screen, so as to extract the gaze fixation time data of each student in the stable gaze area. The data on the duration of fixed gaze includes the coefficient of gaze deviation and the total duration of fixed gaze. Deviation Zone Division Module: This module analyzes the fixed gaze duration data for each target student area, and obtains the distribution results of gaze deviation frequency and gaze deviation degree. The results of the line-of-sight deviation distribution are used to divide the target student region sequence set into multiple deviation region subsets with different degrees of deviation; Index construction module: used to receive the gaze deviation frequency and total gaze fixed duration of each deviation region subset for comprehensive analysis in order to construct a comprehensive attention evaluation index for each deviation region subset; Strategy Validation Module: This module combines teaching quality feedback data from each deviation region subset with the comprehensive attention evaluation index for comprehensive analysis, generating a validation and optimization model to optimize the preset range of the stable viewing area on the teaching screen and formulate strategies to validate the quality and applicability of related teaching content.

Images