Intelligent class patrol method, system and device and medium

Abstract

The embodiment of the invention provides an intelligent class patrol method, system and device and a medium. The method comprises the following steps: acquiring monitoring data through sensing equipment in a classroom; determining teacher and student behavior data and teacher and student text data based on the monitoring data; based on the teacher and student behavior data and the teacher and student text data, generating teaching quality data of each teaching class; based on the teaching quality data and the historical quality data of the associated teaching class of each teaching class, determining the misjudgment confidence of each teaching class, and generating a sampling parameter of each teaching class; updating the teaching quality data based on the sampling parameters to obtain updated teaching quality data; and based on the updated teaching quality data, course patrol parameters are determined.

Description

Technical Field

[0001] This specification relates to the field of smart classroom management technology, and in particular to a smart class inspection method, system, device and medium. Background Art

[0002] Amidst the wave of educational transformation, smart classrooms, a key practice in educational informatization, are rapidly being implemented in schools at all levels to revolutionize traditional education models and improve teaching effectiveness. Currently, most smart classrooms rely on hardware such as surveillance cameras and sensors to build data collection systems, capturing real-time classroom behavior, operation, and environmental data. This replaces manual inspections for classroom supervision and behavior recognition, providing a basis for teaching management. However, over-reliance on hardware data collection in complex environments is susceptible to interference from lighting and occlusion, leading to data distortion, high misjudgment rates, and disruptive teaching. Furthermore, data analysis algorithms driven by pre-set rules struggle to adapt to complex scenarios such as personalized teaching behaviors and unexpected teaching events.

[0003] Therefore, it is hoped to propose a smart class inspection method, system, device and medium to reduce the misjudgment rate and generate reasonable class inspection parameters. Summary of the Invention

[0004] One or more embodiments of this specification provide a smart class inspection method. The smart class inspection method includes: obtaining monitoring data through sensor equipment in the classroom; determining teacher-student behavior data and teacher-student text data based on the monitoring data; generating teaching quality data for each teaching class based on the teacher-student behavior data and the teacher-student text data; the teaching quality data includes an estimated quality score and potential teaching anomalies; based on the teaching quality data and the historical quality data of the associated teaching classes of each teaching class, determining the misjudgment confidence of each teaching class and generating sampling parameters for each teaching class; the sampling parameters include a rejection range; based on the sampling parameters, updating the teaching quality data to obtain updated teaching quality data; based on the updated teaching quality data, determining class inspection parameters.

[0005] One or more embodiments of this specification provide a smart class inspection system. The smart class inspection system includes: a data acquisition module, configured to acquire monitoring data through sensor equipment in the classroom; a data determination module, configured to: determine teacher-student behavior data and teacher-student text data based on the monitoring data; generate teaching quality data for each teaching class based on the teacher-student behavior data and the teacher-student text data; the teaching quality data includes an estimated quality score and potential teaching anomalies; and based on the teaching quality data and the historical quality data of the associated teaching classes of each teaching class, determine the misjudgment confidence of each teaching class and generate sampling parameters for each teaching class; the sampling parameters include a rejection range; a data update module, configured to update the teaching quality data based on the sampling parameters to obtain updated teaching quality data; a parameter determination module, configured to determine class inspection parameters based on the updated teaching quality data.

[0006] One or more embodiments of this specification provide a smart class inspection device, including a processor, wherein the processor is used to execute a smart class inspection method.

[0007] One or more embodiments of this specification provide a computer-readable storage medium, which stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the smart class inspection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein: Figure 1 is a schematic diagram of system modules of a smart class inspection system according to some embodiments of this specification; Figure 2 is an exemplary flow chart of a smart class inspection method according to some embodiments of this specification; Figure 3 is an exemplary flow chart of a method for updating teaching quality data according to some embodiments of this specification; Figure 4 is an exemplary schematic diagram of a class inspection evaluation model according to some embodiments of this specification; Figure 5 This is an exemplary flowchart of a method for determining an efficient teaching template according to some embodiments of this specification. DETAILED DESCRIPTION

[0009] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0010] It should be understood that the terms "system," "device," "unit," and / or "module" used herein are a method for distinguishing different components, elements, parts, portions, or assemblies at different levels. However, if other terms can achieve the same purpose, the terms may be replaced by other expressions.

[0011] As used in this specification and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not refer to the singular but also include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0012] Flowcharts are used throughout this specification to illustrate the operations performed by systems according to embodiments of this specification. It should be understood that preceding or following operations do not necessarily need to be performed in exact order. Instead, the steps may be processed in reverse order or simultaneously. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0013] Figure 1 This is a schematic diagram of the system modules of the smart class inspection system according to some embodiments of this specification.

[0014] In some embodiments, as Figure 1 As shown, the smart class inspection system (hereinafter referred to as the system) 100 may include a data acquisition module 110 , a data determination module 120 , a data update module 130 , a parameter determination module 140 and a template determination module 150 .

[0015] In some embodiments, the data acquisition module is configured to acquire monitoring data through sensor devices in the classroom. That is, the data acquisition module can be communicatively connected or physically connected to the sensor devices deployed in the teaching classroom and acquire monitoring data through the sensor devices.

[0016] In some embodiments, the data determination module is configured to determine teacher-student behavior data and teacher-student text data based on monitoring data; generate teaching quality data for each teaching class based on the teacher-student behavior data and teacher-student text data; and determine the misjudgment confidence of each teaching class based on the teaching quality data and the historical quality data of the associated teaching classes of each teaching class, and generate sampling parameters for each teaching class.

[0017] In some embodiments, the data update module is configured to update the teaching quality data based on the sampling parameters to obtain updated teaching quality data. In some embodiments, the data update module is further configured to determine stage division data based on historical teaching data; determine potential teaching anomalies based on the stage division data; determine monitoring sampling parameters based on the potential anomaly teaching points; determine updated teacher-student behavior data and teacher-student text data based on the monitoring sampling parameters and the exclusion range; and update the teaching quality data based on the updated teacher-student behavior data and teacher-student text data.

[0018] In some embodiments, the parameter determination module is configured to determine the class inspection parameters based on the updated teaching quality data. In some embodiments, the parameter determination module is further configured to determine the classroom to be inspected based on the updated teaching quality data; and determine the class inspection parameters based on the classroom to be inspected.

[0019] In some embodiments, the template determination module is configured to determine high-quality classroom data based on teaching quality data and false positive rate confidence; determine efficient features based on high-quality classroom data and preset threshold data; and determine an efficient teaching template based on the efficient features.

[0020] For more information about each module, see Figure 2-Figure 5 and its related descriptions.

[0021] Figure 2 This is an exemplary flow chart of the smart class inspection method according to some embodiments of this specification. Figure 2 As shown, the process 200 includes the following steps. In some embodiments, the process 200 can be executed by a smart class inspection system.

[0022] Step 210: Acquire monitoring data through sensor equipment in the classroom.

[0023] Sensor devices are devices used to collect multi-dimensional data related to classroom teaching. For example, sensor devices may include surveillance cameras, microphones, or microphone arrays deployed in classrooms.

[0024] Monitoring data refers to multi-dimensional data related to the teaching classroom collected by the sensor device. For example, the monitoring data may include image data and voice data related to the teaching classroom. In some embodiments, the data acquisition module can acquire monitoring data at multiple time points.

[0025] Step 220: Determine teacher-student behavior data and teacher-student text data based on the monitoring data.

[0026] In some embodiments, the data determination module may obtain teacher-student behavior data and teacher-student text data at multiple time points based on monitoring data at multiple time points.

[0027] Teacher-student behavior data refers to data reflecting the characteristics of teachers' teaching behaviors and students' listening behaviors. For example, this data may include the teacher's gesture amplitude, teacher position, the teaching phase of the class, and student body movements (such as sitting angle and head direction). Teaching phases may include lectures, teacher-student interaction, and student independent practice.

[0028] In some embodiments, the data determination module can generate valid monitoring data based on the monitoring data, and use an image recognition algorithm to identify and detect the body language of teachers and students based on the image data in the valid monitoring data to obtain teacher-student behavior data. In some embodiments, the image recognition algorithm can include an image recognition model such as ResNet-50, EfficientNet, etc.

[0029] Valid monitoring data refers to data sampled from the monitoring data. In some embodiments, the data determination module may extract the monitoring data based on an extraction interval to obtain valid monitoring data. For example, if the monitoring data includes image data collected by a sensor device in a classroom at 5-second intervals (e.g., image data corresponding to 0s, 5s, 10s, etc.), the valid monitoring data may be image data corresponding to 0s, 20s, 40s, etc. The extraction interval may be manually set based on experience.

[0030] Teacher-student text data refers to the collected conversation data or speech content data between teachers and students. For example, teacher-student text data may include text information obtained by recognizing and converting the conversation information or speech content information between teachers and students.

[0031] In some embodiments, the data determination module may use a speech recognition algorithm based on the speech data included in the valid monitoring data to obtain the teacher-student text data. In some embodiments, the speech recognition algorithm may include a speech recognition model, such as OpenAI Whisper, Mini-Omni, etc.

[0032] Step 230: Generate teaching quality data for each teaching class based on teacher-student behavior data and teacher-student text data.

[0033] In some embodiments, the data determination module may generate teaching quality data at multiple time points.

[0034] Teaching quality data refers to data used to evaluate the teaching effectiveness of teaching and whether the teaching status is normal. In some embodiments, the teaching quality data includes estimated quality scores and potential teaching anomalies.

[0035] An estimated quality score is a quantifiable score used to assess teaching effectiveness and quality. For example, an estimated quality score can be calculated by comprehensively evaluating data from multiple dimensions of a classroom.

[0036] Potential teaching anomalies refer to time points during classroom instruction where the system detects possible teaching anomalies (i.e., abnormal time points) and the types of teaching anomalies. For example, potential teaching anomalies may include detected time points where classroom discipline may be lax, teacher and student enthusiasm may be low, or students may exhibit unusual behavior.

[0037] In some embodiments, the data determination module can count the teacher's orderliness keywords in the teacher-student text data and the frequency of changes in the students' sitting angles in the teacher's behavior data to determine the teaching quality data. For example, the data determination module can identify the frequency of occurrence of orderliness keywords such as "be quiet" and "pay attention to the lecture" per minute in the teacher-student text data, and the average frequency of changes in the sitting angles of all students per minute, and query them in the first preset table to obtain the teaching quality data. The first preset table includes multiple groups of orderliness keyword frequencies, student sitting angle change frequencies, and their corresponding teaching quality data. The first preset table is constructed based on historical data.

[0038] In some embodiments, the data determination module can obtain the classroom type of the teaching classroom; and determine the teaching quality data using a preset rule library based on teacher-student behavior data, teacher-student text data, classroom type, and effective monitoring data.

[0039] Classroom type refers to a classification type obtained by classifying the course nature of the teaching class. For example, the class type may include a theoretical class that focuses on knowledge transfer, such as a Chinese class. For example, the class type may also include a laboratory class that focuses on hands-on operations, such as a physics laboratory class. For another example, the class type may also include practical classes such as physical education and programming classes. In some embodiments, the data determination module may obtain the class type based on the teacher's reporting information for each teaching class.

[0040] The preset rule library is a set of basic rules for evaluating teaching quality, including evaluation dimensions for different classroom types and their corresponding scoring standards. In some embodiments, the preset rule library will set specific preset scoring rules for evaluation dimensions such as teacher explanation quality, classroom discipline, teacher-student interaction, and student engagement. In some embodiments, the preset rule library can be preset based on experience.

[0041] For example, when the class type is a theoretical class, the preset scoring rules included in the preset rule library may include:

[0042] In some embodiments, the data determination module may process the speech data based on spectral subtraction to obtain noise data, and then determine the average noise decibel based on the decibel level of the noise data. For example, the data determination module may obtain speech data collected by microphones installed at different locations in a classroom, process the speech data using spectral subtraction to obtain noise data, and determine the average noise decibel based on the decibel level of the noise data.

[0043] In some embodiments, the data determination module may count the frequency of teachers' questions (i.e., teacher question frequency) based on the teacher-student text data and teacher-student behavior data. For example, the data determination module may count the frequency of occurrence of keywords such as "Please answer," "Do you have any questions?", and "What does this phenomenon indicate?" in the teacher-student text data to obtain the teacher question frequency.

[0044] In some embodiments, the data determination module may count the proportion of students who raise their hands each time the teacher asks a question based on the teacher-student behavior data, so as to obtain the proportion of students who raise their hands in the class.

[0045] In some embodiments, the data determination module may calculate a first sub-match between the duration of the teaching phase of the current teaching class and the duration of the teaching phase in the teaching template, a second sub-match between the teacher's explanation text data in the teacher-student text data and the explanation text data in the teaching template, and a third sub-match between the teacher's teaching emotion and the teaching emotion in the teaching template to determine the matching degree between the current teaching class and the teaching template. For example, the data determination template may use a weighted average of the first sub-match, the second sub-match, and the third sub-match as the matching degree between the current teaching class and the teaching template.

[0046] In some embodiments, the data determination module determines a first sub-matching degree based on the duration of each teaching phase of the current teaching class and the duration of each teaching phase in the teaching template. For example, if the teaching phase of the teaching class is 8 minutes (when the teaching phase is completed) and the teaching phase in the teaching template is 10 minutes, then the first sub-matching degree is: 1-|8-10| / 10=80% (if the calculated value is negative, the first sub-matching degree is 0%).

[0047] In some embodiments, the data determination module can extract keywords (such as "Newton's First Law") from the teaching template and the current teaching class's explanation text, and calculate the ratio of the number of keywords in the current teaching class's explanation text to the number of keywords in the teaching template's explanation text within the corresponding time period to determine the second sub-matching degree. For example, if the current teaching class completes 0-10 minutes, the number of keywords in the current teaching class's explanation text from 0-10 minutes is 30, while the number of keywords in the teaching template's explanation text from 0-10 minutes is 50, then the second sub-matching degree is: 1-|30-50| / 50=60% (if the calculated value is negative, the second sub-matching degree is 0%).

[0048] In some embodiments, the data determination module may use a sentiment classification model to process the teacher-student text data, obtain the teacher's teaching sentiment (current sentiment) in the current classroom, and compare the current sentiment with the teacher's teaching sentiment in the teaching template (template sentiment) to determine whether the third sub-matching degree is consistent. For example, if the current sentiment is consistent with the template sentiment, the matching degree is 1; otherwise, it is 0. In some embodiments, the sentiment classification model may include a Transformer model, etc.

[0049] In some embodiments, the data determination module can determine the value of the data indicator based on the teacher-student behavior data, teacher-student text data, and effective monitoring data, and search in the preset rule library based on the value of the data indicator and the class type, and determine the score of each evaluation dimension of the current teaching class according to the corresponding evaluation dimension and preset scoring rules in the preset rule library, and then calculate the weighted sum of the score of each evaluation dimension of the current teaching class, and use it as the estimated quality score of the current teaching class. Among them, the weight of the score of each evaluation dimension is related to the class type. For example, when the class type is a theoretical class, the weight of the score corresponding to the two evaluation dimensions of teacher explanation quality and classroom discipline can be set higher.

[0050] In some embodiments, when the calculated score of an evaluation dimension at a certain time point is lower than a preset threshold, it is determined that there is an anomaly in the evaluation dimension at that time point, and the time point is determined to be an abnormal time point, and the type of teaching anomaly is determined based on the abnormal evaluation dimension. For example, if the valid monitoring data is the data of the 5th minute of the current teaching class, and the average decibel of noise determined based on the valid monitoring data is 80 decibels, then based on the preset rule base and the "average decibel of noise is 80 decibels", the score of the dimension of classroom discipline is determined to be 40 points < the preset threshold (such as 60 points). Then, at the 5th minute of the teaching class, there is a potential teaching anomaly point of "lax classroom discipline".

[0051] For more information on how to identify potential teaching anomalies, see Figure 3 3. The relevant description of step 320 in FIG.

[0052] Step 240 : determining the misjudgment confidence of each teaching class based on the teaching quality data and the historical quality data of the teaching classes associated with each teaching class, and generating sampling parameters for each teaching class.

[0053] Related courses refer to courses with a similar teaching model to the current course. For example, related courses may include courses with the same content taught by the current course's teacher or other teachers. In some embodiments, the data determination module may extract historical courses with the same content from the historical data based on the content of the current course (e.g., the teaching topic and the corresponding textbook content) as related courses.

[0054] Historical quality data refers to teaching quality data corresponding to associated teaching courses in the historical data. For example, it may be teaching quality data for associated teaching courses within the past month. In some embodiments, the data determination module may obtain the historical quality data by reading teaching quality data within a preset time period in the historical data. The preset time period may be manually set based on experience.

[0055] The false positive confidence level is an estimate of the probability that a potential teaching anomaly detected by the system is actually an anomaly. In some embodiments, the false positive confidence level can include the probability of false positives for teaching anomalies such as poor classroom discipline, low teacher and student motivation, and unusual student behavior. A higher false positive confidence level indicates a higher likelihood that the potential teaching anomaly is a false positive.

[0056] In some embodiments, the data determination module may calculate the probability of each type of teaching anomaly occurring in all associated teaching classes of the teaching class, and determine the misjudgment confidence level for the current teaching class based on the probability of the type of teaching anomaly of the potential teaching anomaly point existing in the current teaching class occurring in associated teaching classes. For example, the lower the probability of the type of teaching anomaly of the potential teaching anomaly point existing in the current teaching class occurring in associated teaching classes, the higher the misjudgment probability of the potential teaching anomaly point, i.e., the greater the misjudgment confidence level determined.

[0057] In some embodiments, the data determination module may further determine the misjudgment confidence level of the potential teaching anomaly based on the ratio of the duration of the potential teaching anomaly to the total duration. For example, for persistent anomalies (such as teachers deviating from the syllabus and low student and teacher motivation), the greater the ratio of the duration of the potential teaching anomaly to the total duration, the lower the misjudgment confidence level; for sudden anomalies (such as abnormal student behavior and poor classroom discipline), the greater the ratio of the duration of the potential teaching anomaly to the total duration, the greater the misjudgment confidence level.

[0058] For more information on how to determine the confidence level of false positives, see Figure 3 and its related descriptions.

[0059] Sampling parameters refer to relevant parameters used when generating effective monitoring data. In some embodiments, the sampling parameters include a rejection range.

[0060] The exclusion range refers to the interval used to remove abnormal data from valid monitoring data and teaching quality data. For example, if the exclusion range corresponding to a potential teaching anomaly is [1, -1], then the time point corresponding to the potential teaching anomaly, as well as the valid monitoring data and teaching quality data for the previous and next time points, will be excluded.

[0061] In some embodiments, the data determination module can determine whether there is a rejection range and the size of the rejection range based on the numerical value of the misjudgment confidence level corresponding to the potential teaching anomaly points in each teaching class. For example, if the misjudgment confidence level corresponding to the potential teaching anomaly points is greater than a confidence threshold, it is considered that a rejection range exists, and the greater the misjudgment confidence level, the larger the rejection range. The confidence threshold can be set manually based on experience.

[0062] Step 250: Update the teaching quality data based on the sampling parameters to obtain updated teaching quality data.

[0063] The updated teaching quality data refers to data obtained after the teaching quality data is updated. In some embodiments, the updated teaching quality data may include an updated estimated quality score and an updated potential teaching anomaly point.

[0064] In some embodiments, the data update module can eliminate part of the teaching quality data at multiple time points before the update based on the elimination range in the potential teaching anomalies and their corresponding sampling parameters, and use the teaching quality data that has not been eliminated as the updated teaching quality data.

[0065] For more information on how to update teaching quality data, see Figure 3 and its related descriptions.

[0066] Step 260: Determine the class inspection parameters based on the updated teaching quality data.

[0067] The course inspection parameters refer to the guidance parameters for assigning course inspection personnel to conduct course inspection. In some embodiments, the course inspection parameters may include course inspection personnel, the course inspection route corresponding to each course inspection personnel, the stop points and the length of stay.

[0068] The patrol personnel refer to the supervisors who are responsible for patrolling the teaching classroom according to the patrol parameters. In some embodiments, the patrol personnel may include teachers who are currently not teaching, members of the supervision team, etc.

[0069] The course inspection route refers to the route that the course inspector should take during the course inspection. In some embodiments, the course inspection route may include the teaching classrooms that need to be inspected and their inspection order and inspection path instructions.

[0070] A stop point refers to a location where an inspector is required to stop and observe the classroom and record feedback on the teaching situation. In some embodiments, a stop point may include the location of the classroom where the inspector is required to stop and observe. The length of stay refers to the length of time the inspector is required to stay at the stop point.

[0071] In some embodiments, the parameter determination module may determine as a stay point the classroom location of a teaching classroom whose updated estimated quality scores among all teaching classrooms, or the average of the updated estimated quality scores at multiple time points is lower than a quality threshold.

[0072] In some embodiments, the length of stay corresponding to a stay point is negatively correlated with the updated estimated quality score of the teaching class corresponding to the stay point, or the average of the updated estimated quality scores of multiple time points.

[0073] In some embodiments, the parameter determination module may obtain an initial dwell time and, based on the updated estimated quality score or the average of the updated estimated quality scores at multiple time points, increase the initial dwell time by a predetermined amount to obtain the dwell time at the dwell point. In some embodiments, the predetermined amount may be negatively correlated with the updated estimated quality score or the average of the updated estimated quality scores at multiple time points. In some embodiments, the initial dwell time may be preset based on experience.

[0074] In some embodiments, the parameter determination module can sort the stay points from largest to smallest according to the length of stay, and connect these stay points in series to determine the course inspection route. In some embodiments, the parameter determination module can determine the currently idle teacher as the course inspection personnel. In some embodiments, the parameter determination module can also determine the currently idle person closest to the stay point with the longest stay time as the course inspection personnel.

[0075] For more information on how to determine the parameters of the class tour, see Figure 5 and its related descriptions.

[0076] Some embodiments of this specification generate multi-dimensional teaching quality assessment indicators by fusing multi-source sensor data and dynamically analyzing misjudgment confidence levels based on historically correlated course data, effectively reducing the AI's misjudgment rate for complex teaching scenarios and ensuring the scientificity and objectivity of the assessment results. Dynamically eliminating outliers and their adjacent time period data based on misjudgment confidence levels can effectively prevent local misjudgments from affecting the overall assessment. Furthermore, based on updated teaching quality data, optimal class inspection parameters are dynamically generated, prioritizing coverage of low-scoring classrooms. This significantly improves the pertinence and efficiency of manual class inspections and avoids the resource waste of traditional fixed-route inspections.

[0077] Figure 3 FIG. 1 is an exemplary flow chart of a method for updating teaching quality data according to some embodiments of this specification. Figure 3 As shown, the process 300 includes the following steps: In some embodiments, the process 300 may be executed by a data determination module.

[0078] Step 310: Determine stage division data based on historical teaching data.

[0079] Historical teaching data refers to historical record data related to the associated teaching class. In some embodiments, the historical teaching data may include misjudgment confidence corresponding to potential teaching anomalies in the history of the associated teaching class, teacher-student behavior data, teacher-student text data, and historical quality data. In some embodiments, the historical teaching data may include historical quality data and historical misjudgment confidence within the past month. For more information about historical quality data, see Figure 2 and its related descriptions.

[0080] Phase division data refers to multiple consecutive phases obtained by dynamically dividing a teaching class. For example, the phase division data may include dividing a teaching class into three phases (e.g., early stage (e.g., 0-15 minutes), middle stage (e.g., 16-30 minutes), and late stage (e.g., 31-45 minutes) according to a preset time period (e.g., 15 minutes). Another example may be divided into an explanation phase, an interactive discussion phase, and a practice phase. In some embodiments, the preset time period may be based on experience.

[0081] Step 320: Determine potential teaching anomalies based on the stage division data.

[0082] In some embodiments, the evaluation reference range for teaching behavior at each stage in the stage-division data may be defined as historical baseline data. In some embodiments, the historical baseline data may include baseline thresholds for teaching characteristics used to evaluate teaching quality at each stage. Teaching characteristics may include student posture, vocal activity, and the like.

[0083] Student posture refers to the student's body posture data. For example, the student posture may include the average sitting angle, posture change frequency, head orientation, etc. In some embodiments, the data update module can extract the student posture from the teacher-student behavior data. Exemplarily, the data update module can calculate the average sitting angle of all students in the teacher-student behavior data to obtain the average sitting angle, and calculate the number of posture changes of students in the teacher-student behavior data per unit time (such as 1 minute) to obtain the posture change frequency. For more information about teacher-student behavior data, see Figure 2 and its related descriptions.

[0084] Sound activity refers to the intensity of sound in a classroom. In some embodiments, the sound activity may include the decibel level of the voice data. In some embodiments, the data update module may obtain the sound activity based on the decibel level of the voice data recording. For more information about voice data, see Figure 2 and its related descriptions.

[0085] In some embodiments, the data update module can obtain the teacher-student text data and teacher-student behavior data of the associated teaching class according to a preset time window (such as every 5 minutes), generate classroom data subsets of multiple time windows, extract teaching features from the classroom data subset of each time window, and standardize the teaching features of different dimensions, convert all teaching features into a unified range (for example, the interval of 0-1), and based on the standardized teaching features, generate baseline thresholds for each division stage according to the division stages of the teaching classroom.

[0086] In some embodiments, the data update module can determine the baseline threshold based on the mean and standard deviation of the standardized teaching features. For example, if the mean of the historical sound activity corresponding to a division stage is 55 decibels and the standard deviation is 5 decibels, then the baseline threshold is 55+2×5=65 decibels.

[0087] In some embodiments, the data update module can also obtain phase-by-phase statistical data for multiple related teaching courses and, based on the phase-by-phase statistical data, determine the frequency of occurrence of teaching anomaly types at each phase of the multiple related teaching courses, thereby dynamically adjusting the historical baseline data. For example, the data update module can adjust historical baseline data containing a specific teaching anomaly type based on preset adjustment data. The preset adjustment data includes the adjustment direction (including downward adjustment, upward adjustment, or narrowing of the range) and the adjustment amplitude.

[0088] In some embodiments, the adjustment direction for historical baseline data with a certain teaching anomaly type during a specific period can be preset, and the magnitude of the adjustment can be positively correlated with the frequency of occurrence of the teaching anomaly type during that period. For example, if the frequency of the teaching anomaly type "lack of classroom discipline" during the early stage is 50%, the preset adjustment data can be determined to reduce the baseline threshold corresponding to voice activity by 10%.

[0089] Stage statistical data refers to statistical data of the same divided stage in the associated teaching class obtained for each divided stage of the current teaching class. In some embodiments, the stage statistical data may include the number of occurrences of the teaching anomaly type in each divided stage in the associated teaching class. In some embodiments, the data update module may use the same division method as the current teaching class for the associated teaching class, obtain the stage division data of each associated teaching class, and count the number of occurrences of the teaching anomaly type in each divided stage to obtain stage statistical data. In some embodiments, the number of occurrences of a certain teaching anomaly type in a certain divided stage / the number of such divided stages in the total associated classes can be used to determine the frequency of occurrence of a specific teaching anomaly type in a divided stage.

[0090] In some embodiments, the data update module can determine potential teaching anomaly points based on historical baseline data and teaching anomaly types related to teaching characteristics. In some embodiments, the teaching anomaly type related to each teaching characteristic can be manually preset. For example, the teaching anomaly type corresponding to sound activity is "lax classroom discipline". If the current historical baseline data includes a baseline threshold of 60 decibels for sound activity in the early stage, then when the sound activity in the early stage of the current teaching class exceeds 60 decibels, it is considered that there is a teaching anomaly type of lax classroom discipline, and the corresponding time point is the abnormal time point.

[0091] Step 330: Determine monitoring sampling parameters based on potential teaching abnormal points.

[0092] In some embodiments, the sampling parameters also include monitoring sampling parameters. Monitoring sampling parameters refer to relevant parameters used to select samples for monitoring data. For example, the monitoring sampling parameters may include a sampling interval.

[0093] In some embodiments, the data update module can sample and select the monitoring data based on the monitoring sampling parameters to obtain sampled monitoring data. Exemplarily, when the current sampling interval is 1, one monitoring data is obtained as sampled monitoring data every time a monitoring data is obtained. For example, the monitoring data includes image data collected by the sensing device in a teaching classroom at 0s, 5s, 10s, 20s, and 25s. When the sampling interval is 1, the sampled monitoring data includes image data corresponding to 0s, 10s, and 25s.

[0094] In some embodiments, the monitoring sampling parameters can be manually preset based on experience. In some embodiments, when a potential teaching anomaly occurs, the data update module can also adjust the extraction interval based on the misjudgment confidence level corresponding to the potential teaching anomaly to obtain a sampling interval, thereby obtaining the monitoring sampling parameters. In some embodiments, the data update module can shorten the extraction interval by a preset adjustment range based on the misjudgment confidence level to obtain the sampling interval. The preset adjustment range is positively correlated with the misjudgment confidence level. Step 340: Determine the updated teacher-student behavior data and teacher-student text data based on the monitoring sampling parameters and the elimination range.

[0095] In some embodiments, the data update module can remove the data at the time point corresponding to the removal range from the current valid monitoring data, and use the monitoring sampling parameters to sample the monitoring data corresponding to the removed valid monitoring data to obtain sampled monitoring data, and determine the updated teacher-student behavior data and teacher-student text data based on the sampled monitoring data. For more information about the removal range, see Figure 2 The relevant description of step 250 in FIG.

[0096] For example, when the monitoring data includes image data collected by the sensor device every 5 seconds in a teaching class (such as image data corresponding to 0s, 5s, 10s, ..., 100s in the class), if the extraction interval is 3 (i.e., sampling every 20s), the extracted valid monitoring data are image data corresponding to 0s, 20s, 40s, 60s, 80s, and 100s. The current data update module determines that a potential teaching anomaly occurs at the 60th second and its corresponding misjudgment confidence is greater than the confidence threshold, and calculates the elimination range at this time as [-1, 1]. ; Then, after the data update module eliminates the valid monitoring data based on the elimination range, the remaining valid monitoring data that are not eliminated are the image data corresponding to 0s, 20s and 100s; if at this time, the sampling interval determined by the data update module is 2 (that is, sampling every 15s), then after sampling the monitoring data corresponding to the eliminated valid monitoring data (that is, the image data within 20s~100s (excluding 20s and 100s)), the sampled monitoring data obtained is the image data corresponding to 35s, 50s, 65s, 80s, and 95s.

[0097] The data update module determines the updated teacher-student behavior data and teacher-student text data based on the sampled monitoring data, and Figure 2 The method of determining the teacher-student behavior data and teacher-student text data based on the effective monitoring data in step 220 is similar and will not be repeated here.

[0098] Step 350: Update the teaching quality data based on the updated teacher-student behavior data and teacher-student text data.

[0099] In some embodiments, the data update module can obtain the revised quality data based on the quality determination model; and combine the revised quality data with the teaching quality data that has not been eliminated to update the teaching quality data and generate updated teaching quality data. For more information about the teaching quality data that has not been eliminated, please refer to Figure 2 The relevant description of step 250 in FIG.

[0100] Correcting quality data refers to re-updating the calculated data for the eliminated teaching quality data.

[0101] A quality determination model refers to a model used to determine and correct quality data. In some embodiments, the quality determination model can be a machine learning model, such as a convolutional neural network (CNN) model or a combination of one or more other custom models.

[0102] In some embodiments, the input of the quality determination model includes updated teacher-student behavior data, updated teacher-student text data, sample monitoring data, extracurricular performance data, and teacher teaching characteristics, and the output of the quality determination model includes revised quality data.

[0103] Extracurricular performance data refers to quantified data on students' learning behaviors and achievements outside of the classroom. For example, extracurricular performance data may include student homework completion status, student test data, etc. In some embodiments, the data update module may obtain extracurricular performance data based on student data uploaded by teachers.

[0104] Teacher teaching characteristics refer to data characteristics related to the instructor. For example, teacher teaching characteristics may include age, teaching experience, academic qualifications (including major), technical tool operation ability (which can be obtained through teacher testing), teaching evaluation and satisfaction (obtained from student teaching evaluation systems, peer review data, and supervisory evaluation reports), etc.

[0105] In some embodiments, the data update module may use a first training sample set to train a quality determination model. In some embodiments, the first training sample set includes multiple first training samples with first training labels. For example, the data update module may input multiple first training samples with first training labels into the initial quality determination model, construct a loss function based on the output of the initial quality determination model and the first training labels, iteratively update the parameters of the initial quality determination model based on the loss function, and terminate the iteration when an iteration completion condition is met, thereby obtaining a trained quality determination model. Iterative update methods include gradient descent, etc. Iterative completion conditions include, for example, convergence of the loss function or reaching a threshold number of iterations.

[0106] In some embodiments, the first training sample includes sample teacher-student behavior data, sample teacher-student text data, sample sampling monitoring data, sample extracurricular performance data, and sample teacher teaching characteristics. The first training label may include actual teaching quality data corresponding to the first training sample. In some embodiments, the first inspection sample and the first training label may be obtained based on historical data.

[0107] In some embodiments, training of the quality determination model may include an initial training phase and an intensive training phase. The first training sample set in the initial training phase may be obtained based on historical classroom data from other schools (covering different course types and teacher styles), while the first training sample set in the intensive training phase may be obtained based on historical classroom data from the current school.

[0108] Some embodiments of this specification, based on historical teaching data and phased data, identify time periods with high-frequency anomalies and dynamically adjust monitoring sampling parameters to ensure high-precision data collection during critical periods, avoiding the waste of resources caused by high-load sampling throughout the entire period. Furthermore, by combining historical baseline data, misjudgments caused by differences in teacher style or course type are filtered out, significantly reducing the false alarm rate. Furthermore, teaching quality data is recalculated based on updated teacher-student behavior data and teacher-student text data, improving its accuracy.

[0109] In some embodiments, the data update module may further determine the stage division data based on the historical teaching data; and determine the misjudgment confidence based on the stage division data and potential teaching anomalies.

[0110] In some embodiments, the data update module may further determine the stage division data based on the average duration of each division stage in the historical teaching data. For example, if the average duration of the explanation stage in the historical teaching data is 0-16 minutes, the data update module may divide the explanation stage of the teaching class into 0-16 minutes.

[0111] In some embodiments, the data determination module update module may determine the misjudgment confidence level for the current teaching class based on the probability of the teaching anomaly type of the potential teaching anomaly points existing in each division stage of the currently determined teaching class occurring in the same division stage of the associated teaching class. For example, the lower the probability of the teaching anomaly type of the potential teaching anomaly points existing in a certain division stage of the current teaching class occurring in the same division stage of the associated teaching class, the higher the misjudgment probability of the potential teaching anomaly points occurring in that division stage, i.e., the greater the misjudgment confidence level determined.

[0112] In some embodiments of this specification, the stage division is determined based on the average duration of each divided stage in historical teaching data, which can make the stage division more in line with the actual teaching rhythm, and by analyzing abnormal misjudgment cases in the same divided stage in historical data, the misjudgment confidence of the current stage is dynamically corrected, thereby improving the accuracy of the misjudgment confidence.

[0113] In some embodiments, the parameter determination module can determine the classroom to be inspected based on the updated teaching quality data; and determine the inspection parameters based on the classroom to be inspected. For more information about the updated teaching quality data and inspection parameters, please refer to Figure 2 and its related descriptions.

[0114] Classrooms awaiting inspection refer to classrooms requiring manual inspection. For example, classrooms awaiting inspection may include classrooms with poor teaching quality data and high misjudgment confidence. In some embodiments, the parameter determination module may identify classrooms whose updated estimated quality scores, or the average of the updated estimated quality scores at multiple time points, as awaiting inspection.

[0115] In some embodiments, the preset inspection threshold can be determined based on the historical anomaly frequency and class size of the class corresponding to the teaching class. For example, the higher the historical anomaly frequency and the larger the class size, the higher the preset inspection threshold. The historical anomaly frequency refers to the probability of potential teaching anomalies occurring in the historical data of the class.

[0116] The parameter determination module can determine the class inspection parameters based on the classroom to be inspected in a variety of ways. In some embodiments, the parameter determination module can determine multiple candidate class inspection parameters based on the classroom to be inspected; based on the multiple candidate class inspection parameters, determine the class inspection scores corresponding to the multiple candidate class inspection parameters using a class inspection evaluation model; and determine the class inspection parameters based on the class inspection scores.

[0117] Candidate class inspection parameters refer to multiple class inspection parameters to be determined. In some embodiments, the candidate class inspection parameters can be determined in a variety of ways. For example, the parameter determination module can filter out multiple candidate class inspection parameters from historical inspection records based on the classroom to be inspected.

[0118] In some embodiments, the parameter determination module can generate multiple inspection routes based on the classrooms to be inspected using a path planning algorithm. The path planning algorithm can be, for example, a Traveling Salesman Problem (TSP) algorithm. For the same inspection route, the parameter determination module can randomly set a dwell time for each stop, thereby generating multiple candidate inspection parameters.

[0119] In some embodiments, each classroom to be inspected may be inspected once or multiple times during a single inspection, and the upper limit of the inspection number does not exceed a preset value. The preset value may be preset based on prior experience.

[0120] The class inspection score is an indicator used to measure the class inspection effect corresponding to the candidate class inspection parameters. In some embodiments, the higher the class inspection score, the better the class inspection effect of the corresponding candidate class inspection parameter. In some embodiments, each candidate class inspection parameter corresponds to a class inspection score.

[0121] In some embodiments, the parameter determination module may determine the class inspection scores corresponding to the multiple candidate class inspection parameters through a class inspection evaluation model based on the multiple candidate class inspection parameters.

[0122] The class inspection evaluation model is a model used to determine the class inspection score. In some embodiments, the class inspection evaluation model can be a machine learning model. For example, the class inspection evaluation model can be any one or a combination of a neural network (NN) model or other custom model structures.

[0123] Figure 4 It is an exemplary schematic diagram of the class inspection evaluation model shown in some embodiments of this specification.

[0124] In some embodiments, as Figure 4 As shown, the input of the class tour evaluation model 420 includes multiple candidate class tour parameters (such as candidate class tour parameter 1 (410-1), candidate class tour parameter 2 (410-2),..., candidate class tour parameter N (410-N)), and the output of the class tour evaluation model 420 includes the class tour scores corresponding to the multiple candidate class tour parameters (such as class tour score 1 (430-1), class tour score 2 (430-2),..., class tour score N (430-N)).

[0125] In some embodiments, the input of the class inspection evaluation model also includes teaching quality data, misjudgment confidence, class type, class size, and teacher teaching characteristics. For more information about teaching quality data, misjudgment confidence, and class type, please refer to Figure 2 For more information about teacher teaching characteristics, please refer to Figure 3 and its related descriptions.

[0126] Class size refers to the parameters of a class taught by a single teacher during the same teaching period. In some embodiments, class size includes the number of students, the size of the classroom, etc. In some embodiments, class size can be obtained through sensor devices in the classroom. For more information about sensor devices, see Figure 2 and its related descriptions.

[0127] In some embodiments, the parameter determination module can obtain a class inspection evaluation model through training based on multiple second training samples with second training labels. The training process of the class inspection evaluation model is similar to the training process of the teaching quality determination model and will not be repeated here.

[0128] In some embodiments, the second training sample includes sample tour parameters, sample teaching quality data, sample misjudgment confidence, sample misjudgment rate, sample classroom type, sample classroom size, and sample teacher teaching characteristics. The second training label includes the actual tour score corresponding to the tour parameters of the second training sample.

[0129] In some embodiments, the second training sample can be obtained based on historical data. In some embodiments, the parameter determination module can collect historical class inspection records, record the class inspection effect of each inspection, and use the class inspection effect as the second label. The class inspection effect can be represented by the increase in the estimated quality score before and after the inspection. The greater the increase in the estimated quality score, the higher the class inspection score.

[0130] In some embodiments, the parameter determination module may sort the class inspection scores corresponding to multiple candidate class inspection parameters (ie, multiple class inspection scores) from large to small, and select the class inspection score with the highest score as the class inspection parameter.

[0131] In some embodiments of this specification, based on updated teaching quality data (such as the corrected number of outliers and misjudgment confidence), combined with indicators such as misjudgment confidence and classroom type, final classroom inspection parameters can prioritize low-scoring classrooms or areas with high anomaly densities, avoiding redundant inspections of normal classrooms. Furthermore, by generating inspection scores through a classroom inspection evaluation model and constructing a classroom inspection parameter optimization system, the efficiency of classroom inspection management is improved, achieving intelligent classroom inspection management.

[0132] Figure 5 This is an exemplary flow chart of a method for determining an efficient teaching template according to some embodiments of this specification. Figure 5 As shown, the process 500 includes the following steps: In some embodiments, the process 500 may be executed by a template determination module.

[0133] Step 510: Determine high-quality classroom data based on the teaching quality data and the misjudgment confidence level.

[0134] For an explanation of teaching quality data and misjudgment confidence, see Figure 2 and its related descriptions.

[0135] High-quality classroom data refers to classroom data that meets the needs of teaching improvement. Classroom data refers to the collection of data and / or information generated during the inspection process, including teacher-student behavior data and teacher-student text data. For more information about teacher-student behavior data and teacher-student text data, see Figure 2 and its related descriptions.

[0136] In some embodiments, the template determination module can determine high-quality classroom data using various methods. For example, the template determination module may identify classroom data as high-quality classroom data if the estimated quality score in the teaching quality data is greater than a quality score threshold, the number of potential teaching anomalies (e.g., teaching anomaly types) is less than an anomaly threshold, and the misjudgment confidence level is less than a confidence threshold. The quality score threshold, anomaly threshold, and confidence threshold can be preset based on experience.

[0137] Step 520: Determine efficient features based on high-quality classroom data and preset threshold data.

[0138] The preset threshold data is an indicator used to determine the quality of data and / or features. In some embodiments, the preset threshold data can be preset based on prior experience. In some embodiments, the preset threshold data is a minimum support. For more information on minimum support, see below.

[0139] High-efficiency features are classroom characteristics that demonstrate effective teaching performance. Classroom characteristics refer to performance characteristics related to a classroom. In some embodiments, classroom characteristics may include the frequency of teacher-student interaction at each level, the time allocated for each stage (e.g., explanation, interaction, practice, etc.) at each level, the frequency of teacher questions at each level, the classroom decibel level at each level, and the duration of use of teaching tools at each level. Levels can be represented by "low," "medium," "high," and so on. In some embodiments, the template determination module can perform data binning by binning continuous variables. For example, with respect to the frequency of teacher-student interaction, a frequency of less than 3 times / 10 minutes is defined as low interaction; a frequency of (3 to 5 times) / 10 minutes is defined as medium interaction; and a frequency of greater than 5 times / 10 minutes is defined as high interaction.

[0140] The template determination module can determine the high-performance features in a variety of ways. In some embodiments, the template determination module can determine the correlation between teaching classroom features and teaching performance effects based on high-quality classroom data and preset threshold data through an association rule mining algorithm; and determine the high-performance features based on the correlation.

[0141] Teaching performance refers to data related to teaching performance in the classroom. In some embodiments, teaching performance may include student focus at various levels, student engagement at various levels, and test accuracy at various levels. In some embodiments, with respect to test accuracy, the template determination module may define a test accuracy below 60% as low, a test accuracy between 60% and 80% as medium, and a test accuracy greater than 80% as high.

[0142] In some embodiments, a set of association relationships includes one or more teaching classroom characteristics and one or more classroom performance effects. For example, the association relationship is (("high teacher questioning frequency", "high teaching tool usage time"), ("high student concentration")), which means that if the preset conditions of high teacher-student interaction rate and high teaching tool usage time can be achieved in the teaching classroom, then the classroom performance effect of high student concentration will be obtained. In some embodiments, the template determination module can determine the teaching classroom characteristics corresponding to the positive classroom performance effect as efficient characteristics based on the association relationship. For example, the above-mentioned "high student concentration" is a positive classroom performance effect, then the "high teaching questioning rate" and "high teaching power usage time" can be used as efficient characteristics.

[0143] In some embodiments, the template determination module can represent different levels of classroom characteristics and different levels of teaching performance as events. For example, "high teacher-student interaction rate" in the classroom characteristics can be an event, "high teaching tool usage time" can be an event, etc. For another example, "high student concentration" in the classroom performance effect can be an event, etc.

[0144] In some embodiments, the template determination module can use association rule mining algorithms based on high-quality classroom data to calculate the support of different event sets, determine frequent item sets based on the support, and establish candidate association relationships based on the frequent item sets, and then determine the association relationship between teaching classroom characteristics and teaching performance effects based on the confidence and improvement of the candidate association relationships.

[0145] Frequent item sets are combinations of events that occur frequently. For example, event combinations could include ("low teacher question frequency," "high explanation duration," "middle school student test accuracy"), etc. In some embodiments, the template determination module may combine multiple events corresponding to classroom characteristics extracted from high-quality classroom data with multiple events corresponding to classroom performance exceeding a preset standard extracted from high-quality classroom data to obtain multiple event combinations, and calculate the support of each event combination to obtain frequent item sets.

[0146] In some embodiments, the template determination module can be based on high-quality classroom data and extracurricular performance data, and determine multiple events, event combinations, and their occurrence times in high-quality classroom data through statistics. For example, the number of occurrences of high teacher question frequency in high-quality classroom data. For more information on how to obtain teacher question frequency, please refer to Figure 2 and its related descriptions. For the description of extracurricular performance data, see Figure 3 and its related descriptions.

[0147] In some embodiments, the template determination module may determine the event combination whose corresponding support is greater than the minimum support as a frequent set item. Support refers to the frequency of occurrence of the event combination. In some embodiments, support may be related to the number of occurrences of the event combination in high-quality classroom data. For example, support may be positively correlated with the number of occurrences of the event combination in high-quality classroom data. The template determination module may calculate support using the following formula (1): (1) in, represents the support of the event combination, Indicates the number of occurrences of the event combination, Indicates the number of all events in the historical data.

[0148] The minimum support is the ratio of the data volume used to indicate that multiple specific events or features occur in the association rule mining algorithm to the total data volume. In some embodiments, the minimum support can filter out association rules and eliminate contingencies.

[0149] A candidate association relationship refers to the association relationship between the classroom characteristics to be determined and the classroom performance effect. In some embodiments, the template determination module can establish a candidate association relationship based on the classroom characteristics and classroom performance effects in the frequent item set. For example, a candidate association relationship is represented as (("high teacher question frequency", "high teaching tool usage time"), ("high student attention")).

[0150] Confidence refers to the credibility of a candidate relationship. Lift refers to the direction of the correlation between classroom characteristics and classroom performance in a candidate relationship. Correlation directions can include positive and negative correlations. A lift of 1 or higher indicates a positive correlation between classroom characteristics and classroom performance in the relationship; a lift of less than 1 indicates a negative correlation between classroom characteristics and classroom performance in the relationship.

[0151] In some embodiments, the template determination module may calculate the confidence corresponding to the candidate association relationship through a confidence algorithm. The confidence algorithm is shown in the following formula (2): (2) in, represents the confidence of the candidate association relationship, Indicates the number of occurrences of the event combination, It represents the number of occurrences of events corresponding to the classroom teaching features in the event combination in high-quality classroom data.

[0152] In some embodiments, the template determination module may calculate the confidence corresponding to the candidate association relationship by using a lifting algorithm. The lifting algorithm is shown in the following formula (3): (3) in, represents the lift of the candidate relationship, Indicates the number of occurrences of the event combination, It represents the number of occurrences of the event corresponding to the classroom performance effect in the event combination in the high-quality classroom data. The representational meanings of T and n1 are as mentioned above.

[0153] In some embodiments, the template determination module may use candidate associations with a lift greater than a lift threshold and a confidence greater than a minimum confidence as associations between classroom characteristics and classroom performance. The minimum confidence is a parameter used in association rule mining algorithms to determine the minimum likelihood of a review condition in an association rule. In some embodiments, the minimum confidence can measure the credibility of the current association rule.

[0154] Step 530: Determine an efficient teaching template based on the efficient features.

[0155] The efficient teaching template is used to instruct teachers to conduct classroom teaching. In some embodiments, the template determination module can use efficient features as standard conditions for the teaching template to obtain an efficient teaching template. For example, if "high teacher questioning frequency" has been determined to be an efficient feature, then the efficient teaching template is allocated a requirement that the teacher's questioning rate is not less than 2 times / min. For another example, if "high teaching tool usage time" is determined to be an efficient feature, then 25 minutes of class time is allocated in the efficient teaching template for tool operation practice.

[0156] In some embodiments of this specification, high-quality classroom data can be accurately identified through teaching quality data and misjudgment confidence, and combined with association rule mining algorithms, efficient features can be accurately extracted, and quantifiable university teaching templates can be generated to ensure the quality of template generation, quantify teachers' execution of templates, and improve classroom efficiency.

[0157] It should be noted that the above descriptions of processes 200, 300, and 500 are for illustration and purpose only and do not limit the scope of this specification. Those skilled in the art may, under the guidance of this specification, make various modifications and alterations to processes 200, 300, and 500. However, such modifications and alterations remain within the scope of this specification.

[0158] One or more embodiments of this specification provide a smart class inspection device, including a processor, which is used to execute the aforementioned smart class inspection method.

[0159] One or more embodiments of this specification provide a computer-readable storage medium, which stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the aforementioned smart class inspection method.

[0160] While the basic concepts have been described above, it will be apparent to those skilled in the art that the detailed disclosure is merely illustrative and does not limit this specification. Although not explicitly stated herein, various modifications, improvements, and revisions to this specification may be made by those skilled in the art. Such modifications, improvements, and revisions are suggested in this specification and remain within the spirit and scope of the exemplary embodiments of this specification.

Claims

1. A smart class inspection method, characterized in that: include: Acquire monitoring data through sensor devices in the classroom; Determining teacher-student behavior data and teacher-student text data based on the monitoring data; Generating teaching quality data of each teaching class based on the teacher-student behavior data and the teacher-student text data; The teaching quality data includes estimated quality scores and potential teaching anomalies; Determining the misjudgment confidence of each teaching class based on the teaching quality data and the historical quality data of the teaching classes associated with each teaching class, and generating sampling parameters for each teaching class; The sampling parameters include a rejection range; Based on the sampling parameters, the teaching quality data is updated to obtain updated teaching quality data; Based on the updated teaching quality data, class inspection parameters are determined.

2. The smart class inspection method according to claim 1, characterized in that: The sampling parameters also include monitoring sampling parameters; The updating of the teaching quality data based on the sampling parameters includes: Determine the stage division data based on historical teaching data; Determining the potential teaching abnormality points based on the stage division data; Determining the monitoring sampling parameters based on the potential teaching abnormal points; Determining updated teacher-student behavior data and teacher-student text data based on the monitoring sampling parameters and the elimination range; The teaching quality data is updated based on the updated teacher-student behavior data and teacher-student text data.

3. The smart class inspection method according to claim 1, characterized in that: Determining the class inspection parameters based on the updated teaching quality data includes: Determining the classrooms to be inspected based on the updated teaching quality data; Based on the classroom to be inspected, the class inspection parameters are determined.

4. The smart class inspection method according to claim 1, wherein: Also includes: Determining high-quality classroom data based on the teaching quality data and the misjudgment confidence level; Determining efficient features based on the high-quality classroom data and preset threshold data; Based on the efficient features, an efficient teaching template is determined.

5. A smart class inspection system, characterized by: include: A data acquisition module is configured to acquire monitoring data through sensor devices in the classroom; The data determination module is configured to: Based on the monitoring data, determining teacher-student behavior data and teacher-student text data; generating teaching quality data for each teaching class based on the teacher-student behavior data and the teacher-student text data; the teaching quality data includes an estimated quality score and potential teaching anomalies; and determining the misjudgment confidence of each teaching class based on the teaching quality data and the historical quality data of the teaching classes associated with each teaching class, and generating sampling parameters for each teaching class; The sampling parameters include a rejection range; a data updating module, configured to update the teaching quality data based on the sampling parameters to obtain updated teaching quality data; The parameter determination module is configured to determine the class inspection parameters based on the updated teaching quality data.

6. The intelligent class inspection system according to claim 5, characterized in that: The sampling parameters also include monitoring sampling parameters; The data update module is further configured to: Determine the stage division data based on historical teaching data; Determining the potential teaching abnormality points based on the stage division data; Determining the monitoring sampling parameters based on the potential teaching abnormal points; Determining updated teacher-student behavior data and teacher-student text data based on the monitoring sampling parameters and the elimination range; The teaching quality data is updated based on the updated teacher-student behavior data and teacher-student text data.

7. The intelligent class inspection system according to claim 5, characterized in that: The parameter determination module is further configured to: Determining the classrooms to be inspected based on the updated teaching quality data; Based on the classroom to be inspected, the class inspection parameters are determined.

8. The intelligent class inspection system according to claim 5, characterized in that: Also includes: The template determination module is configured to: Determining high-quality classroom data based on the teaching quality data and the misjudgment confidence level; Determining efficient features based on the high-quality classroom data and preset threshold data; Based on the efficient features, an efficient teaching template is determined.

9. A smart class inspection device, comprising a processor, characterized in that: The processor is used to execute the smart class inspection method as described in any one of claims 1 to 4.

10. A computer-readable storage medium storing computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the smart class inspection method according to any one of claims 1 to 4.

Images