A smart learning management system and method based on learning terminals

Abstract

This invention discloses a smart learning management system and method based on a learning terminal, belonging to the field of learning management systems. It includes: a user authentication module for receiving user login information and verifying identity; a learning record storage module for storing the user's historical learning data; a practice question generation module in response to successful authentication; a practice evaluation module to determine whether the user's learning mastery has met the required standards; a content linkage analysis module to calculate the linkage degree S; a forced review control module to force the user to complete review based on the linkage degree S; a learning status monitoring module for collecting video streams of the user's face and limbs during the learning process; and a non-verbal behavior analysis engine to identify abnormal attentional behaviors. This invention can adjust the forced review strategy based on the results of the previous learning content assessment and the correlation between new and old learning content, thereby ensuring that the user's final learning effect reaches the expected level.

Description

Technical Field

[0001] This invention relates to the field of learning management systems, specifically a smart learning management system and method based on a learning terminal. Background Technology

[0002] Smart learning is a process in which learners access learning resources on demand in a smart environment, flexibly conduct learning activities, and rapidly build knowledge and interpersonal networks. Its ultimate goal is to develop learners' learning wisdom and innovative abilities. Smart learning based on learning terminals combines intelligent devices with advanced technologies to construct a personalized, efficient, immersive, and continuous learning environment. Its core lies in being learner-centered, using terminal data to drive teaching decisions, achieving precise resource delivery, interactive collaboration, and real-time feedback, ultimately improving learning outcomes and innovative abilities. Typical applications include tablet-based teaching in junior high school mathematics and smart campus cloud platforms, demonstrating the profound transformation of education models through technology.

[0003] Existing patent 202311490140.7 discloses a learning management method and system based on a smart learning terminal, comprising: borrowing a smart learning terminal from a smart base station cabin via facial recognition; collecting user operation data based on the smart learning terminal to generate a database file; after the student returns the smart learning terminal, the smart base station cabin synchronizes the structured data in the database file to a central control service center via a data processing tool; the central control service center sends the synchronized data to a server, where supervisors perform corresponding approval processing and issue new tasks and content; the server synchronizes requested structured and unstructured data to the central control service center according to the user's needs, and the central control service center simultaneously distributes the data to the smart learning terminal via a data processing tool.

[0004] Existing technologies suffer from the following problems: before starting a new round of learning, users are not specifically assessed on the content learned previously, thus their mastery of that content is unknown. Furthermore, many adjacent knowledge units are interconnected, meaning that a lack of mastery of previously learned material can negatively impact the quality and efficiency of the new learning round. Existing review mechanisms are rigid and unable to dynamically adjust forced review strategies based on the connections between new and old knowledge, ultimately preventing users from achieving their expected learning outcomes. Therefore, those skilled in the art provide a smart learning management system and method based on a learning terminal to address the problems mentioned in the background. Summary of the Invention

[0005] The purpose of this invention is to provide a smart learning management system and method based on a learning terminal, which can conduct targeted assessments of the previous learning content before the user starts a new round of learning, and then adjust the forced review strategy according to the assessment results and the correlation between the new and old learning content, so as to ensure that the user's final learning effect reaches the expected level, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A smart learning management system based on a learning terminal, comprising:

[0008] The user authentication module is used to receive user login information and verify user identity.

[0009] The learning record storage module is used to store the user's historical learning data, which includes learning content identifiers, learning timestamps, and the duration of time spent on each knowledge unit.

[0010] The exercise generation module, in response to successful identity authentication, retrieves the latest learning record corresponding to the user, identifies at least three high-time-consuming knowledge units based on the user's dwell time in each knowledge unit in the learning record, and extracts a set of exercise questions from the question bank corresponding to the high-time-consuming knowledge units;

[0011] The practice assessment module is used to receive the user's answer data to the set of practice questions in real time, calculate the real-time score, and judge whether the learning mastery ability meets the standard based on the comparison between the learning time interval and the preset achievement rules.

[0012] The content linkage analysis module is activated when the practice assessment module determines that the learning mastery ability has not met the standard. It is used to analyze the correlation between the user's previous learning content and the new content to be learned, and to calculate the linkage degree S.

[0013] The review-forced control module forces users to complete a review for a duration of t1, t2, or t3 based on the degree of linkage S, where t1 < t2 < t3.

[0014] The learning status monitoring module includes a camera deployed on the learning terminal, used to collect video streams of the user's face and limbs during the learning process;

[0015] A non-verbal behavior analysis engine receives the video stream and identifies abnormal attentional behaviors using computer vision algorithms.

[0016] The learning anomaly weight calculation module calculates the learning anomaly weight value V based on attention anomaly behavior data;

[0017] The learning outcome correction module adjusts the learning outcome rating based on the V-value.

[0018] As a further aspect of the present invention: the specific process by which the exercise generation module extracts the exercise set is as follows:

[0019] (1) Extract the dwell time Ti for each knowledge unit, i=1···n, where n is a positive integer;

[0020] (2) Calculate the average length of stay ;

[0021] (3) Screening Ti > k The knowledge unit is considered as a high-time-consuming knowledge unit, where k is the difficulty coefficient adjustment factor and k≥1.5;

[0022] (4) Sort the knowledge units with the longest dwell time Ti from largest to smallest, and select the top m units with the longest dwell time as selected units, where m is a positive integer and m≥3;

[0023] (5) For each selected unit, randomly select q questions from the question bank to form a subset of practice questions, where q is a positive integer;

[0024] (6) Aggregate all subsets of practice questions to form the final set of practice questions.

[0025] As a further aspect of the present invention: the preset achievement rules in the practice assessment module are specifically as follows:

[0026] Preset pass score threshold P th The functional relationship with the time interval Δt is as follows:

[0027] ;

[0028] Among them, P max The highest threshold for achieving the target when Δt=0, 90%≤ and P max ≤100%, λ is the attenuation coefficient, and 0<λ<0.05, the unit of Δt is days;

[0029] If the user's real-time score P 实时 ≥P th If the user's real-time score P is within the acceptable range, it indicates that the user's learning and mastery ability has met the standard. 实时 <P th If so, it indicates that the user's learning and mastery ability has not met the standard.

[0030] As a further aspect of the present invention: the specific process by which the content linkage analysis module calculates the linkage degree S is as follows:

[0031] (a) Construct a knowledge graph, where nodes represent knowledge units and edges represent prerequisite relationships;

[0032] (b) Locate the previously learned content node set A and the content node set to be learned B;

[0033] (c) Calculate the set of shortest path lengths from A to B {L1, L2, ..., Lp};

[0034] (d) According to the formula Calculate the degree of linkage, where p is the number of elements in the set of shortest path lengths mentioned above.

[0035] As a further aspect of the present invention: the specific process of the forced review control module is as follows:

[0036] When S≤30%, mandatory review t1=5-10 minutes;

[0037] When 30% < S ≤ 60%, forced review t2 = 11-20 minutes;

[0038] When S > 60%, forced review t3 = 21-30 minutes;

[0039] Lock access to new content during review periods.

[0040] As a further aspect of the present invention: the non-verbal behavior analysis engine identifies the following abnormal attention behaviors:

[0041] (S1) The pupil diameter change rate is lower than the threshold or the eye movement frequency is lower than the preset value;

[0042] (S2) Blinking frequency is not within the preset range;

[0043] (S3) A single fixation lasts longer than 10 seconds without blinking;

[0044] (S4) The percentage of time the viewer's gaze is off the learning interface is ≥30%;

[0045] (S5) The tension of the interbrow muscles decreases more than the baseline value or the angle of trunk extension continues to increase;

[0046] (S6) Non-learning action frequency exceeds the standard (pen spinning > 5 times / min or leg shaking > 8 times / min);

[0047] (S7) Any combination of behaviors from (S1) to (S6) occurs for 3 consecutive minutes.

[0048] As a further aspect of the present invention: the learning anomaly weight calculation module calculates the V value using a formula, specifically:

[0049] ;

[0050] Where N is the number of anomalies and T is the cumulative duration of anomalies (minutes). When V≥0.7, it is judged as a significant learning anomaly.

[0051] As a further aspect of the present invention, it also includes a feedback adjustment module, used for:

[0052] Predict the optimal learning duration based on historical V-value sequences;

[0053] When the V value is ≥0.7 for 3 consecutive times, the single learning session is shortened by 10%;

[0054] Link abnormal behavior to the error rate analysis of practice questions;

[0055] When V≥0.8, push relaxation training; when 0.7≤V<0.8, reduce the difficulty of the exercises.

[0056] As a further aspect of the present invention: the specific process by which the learning effect correction module corrects the learning effect rating based on the V value is as follows:

[0057] Generate a learning diagnostic report that includes a heatmap of abnormal behavior;

[0058] When V≥0.8, learning is paused and a preset fatigue relief program is activated;

[0059] When 0.7 ≤ V < 0.8, the cognitive load intensity of subsequent learning content is dynamically adjusted;

[0060] The V value is mapped to a learning efficiency index of 0-100 and stored in the user profile.

[0061] This application also discloses a smart learning management method based on a learning terminal, which employs a smart learning management system based on a learning terminal and includes the following steps:

[0062] Store users' historical learning data;

[0063] Receive user login information and verify identity;

[0064] In response to successful identity authentication, the latest learning record of the user is retrieved. Based on the time the user spends in each knowledge unit in the learning record, at least three high-time-consuming knowledge units are identified, and a set of practice questions is extracted from the question bank corresponding to the high-time-consuming knowledge units.

[0065] The system receives user responses to the set of practice questions in real time, calculates real-time scores, and determines whether the user's learning and mastery ability meets the standards based on a comparison between the learning time interval and the preset achievement rules.

[0066] When the practice assessment module determines that the learning mastery ability has not met the standard, it is activated to analyze the correlation between the user's previous learning content and the new content to be learned, and to calculate the degree of linkage S.

[0067] Based on the degree of linkage S, users are forced to complete a review for a duration of t1, t2, or t3, where t1 < t2 < t3.

[0068] The system uses cameras deployed on learning terminals to capture video streams of the user's face and limbs during the learning process.

[0069] Receive the video stream and identify abnormal attentional behaviors using computer vision algorithms;

[0070] Calculate the abnormal learning weight value V based on data of abnormal attention behavior;

[0071] The learning performance rating is adjusted based on the V-value.

[0072] Compared with the prior art, the beneficial effects of the present invention are:

[0073] This invention significantly improves review efficiency and avoids ineffective practice by dynamically identifying time-consuming knowledge units and generating targeted exercises. A time-decay threshold model quantifies the forgetting curve, making the assessment of mastery more aligned with cognitive patterns. An innovative three-level mandatory review mechanism with a linkage level S effectively prevents the spread of knowledge gaps. A non-verbal behavior analysis engine accurately detects multiple types of attentional anomalies and calculates the learning anomaly weight value V by integrating the number of anomalies and their duration, overcoming the limitations of traditional learning systems that rely solely on answer data. When V ≥ 0.7, tiered intervention (reducing difficulty / shortening time / pausing learning) is automatically triggered, combined with dynamic optimization of cognitive load by the feedback adjustment module, thus improving the learning efficiency index. Finally, a closed-loop system of "practice assessment - content linkage - status monitoring - effect correction" is constructed, which, through practical testing, reduces review time costs, decreases ineffective learning time, and improves the accuracy of knowledge relevance identification. Attached Figure Description

[0074] Figure 1 This is a structural block diagram of a smart learning management system based on a learning terminal. Detailed Implementation

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] As mentioned in the background section of this application, research has found that users do not undergo targeted assessments of their previous learning content before starting a new round of learning. Consequently, it is unclear whether users have mastered the content from the previous learning. Furthermore, many adjacent knowledge units are interconnected, which means that if the knowledge from the previous learning is not mastered, it will affect the quality and efficiency of the new round of learning. The existing review mechanism is rigid and cannot dynamically adjust the forced review strategy according to the correlation between new and old knowledge, thus causing the user's final learning effect to fall short of expectations, which has certain shortcomings.

[0077] To address the aforementioned shortcomings, this application discloses a smart learning management system and method based on a learning terminal. This system can conduct targeted assessments of the previous learning content before a user begins a new round of learning, and then adjust the mandatory review strategy based on the assessment results and the correlation between the new and old learning content, thereby ensuring that the user's final learning effect reaches the expected level.

[0078] The following will describe in detail, with reference to the accompanying drawings, how the solution of this application solves the above-mentioned technical problems.

[0079] Please see Figure 1In this embodiment of the invention, a smart learning management system based on a learning terminal includes: a user authentication module for receiving user login information and verifying identity; a learning record storage module for storing the user's historical learning data, which includes learning content identifiers, learning timestamps, and dwell time for each knowledge unit; a practice question generation module that, in response to successful authentication, retrieves the user's latest learning record, identifies at least three high-time-consuming knowledge units based on the user's dwell time in each knowledge unit, and extracts a set of practice questions from the question bank corresponding to the high-time-consuming knowledge units; and a practice evaluation module for receiving the user's answers to the practice question set in real time, calculating real-time scores, and judging based on a comparison between the learning time interval and preset achievement rules. The learning process includes: a content linkage analysis module (activated when the practice assessment module determines that the learning mastery is not up to standard) to analyze the correlation between the user's previous learning content and the new content to be learned, calculating the linkage degree S; a forced review control module (forced to complete a review for a duration of t1, t2, or t3 based on the linkage degree S, where t1 < t2 < t3); a learning status monitoring module (including a camera deployed on the learning terminal to collect video streams of the user's face and limbs during the learning process); a non-verbal behavior analysis engine (receiving the video stream and identifying abnormal attention behaviors through computer vision algorithms); a learning anomaly weight calculation module (calculating the learning anomaly weight value V based on the abnormal attention behavior data); and a learning effect correction module (correcting the learning effect rating based on the V value). This application constructs a smart learning management framework based on multi-module collaboration, realizing a closed-loop process of "identity authentication → learning record analysis → dynamic practice question generation → mastery assessment → content linkage decision → learning status monitoring".

[0080] In this embodiment, the specific process of the exercise generation module extracting the exercise set is as follows: (1) Extract the dwell time Ti of each knowledge unit, i=1···n, where n is a positive integer; (2) Calculate the average dwell time. (3) Screening Ti > k The knowledge units are used as high-time-consuming knowledge units, where k is the difficulty coefficient adjustment factor and k≥1.5; (4) Sort the high-time-consuming knowledge units from largest to smallest according to the dwell time Ti, and select the first m high-time-consuming knowledge units as selected units, where m is a positive integer and m≥3; (5) For each selected unit, randomly select q questions from the question bank to form a subset of practice questions, where q is a positive integer; (6) Aggregate all the subsets of practice questions to form the final set of practice questions. Through dynamic threshold (Ti>k· The system accurately quantifies "time-consuming knowledge units" to ensure that practice questions are generated to address users' real weaknesses. This feature solves the problem of traditional systems blindly extracting practice questions.

[0081] In this embodiment, the preset achievement rule in the practice assessment module is specifically: a preset achievement score threshold P. th The functional relationship with the time interval Δt is as follows: Among them, P max The highest threshold for achieving the target when Δt=0, 90%≤ and P max ≤100%, λ is the attenuation coefficient, and 0<λ<0.05, Δt is in days; if the user's real-time score P 实时 ≥P th If the user's real-time score P is within the acceptable range, it indicates that the user's learning and mastery ability has met the standard. 实时 <P th If this is not the case, it indicates that the user's learning and mastery ability has not met the standard. This setting uses an exponential decay function. The standards for achieving the target are dynamically adjusted to conform to the forgetting curve.

[0082] In this embodiment, the specific process of the content linkage analysis module calculating the linkage degree S is as follows: (a) Constructing a knowledge graph, where nodes represent knowledge units and edges represent prerequisite relationships; (b) Locating the previously learned content node set A and the content node set to be learned B; (c) Calculating the set of shortest path lengths from A to B {L1, L2, ..., Lp}; (d) According to the formula Calculate the degree of linkage. This setting calculates the degree of linkage S based on the topological relationship of the knowledge graph (the inverse mean of the shortest path), realizing an objective measurement of content relevance and breaking through the limitations of traditional systems that rely solely on manual annotation of relevance.

[0083] In this embodiment, the forced review process of the forced review control module is as follows: when S≤30%, forced review t1=5-10 minutes; when 30%<S≤60%, forced review t2=11-20 minutes; when S>60%, forced review t3=21-30 minutes; and new content access is locked during the review period. This setting triggers differentiated forced review durations (t1 / t2 / t3) according to the linkage level S, balancing learning efficiency and knowledge coherence (longer review is required when S>60%), and the time threshold design (5-30 minutes) conforms to the attention maintenance limit.

[0084] In this embodiment, the nonverbal behavior analysis engine identifies the following abnormal attention behaviors: (S1) The pupil diameter change rate is lower than the threshold or the eye movement frequency is lower than the preset value. The pupil diameter change rate per unit time is calculated using a pupil tracking algorithm (such as Ellipse Fitting). Under normal cognitive activity, the pupil will periodically contract and expand (average change rate ≥ 0.5 mm / s). When the change rate is continuously < 0.2 mm / s, it is judged as "dullness". The eye movement frequency is calculated by the displacement vector of iris feature points. During normal reading, the eye scanning rate is 2-4 times / second. If < 0.5 times / second, an alarm is triggered; (S2) The blinking frequency is not within the preset range. The normal physiological range is 10-25 times / minute; (S3) A single gaze lasts for more than 10 seconds without blinking. If it lasts for 10 seconds, it is judged as "false focus"; (S4) The duration of the gaze deviating from the learning interface is ≥ 30%. At this time, the knowledge absorption rate drops significantly; (S5) The tension of the interbrow muscles decreases more than the baseline value or the trunk tilt angle continues to increase. The FACS (Face Angle Classification System) is used to detect this. The system calculates the rate of change in the depth of frown lines, with tension fluctuating within ±15% during normal thinking. A decrease of >25% is considered relaxation. The system detects the angle θ between the shoulder-hip line and the vertical axis at key skeletal points. If θ consistently exceeds 15° and increases at a rate >0.5° / s, it is considered "slouching." It integrates biomechanics (trunk angle) and micro-expression analysis (between the eyebrows) to verify negative states using dual signals. (S6) Excessive frequency of non-learning actions (e.g., pen spinning >5 times / min or leg shaking >8 times / min); (S7) Any combination of behaviors from (S1) to (S6) occurring continuously for 3 minutes. This setting transforms subjective teaching experiences (such as "daydreaming" or "drowsiness") into machine-detectable indicators. The combination condition design of S7 significantly reduces the false alarm rate (requires continuous 3 minutes of abnormality).

[0085] In this embodiment, the learning anomaly weight calculation module calculates the V value using a formula, specifically: Where N is the number of anomalies and T is the cumulative duration of anomalies (minutes). A significant learning anomaly is defined as V≥0.7. Anomaly frequency and duration are fused using a weighted formula to output a standardized evaluation value. The weighting coefficient (0.4 / 0.6) reflects the cognitive principle that duration is more important than frequency, and the threshold of V≥0.7 provides a unified intervention trigger point.

[0086] This embodiment also includes a feedback adjustment module, used for: predicting the optimal learning duration based on historical V-value sequences; shortening the single learning session by 10% when the V-value is ≥0.7 for three consecutive times; associating abnormal behavior with the error rate analysis of practice questions; pushing relaxation training when V ≥0.8, and reducing the difficulty of practice questions when 0.7 ≤ V < 0.8. This setting dynamically optimizes learning parameters (duration / difficulty) based on historical V-value data and associates abnormal behavior with error analysis, forming a self-evolutionary cycle of "monitoring → diagnosis → adjustment". Shortening the learning duration by 10% when V ≥0.7 for three consecutive times prevents the cumulative effect of fatigue.

[0087] In this embodiment, the specific process of the learning effect correction module correcting the learning effect rating based on the V value is as follows: A learning diagnostic report containing a heatmap of abnormal behavior is generated; when V ≥ 0.8, learning is paused and a preset fatigue relief program is initiated; when 0.7 ≤ V < 0.8, the cognitive load intensity of subsequent learning content is dynamically adjusted; the V value is mapped to a learning efficiency index of 0-100 and stored in the user profile. The V value is mapped to operable commands (pause learning / reduce load / generate report), and the learning efficiency index is output. The heatmap report visualizes learning obstacles (improving user engagement), and the learning efficiency index quantifies the user profile (laying the foundation for personalized recommendations).

[0088] This application also discloses a smart learning management method based on a learning terminal, employing a smart learning management system based on a learning terminal, including the following steps: storing the user's historical learning data; receiving user login information and verifying identity; in response to successful identity verification, retrieving the user's latest learning record, identifying at least three high-time-consuming knowledge units based on the user's dwell time in each knowledge unit in the learning record, and extracting a set of practice questions from the question bank corresponding to the high-time-consuming knowledge units; receiving the user's answer data to the set of practice questions in real time, calculating the real-time score, and judging whether the learning mastery ability meets the standard based on the comparison between the learning time interval and the preset achievement rules; when the practice evaluation module determines that the learning mastery ability does not meet the standard, it is activated to analyze the correlation between the user's previous learning content and the new content to be learned, and calculate the linkage degree S; according to the linkage degree S, the user is forced to complete a review for a duration of t1, t2, or t3, where t1 < t2 < t3; capturing video streams of the user's face and limbs during the learning process through a camera deployed on the learning terminal; receiving the video stream and identifying abnormal attention behavior through computer vision algorithms; calculating a learning abnormality weight value V based on the abnormal attention behavior data; and correcting the learning effect rating based on the V value.

[0089] This invention significantly improves review efficiency and avoids ineffective practice by dynamically identifying time-consuming knowledge units and generating targeted exercises. A time-decay threshold model quantifies the forgetting curve, making the assessment of mastery more aligned with cognitive patterns. An innovative three-level mandatory review mechanism with a linkage level S effectively prevents the spread of knowledge gaps. A non-verbal behavior analysis engine accurately detects multiple types of attentional anomalies and calculates the learning anomaly weight value V by integrating the number of anomalies and their duration, overcoming the limitations of traditional learning systems that rely solely on answer data. When V ≥ 0.7, tiered intervention (reducing difficulty / shortening time / pausing learning) is automatically triggered, combined with dynamic optimization of cognitive load by the feedback adjustment module, thus improving the learning efficiency index. Finally, a closed-loop system of "practice assessment - content linkage - status monitoring - effect correction" is constructed, which, through practical testing, reduces review time costs, decreases ineffective learning time, and improves the accuracy of knowledge relevance identification.

[0090] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0091] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A smart learning management system based on a learning terminal, characterized in that, include: The user authentication module is used to receive user login information and verify user identity. The learning record storage module is used to store the user's historical learning data, which includes learning content identifiers, learning timestamps, and the duration of time spent on each knowledge unit. The exercise generation module, in response to successful identity authentication, retrieves the latest learning record corresponding to the user, identifies at least three high-time-consuming knowledge units based on the user's dwell time in each knowledge unit in the learning record, and extracts a set of exercise questions from the question bank corresponding to the high-time-consuming knowledge units; The practice assessment module is used to receive the user's answer data to the set of practice questions in real time, calculate the real-time score, and judge whether the learning mastery ability meets the standard based on the comparison between the learning time interval and the preset achievement rules. The content linkage analysis module is activated when the practice assessment module determines that the user's learning mastery ability has not met the standard. It analyzes the correlation between the user's previous learning content and the new content to be learned, calculating the linkage degree S. The specific process of calculating the linkage degree S by the content linkage analysis module is as follows: (a) Constructing a knowledge graph, where nodes represent knowledge units and edges represent prerequisite relationships; (b) Locating the node set A of the previously learned content and the node set B of the content to be learned; (c) Calculating the set of shortest path lengths from A to B {L1, L2, ..., Lp}; (d) According to the formula... Calculate the degree of linkage; The review-forced control module forces users to complete a review for a duration of t1, t2, or t3 based on the degree of linkage S, where t1 < t2 < t3. The learning status monitoring module includes a camera deployed on the learning terminal, used to collect video streams of the user's face and limbs during the learning process; A non-verbal behavior analysis engine receives the video stream and identifies abnormal attentional behaviors using computer vision algorithms. The learning anomaly weight calculation module calculates the learning anomaly weight value V based on attention anomaly behavior data; The learning outcome correction module adjusts the learning outcome rating based on the V-value.

2. The intelligent learning management system based on a learning terminal according to claim 1, characterized in that, The specific process by which the exercise generation module extracts the exercise set is as follows: (1) Extract the dwell time Ti for each knowledge unit, i=1···n, where n is a positive integer; (2) Calculate the average length of stay ; (3) Screening Ti > k The knowledge unit is considered as a high-time-consuming knowledge unit, where k is the difficulty coefficient adjustment factor and k≥1.5; (4) Sort the knowledge units with the longest dwell time Ti from largest to smallest, and select the top m units with the longest dwell time as selected units, where m is a positive integer and m≥3; (5) For each selected unit, randomly select q questions from the question bank to form a subset of practice questions, where q is a positive integer; (6) Aggregate all subsets of practice questions to form the final set of practice questions.

3. The intelligent learning management system based on a learning terminal according to claim 2, characterized in that, The preset achievement rules in the practice assessment module are as follows: Preset pass score threshold P th The functional relationship with the time interval Δt is as follows: ； Among them, P max The highest threshold for achieving the target is when Δt=0, and 90%≤P max ≤100%, λ is the attenuation coefficient, and 0<λ<0.05, the unit of Δt is days; If the user's real-time score P 实时 ≥P th If the user's real-time score P is within the acceptable range, it indicates that the user's learning and mastery ability has met the standard. 实时 <P th If so, it indicates that the user's learning and mastery ability has not met the standard.

4. The intelligent learning management system based on a learning terminal according to claim 3, characterized in that, The specific process of forced review in the forced review control module is as follows: When S≤30%, mandatory review t1=5-10 minutes; When 30% < S ≤ 60%, forced review t2 = 11-20 minutes; When S > 60%, forced review t3 = 21-30 minutes; Lock access to new content during review periods.

5. The intelligent learning management system based on a learning terminal according to claim 4, characterized in that, The non-verbal behavior analysis engine identifies the following abnormal attentional behaviors: (S1) The pupil diameter change rate is lower than the threshold or the eye movement frequency is lower than the preset value; (S2) Blinking frequency is not within the preset range; (S3) A single fixation lasts longer than 10 seconds without blinking; (S4) The percentage of time the viewer's gaze is off the learning interface is ≥30%; (S5) The tension of the interbrow muscles decreases more than the baseline value or the angle of trunk extension continues to increase; (S6) Non-learning action frequency exceeds the standard; (S7) Any combination of behaviors from (S1) to (S6) occurs for 3 consecutive minutes.

6. The intelligent learning management system based on a learning terminal according to claim 5, characterized in that, The learning anomaly weight calculation module calculates the V value using a formula, specifically: ； Where N is the number of anomalies and T is the cumulative duration of anomalies. When V≥0.7, it is judged as a significant learning anomaly.

7. A smart learning management system based on a learning terminal according to claim 6, characterized in that, It also includes a feedback adjustment module, used for: Predict the optimal learning duration based on historical V-value sequences; When the V value is ≥0.7 for 3 consecutive times, the single learning session is shortened by 10%; Link abnormal behavior to the error rate analysis of practice questions; When V≥0.8, push relaxation training; when 0.7≤V<0.8, reduce the difficulty of the exercises.

8. The intelligent learning management system based on a learning terminal according to claim 7, characterized in that, The specific process by which the learning performance correction module corrects the learning performance rating based on the V value is as follows: Generate a learning diagnostic report that includes a heatmap of abnormal behavior; When V≥0.8, learning is paused and a preset fatigue relief program is activated; When 0.7 ≤ V < 0.8, the cognitive load intensity of subsequent learning content is dynamically adjusted; The V value is mapped to a learning efficiency index of 0-100 and stored in the user profile.

9. A smart learning management method based on a learning terminal, characterized in that, The intelligent learning management system based on a learning terminal as described in any one of claims 1-8 includes the following steps: Store users' historical learning data; Receive user login information and verify identity; In response to successful identity authentication, the latest learning record of the user is retrieved. Based on the time the user spends in each knowledge unit in the learning record, at least three high-time-consuming knowledge units are identified, and a set of practice questions is extracted from the question bank corresponding to the high-time-consuming knowledge units. The system receives user responses to the set of practice questions in real time, calculates real-time scores, and determines whether the user's learning and mastery ability meets the standards based on a comparison between the learning time interval and the preset achievement rules. When the practice assessment module determines that the learning mastery ability has not met the standard, it is activated to analyze the correlation between the user's previous learning content and the new content to be learned, and to calculate the degree of linkage S. Based on the degree of linkage S, users are forced to complete a review for a duration of t1, t2, or t3, where t1 < t2 < t3. The system uses cameras deployed on learning terminals to capture video streams of the user's face and limbs during the learning process. Receive the video stream and identify abnormal attentional behaviors using computer vision algorithms; Calculate the abnormal learning weight value V based on data of abnormal attention behavior; The learning performance rating is adjusted based on the V-value.

Images