Search

CN-121979399-A - Intelligent student learning state monitoring method and system based on behavior analysis

CN121979399ACN 121979399 ACN121979399 ACN 121979399ACN-121979399-A

Abstract

The invention relates to the technical field of student learning state monitoring, in particular to an intelligent student learning state monitoring method and system based on behavior analysis, wherein a data acquisition unit acquires a glance path sequence and glance drop point coordinates of a student in a question examination stage in real time through eye movement tracking equipment; the glance judging unit builds a glance feature judging model, dynamically optimizes an angle deviation accumulation value threshold, strengthens abnormal steering contribution by combining a direction weight factor, distinguishes target and roaming glance through speed fluctuation double-threshold secondary verification in a critical interval, generates a keyword area through natural language processing and a visual saliency model, rapidly calculates Euclidean distance based on R tree index and coordinate normalization, maps the Euclidean distance to a matching degree score and divides a high-low matching area, and outputs a concentration state through a quaternary matrix by a state decision unit for continuously N times of target glance and the high-low matching area, accurately monitors a learning state and provides a reliable basis for teaching intervention.

Inventors

  • LIU YONG

Assignees

  • 北京中航未来科技集团有限公司

Dates

Publication Date
20260505
Application Date
20260408

Claims (10)

  1. 1. A student learning state intelligent monitoring method based on behavior analysis is characterized by comprising the following steps: s1, acquiring eye movement data of students in a question examination stage in real time through eye movement tracking equipment, wherein the eye movement data comprise a glance path sequence and glance drop point coordinate data; S2, constructing a glance feature judgment model, wherein the glance feature judgment model calculates angle deviation accumulated values among all continuous points on each glance path in a glance path sequence based on a preset path analysis algorithm, compares the angle deviation accumulated values with a preset threshold value, judges targeted glance and roaming glance according to comparison results, and performs secondary verification by combining glance speed fluctuation features when the angle deviation accumulated values are in a preset threshold value critical interval, wherein the glance speed fluctuation features are extracted based on the glance path sequence; S3, predefining boundary coordinates of a plurality of keyword areas based on topic content, generating a keyword area, calculating Euclidean distance between each glance falling point coordinate and the center of the nearest keyword area, mapping the Euclidean distance into a matching degree score, and dividing the matching degree score into a high matching area and a low matching area by comparing the matching degree score with a set distance threshold, wherein the falling point distance is smaller than a first threshold and is determined as the high matching area, and the falling point distance is larger than the first threshold and is determined as the low matching area; And S4, combining the targeted glance with the roaming glance and the high matching area and the low matching area to monitor and output the learning state of the students.
  2. 2. The intelligent monitoring method for student learning state based on behavior analysis according to claim 1, wherein the construction process of the glance characteristic judgment model comprises dynamically optimizing a preset threshold value, and specifically comprises the following steps: The machine learning module is used for analyzing the sample data distribution rule of the targeted glance and the roaming glance in the historical examination stage, automatically adjusting the judging threshold value of the angle deviation accumulated value, introducing the attention distribution theoretical model, adaptively correcting the verification intensity of the glance speed fluctuation feature according to the complexity coefficient of the topic text, enabling the attention distribution theoretical model to strengthen the path analysis algorithm weight in a simple topic, and improving the verification priority of the glance speed fluctuation feature in a complex topic.
  3. 3. The intelligent monitoring method for student learning state based on behavior analysis of claim 2, wherein the calculating of the angle deviation accumulation value between all continuous points in each glance path in the glance path sequence is specifically to traverse a continuous vector group formed by adjacent three points in the glance path sequence, calculate the absolute value of steering angle between every two adjacent vectors, and accumulate all absolute values of steering angles in the glance path to generate the angle deviation accumulation value, wherein the steering angle calculation adopts a vector angle cosine theorem reverse-pushing mechanism.
  4. 4. The intelligent monitoring method for student learning state based on behavior analysis of claim 3, wherein a direction weight factor is introduced when calculating an absolute value of a steering angle, the steering angle weight is reduced when the direction change of a continuous vector group accords with the subject reading directivity rule, the weight is increased if the direction mutation collides with the reading directivity rule, the contribution degree of abnormal steering is enhanced in an angle deviation accumulation value, and reasonable path deviation according with the cognition rule is weakened.
  5. 5. The intelligent student learning state monitoring method based on behavior analysis of claim 1, wherein the workflow of performing secondary verification in combination with glance speed fluctuation features is as follows: When the angle deviation accumulated value is in a preset threshold critical interval, an instantaneous speed sequence between adjacent sampling points in the glance path sequence is extracted, the standard deviation of the instantaneous speed sequence is calculated as a speed fluctuation characteristic value, a speed fluctuation double threshold is set, if the speed fluctuation characteristic value is lower than a first speed threshold, the target glance is judged, if the speed fluctuation characteristic value is higher than a second speed threshold, the roaming glance is judged, and if the speed fluctuation characteristic value is between the two speed thresholds, the preliminary judgment result of the angle deviation accumulated value is inherited.
  6. 6. The intelligent monitoring method for student learning state based on behavior analysis of claim 1, wherein the method is characterized in that a plurality of keyword area boundary coordinates are predefined based on topic content, topic text is analyzed specifically through a natural language processing model, core terms and associated logic words in a topic stem are identified as keywords, then the spatial concentration degree of each keyword in a topic page is calculated by combining a visual saliency model, adjacent keywords with the spatial concentration degree higher than a set value are combined into the same keyword area, and finally a keyword area set with hierarchical boundary coordinates is output.
  7. 7. The intelligent student learning state monitoring method based on behavior analysis of claim 1, wherein the Euclidean distance between each glance drop point coordinate and the center of the nearest keyword region is calculated, an R tree index structure of the keyword region is established by adopting a spatial index optimization algorithm, the center point of the keyword region nearest to the glance drop point is positioned by a pruning strategy, global traversal calculation is avoided, and meanwhile, page coordinate system normalization processing is introduced.
  8. 8. The intelligent student learning state monitoring method based on behavior analysis of claim 1, wherein the mapping of Euclidean distance to matching degree score adopts an inverse proportion function conversion mechanism, a maximum effective matching distance threshold is set, the matching degree score is maximum when the Euclidean distance is zero, the matching degree score is reduced to zero when the Euclidean distance reaches the maximum effective matching distance threshold, the division of a high matching region and a low matching region is realized through the comparison of the matching degree score and a dynamic matching threshold, and the dynamic matching threshold is adaptively adjusted according to the proportion of the total area of a keyword region to the total area of a subject page.
  9. 9. The intelligent student learning state monitoring method based on behavior analysis of claim 1, wherein the combination of targeted glance and roaming glance and high and low matching areas specifically comprises: And constructing a quaternary state judgment matrix, outputting a focus examination question state if all continuous N glance events are judged to be targeted glances and are high matching areas, triggering a distraction early warning candidate signal if a single glance event is judged to be roaming glance or low matching areas, and activating distraction early warning and recording abnormal behavior fragments when the distraction early warning candidate signal is accumulated for a threshold number of times in a preset time window.
  10. 10. A system for implementing a student learning state intelligent monitoring method based on behavioral analysis as claimed in any one of claims 1-9, comprising: The data acquisition unit (1) integrates a hardware interface of eye movement tracking equipment, captures original eye movement data of students in a question examination stage in real time, outputs a glance path sequence containing a time stamp and glance drop point coordinate data stream, and provides time sequence input for subsequent analysis; The glance judging unit (2) analyzes historical sample distribution based on a machine learning module, dynamically optimizes a judging threshold value of an angle deviation accumulated value, adopts a vector included angle cosine theorem back-pushing mechanism to calculate steering angles among continuous points in a traversing mode, superimposes direction weight factors to generate the angle deviation accumulated value, extracts an instantaneous speed sequence standard deviation when the angle deviation value is in a critical interval, and secondarily verifies glance types through double thresholds; The regional matching unit (3) identifies topic keywords through a natural language processing model, a hierarchical keyword regional boundary is generated by a linkage visual saliency model, quick Euclidean distance calculation between a glancing drop point and the center of the nearest keyword regional is realized based on an R tree index, page coordinate normalization processing is assisted, the Euclidean distance is converted into a matching degree score by adopting an inverse proportion function, and a high matching region and a low matching region are divided according to a dynamic matching threshold; The state decision unit (4) constructs a quaternary state decision matrix, outputs concentration states for N continuous targeting glances and high matching areas, generates a distraction early warning candidate signal for a roaming glance or low matching area, and activates distraction early warning and stores abnormal behavior fragments when the distraction early warning candidate signal accumulates over-threshold in a time window.

Description

Intelligent student learning state monitoring method and system based on behavior analysis Technical Field The invention relates to the technical field of student learning state monitoring, in particular to an intelligent student learning state monitoring method and system based on behavior analysis. Background The student learning state monitoring technology is an important technology, is particularly applied to a learning state monitoring link in a student examination stage, and is characterized in that the concentration degree is judged by accurately analyzing eye movement behaviors, a basis is provided for teaching intervention, the core requirement of accurately grasping the student learning state in real time is adapted to an education scene, the concentration state in the student examination stage is reflected by eye movement behaviors such as glancing paths, drop point distribution and the like, the eye movement characteristic performance is influenced due to the fact that different student examination habits are different, the targeting (with an explicit target) and focusing (focusing core information) of the eye movement behaviors are difficult to accurately distinguish due to the factors, misjudgment is easily caused due to single characteristic judgment in a traditional monitoring mode, the student distraction state or misjudgment concentration state cannot be timely identified, effective control and targeted intervention on the learning state are influenced, and in order to solve the technical problem, the intelligent student learning state monitoring method and system based on the behavior analysis are provided. Disclosure of Invention The invention aims to provide an intelligent student learning state monitoring method and system based on behavior analysis, which are used for solving the problems in the background technology. In order to achieve the above purpose, one of the purposes of the present invention is to provide an intelligent monitoring method for learning status of students based on behavior analysis, comprising the following steps: s1, acquiring eye movement data of students in a question examination stage in real time through eye movement tracking equipment, wherein the eye movement data comprise a glance path sequence and glance drop point coordinate data; S2, constructing a glance feature judgment model, wherein the glance feature judgment model calculates angle deviation accumulated values among all continuous points on each glance path in a glance path sequence based on a preset path analysis algorithm, compares the angle deviation accumulated values with a preset threshold value, judges targeted glance and roaming glance according to comparison results, and performs secondary verification by combining glance speed fluctuation features when the angle deviation accumulated values are in a preset threshold value critical interval, wherein the glance speed fluctuation features are extracted based on the glance path sequence; S3, predefining boundary coordinates of a plurality of keyword areas based on topic content, generating a keyword area, calculating Euclidean distance between each glance falling point coordinate and the center of the nearest keyword area, mapping the Euclidean distance into a matching degree score, and dividing the matching degree score into a high matching area and a low matching area by comparing the matching degree score with a set distance threshold, wherein the falling point distance is smaller than a first threshold and is determined as the high matching area, and the falling point distance is larger than the first threshold and is determined as the low matching area; And S4, combining the targeted glance with the roaming glance and the high matching area and the low matching area to monitor and output the learning state of the students. The second object of the present invention is to provide a system for implementing the intelligent monitoring method for learning status of students based on behavior analysis, which comprises: The data acquisition unit integrates a hardware interface of eye movement tracking equipment, captures original eye movement data of students in a question examination stage in real time, outputs a glance path sequence containing a time stamp and glance drop point coordinate data stream, and provides time-sequence input for subsequent analysis; The glance judging unit analyzes historical sample distribution based on the machine learning module, dynamically optimizes a judging threshold value of an angle deviation accumulated value, adopts a vector included angle cosine theorem back-pushing mechanism to calculate steering angles among continuous points in a traversing mode, superimposes direction weight factors to generate the angle deviation accumulated value, extracts an instantaneous speed sequence standard deviation when the angle deviation value is in a critical interval, and secondarily verifies glance types through doubl