Online examination system and method for adaptive adjustment of cognitive pressure

Abstract

The invention provides an online examination system and method capable of adaptively adjusting cognitive pressure, the system comprises an examinee interaction terminal, a physiological signal acquisition device and a processing device, and the examinee interaction terminal presents examination content to examinees and receives interaction input of the examinees; the physiological signal acquisition equipment acquires multi-modal physiological signals of examinees in real time; the processing equipment obtains a cognitive pressure index representing the cognitive pressure level of the examinee in real time through a pressure evaluation model based on the multi-modal physiological signal of the examinee; dynamically generating a regulation and control instruction according to the cognitive pressure index and a preset self-adaptive rule; and adjusting examination parameters presented to the examinee and / or triggering pressure relief intervention measures based on the regulation and control instruction. According to the method, a closed-loop process including signal acquisition, pressure evaluation, decision regulation and control and intervention execution is constructed, so that the cognitive pressure of the examinee can be objectively quantified and adjusted in real time, personalized self-adaptive examination experience is realized, and the performance of the examinee is improved while the examination validity is ensured.

Description

Technical Field

[0001] This invention relates to the fields of online education and human-computer interaction technology, and in particular to an online examination system and method with adaptive adjustment of cognitive stress. Background Technology

[0002] With the development of information technology, online examination systems have become an important tool in the fields of education and professional certification. Traditional online examination systems typically employ fixed test paper structures and uniform time limits, placing all candidates in the same testing environment. This "one-size-fits-all" approach ignores the psychological and physiological differences among individual candidates. This is particularly true in various educational assessment and professional certification systems, where numerous examination scenarios (such as course completion exams, professional qualification exams, and skills level certifications) focus on "accurately assessing the mastery of specific knowledge systems." The fundamental purpose of such exams is to objectively determine whether candidates have reached the preset knowledge and ability standards, and the assessment results should reflect the candidates' true knowledge level to the greatest extent possible.

[0003] However, in important exam scenarios, test takers generally experience varying degrees of cognitive pressure, such as tension, anxiety, and excessive mental load. Moderate pressure can stimulate optimal performance, but excessive pressure can significantly impair cognitive functions, such as memory retrieval, logical reasoning, and decision-making abilities, thus preventing exam scores from accurately reflecting their knowledge level. Conversely, if the exam is too easy, test takers may feel bored or become complacent, also failing to achieve the intended assessment purpose.

[0004] Currently, some advanced adaptive learning systems can dynamically adjust the difficulty of subsequent questions based on the test taker's accuracy rate, but this adjustment only addresses the dimension of knowledge mastery. These systems lack objective and effective means of perception and intervention mechanisms to address the test taker's real-time psychological pressure during the test. Summary of the Invention

[0005] This invention provides an online examination system and method for adaptive adjustment of cognitive stress, which addresses the problem that existing online examination systems cannot objectively and in real time assess and effectively regulate examinee cognitive stress, potentially leading to distorted exam performance and poor user experience, and achieves a personalized adaptive examination experience.

[0006] This invention provides an online examination system with adaptive adjustment of cognitive stress, comprising: The candidate interaction terminal is configured to present exam content to candidates and receive their interactive input. The physiological signal acquisition device is configured to acquire multimodal physiological signals of examinees in real time; The processing device is communicatively connected to the candidate's interactive terminal and the physiological signal acquisition device. The processing device is configured as follows: The multimodal physiological signals received by the physiological signal acquisition device are acquired and processed by a stress assessment model to obtain a cognitive stress index that characterizes the cognitive stress level of the examinee in real time. Based on the cognitive stress index and preset adaptive rules, dynamic control instructions are generated; The control instructions are sent to the candidate's interactive terminal to adjust the examination parameters presented on the candidate's interactive terminal and / or trigger stress relief intervention measures.

[0007] According to the cognitive stress adaptive adjustment online examination system provided by the present invention, the multimodal physiological signals include at least one: Cardiovascular electrophysiological signals or skin conductance signals acquired through contact sensing units; Visual behavioral signals acquired through non-contact sensing units, including at least one of the examinee's facial micro-expressions, eye movement trajectories, or body posture characteristics.

[0008] The online examination system for adaptive adjustment of cognitive stress provided by the present invention further includes an edge computing node, which is located between the physiological signal acquisition device and the processing device, and is configured to: Before sending the multimodal physiological signals to the processing device, at least one of the following preprocessing methods is performed: time synchronization alignment, noise reduction filtering, and feature extraction, to generate multimodal feature vectors. The multimodal feature vector is sent to the processing device.

[0009] According to the online examination system for adaptive adjustment of cognitive stress provided by the present invention, when the processing device dynamically generates adjustment instructions based on the cognitive stress index and preset adaptive rules, it is specifically configured as follows: When the duration of the cognitive stress index within the first preset stress range reaches the first preset duration, a first-level response is generated to adjust the subsequent question selection strategy. The first-level response includes: reducing the average cognitive difficulty level of the subsequently presented questions, and / or extending their average answering time limit. When the duration of the cognitive stress index within the second preset stress range reaches the second preset duration, a secondary response is generated to adjust the question presentation sequence. The secondary response includes: pausing the current question type and inserting a basic question with a cognitive load level lower than the current question. The second preset stress range is higher than the first preset stress range.

[0010] In the online examination system for adaptive adjustment of cognitive stress provided by the present invention, the processing device is further configured to perform at least one of the following operations: When the rate of increase of the cognitive stress index per unit time exceeds a preset threshold, an instruction to trigger the playback of mindfulness-guided content is automatically generated. In response to a stress reduction request initiated by a candidate through the candidate's interactive terminal, stress relief intervention content of corresponding intensity is matched and triggered from a preset intervention content library based on the real-time cognitive stress index at the time the stress reduction request is triggered.

[0011] According to the cognitive stress adaptive adjustment online examination system provided by the present invention, the system further includes a dynamic question scheduling module, which is linked to the processing device and configured to: Access a question bank database pre-labeled with question cognitive load levels and knowledge domain tags; Based on the real-time cognitive stress index and the knowledge domain status of the questions already answered by the examinee, an optimization algorithm is used to dynamically select the next question to be presented.

[0012] According to the cognitive stress adaptive adjustment online examination system provided by the present invention, the system further includes a security and privacy protection module, which is configured to perform at least one of the following operations: End-to-end encrypted transmission is performed on the raw physiological signals acquired from the physiological signal acquisition device; The original physiological signals are subjected to feature extraction or desensitization processing in a trusted execution environment at the processing device end; A preset data retention strategy is applied to the raw physiological signals to automatically destroy or anonymize them after stress assessment is completed.

[0013] This invention also provides an online examination method for adaptive adjustment of cognitive stress, applied to any of the online examination systems described above, comprising the following steps: Real-time acquisition of multimodal physiological signals from test takers; The multimodal physiological signals are analyzed based on a pre-trained stress assessment model to obtain a cognitive stress index that represents the cognitive stress level of examinees in real time. Based on the cognitive stress index and preset adaptive rules, dynamic control instructions are generated; Based on the aforementioned control instructions, the examination parameters presented to the candidates are adjusted, and / or stress relief interventions are triggered.

[0014] The present invention also provides a server for use in any of the above-described online examination systems, the server comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, when the processor executes the computer program, it performs the following steps: Receive multimodal physiological signals from test takers; Based on the aforementioned multimodal physiological signals, and processed through a pre-trained stress assessment model, a cognitive stress index representing the cognitive stress level of examinees is obtained in real time. Based on the cognitive stress index and preset adaptive rules, dynamic control instructions are generated; Based on the aforementioned control instructions, the adjustment of examination parameters and / or the triggering of stress relief interventions are controlled.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an online examination method for adaptive adjustment of cognitive stress as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements an online examination method for adaptive adjustment of cognitive stress as described above.

[0017] The online examination system and method for adaptive adjustment of cognitive stress provided by this invention integrates multimodal physiological signals and uses a stress assessment model for calculation, enabling real-time and objective quantification of examinees' cognitive stress levels. This overcomes the limitations of traditional methods that rely on subjective reports or single behavioral indicators, resulting in more accurate assessment results. Furthermore, a closed-loop system is constructed, encompassing signal acquisition, stress assessment, decision-making regulation, and executive intervention. This closed-loop system dynamically adjusts according to each examinee's real-time status, achieving a personalized examination experience tailored to each individual and helping to maintain examinees in their optimal cognitive state. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the structure of the online examination system for adaptive adjustment of cognitive stress provided by the present invention.

[0020] Figure 2 This is a schematic diagram of the pressure adaptive control principle provided by the present invention.

[0021] Figure 3 This is one of the flowcharts of the online examination method for adaptive adjustment of cognitive stress provided by the present invention.

[0022] Figure 4 This is the second flowchart of the online examination method for adaptive adjustment of cognitive stress provided by the present invention.

[0023] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0026] Example 1: Please see Figure 1 This embodiment provides an online examination system with adaptive adjustment of cognitive stress. The system includes: a candidate interaction terminal 10, a physiological signal acquisition device 20, an optional edge computing node 30, and a processing device 40.

[0027] Among them, the candidate interaction terminal 10 is configured to present the examination content to the candidate and receive the candidate's interactive input; Physiological signal acquisition device 20 is configured to acquire multimodal physiological signals of examinees in real time; Processing device 40 is communicatively connected to the candidate's interactive terminal and the physiological signal acquisition device; The processing device is configured as follows: Based on the multimodal physiological signals received from the physiological signal acquisition device, and processed by a stress assessment model, a cognitive stress index characterizing the cognitive stress level of the examinee is obtained in real time. Based on the cognitive stress index and preset adaptive rules, dynamic control instructions are generated; The control instructions are sent to the candidate's interactive terminal to adjust the examination parameters presented on the candidate's interactive terminal and / or trigger stress relief intervention measures.

[0028] Specifically, the candidate interaction terminal 10 can be a personal computer, tablet computer, or any smart device with a display screen and input devices. It runs examination client software or accesses the examination platform through a browser, and is responsible for presenting test questions, options, multimedia content, and receiving the candidate's input (such as clicks, text input) and active requests (such as requests for stress reduction).

[0029] The physiological signal acquisition device 20 is used to non-invasively acquire the multimodal physiological signals of examinees in real time.

[0030] In one specific embodiment, the multimodal physiological signal includes at least one: Cardiovascular electrophysiological signals or skin conductance signals acquired through contact sensing units; Visual behavioral signals acquired through non-contact sensing units, including at least one of the examinee's facial micro-expressions, eye movement trajectories, or body posture characteristics.

[0031] In a preferred embodiment, the physiological signal acquisition device 20 includes: Non-contact sensing unit 21: Typically a high-definition camera (such as a webcam). It is used to acquire visual behavioral signals including the examinee's face, eyes, and upper body posture.

[0032] Contact sensing unit 22: This can be a wearable device (such as a smart bracelet or a finger clip sensor) that integrates multiple sensors. It is used to collect cardiovascular electrophysiological signals (such as heart rate and heart rate variability (HRV) obtained by photoplethysmography (PPG)) and skin conductance response (GSR) signals.

[0033] In one specific embodiment, the online examination system further includes an edge computing node located between the physiological signal acquisition device and the processing device, and configured to: Before sending the multimodal physiological signals to the processing device, at least one of the following preprocessing methods is performed: time synchronization alignment, noise reduction filtering, and feature extraction, to generate multimodal feature vectors. The multimodal feature vector is sent to the processing device.

[0034] Specifically, edge computing node 30 is deployed closer to the examinee, such as as a standalone process running on the examinee's interactive terminal 10, or a hardware box connected via USB. Its main function is to preprocess the raw, high-frequency physiological signals generated by the acquisition device 20 to reduce network bandwidth usage and the computational burden on the processing device 40. Specific tasks include: Time synchronization alignment: Align data streams from different sensors (camera, wristband) on timestamps.

[0035] Noise reduction filtering: Processing the original signal, such as using a bandpass filter to remove noise from the PPG signal and smoothing the GSR signal.

[0036] Feature extraction: Extracting key features from the preprocessed signal. For example: From visual signals, facial landmark detection and tracking techniques (such as using the Dlib library) are used to extract the activation frequency (related to micro-expressions), average blink frequency, average gaze duration, and head posture change frequency of facial action units.

[0037] Real-time heart rate (HR), time-domain metrics of heart rate variability (such as SDNN, RMSSD), and frequency-domain metrics (such as LF / HF ratio) are calculated from the PPG signal.

[0038] From the GSR signal, the skin conductance level (SCL) and the number of peaks in the skin conductance response (SCRs) are calculated. After processing, the edge computing node 30 sends these lightweight multimodal feature vectors to the processing device 40.

[0039] The processing device 40 is the "brain" of the system. Internally, it logically includes the following functional modules: Stress Level Assessment Module 41: This module receives multimodal feature vectors from edge computing node 30 (or directly from acquisition device 20). Its core is a stress assessment model, preferably a pre-trained Long Short-Term Memory (LSTM) network, as it is particularly well-suited for processing time-series data. The training process for this LSTM model is as follows: A group of subjects is recruited and asked to complete cognitive tasks of varying difficulty levels (such as mental arithmetic and Stroop tests) in an experimental environment. Simultaneously, their multimodal physiological signals and subjective stress reports (such as the State-Trait Anxiety Inventory (STAI)) are collected. The collected feature vectors are used as model input, and the subjective stress scores are used as labels for supervised learning training. After training, the model can output a continuous, normalized cognitive stress index (e.g., a floating-point number from 0 to 1, with higher values ​​representing greater stress) based on the input real-time feature vectors.

[0040] Adaptive Decision Engine 42: This module is the core of the control logic. It continuously receives the cognitive stress index calculated by the stress level assessment module 41 and generates control instructions according to preset adaptive rules.

[0041] In one specific embodiment of the present invention, when the processing device dynamically generates control instructions based on the cognitive stress index and preset adaptive rules, it is specifically configured as follows: When the duration of the cognitive stress index within the first preset stress range reaches the first preset duration, a first-level response is generated to adjust the subsequent question selection strategy. The first-level response includes: reducing the average cognitive difficulty level of the subsequently presented questions, and / or extending their average answering time limit. When the duration of the cognitive stress index within the second preset stress range reaches the second preset duration, a secondary response is generated to adjust the question presentation sequence. The secondary response includes: pausing the current question type and inserting a basic question with a cognitive load level lower than the current question. The second preset stress range is higher than the first preset stress range.

[0042] Specifically, such as Figure 2 As shown, the decision engine can define multiple pressure state ranges, such as: comfort zone (0-0.4), optimal zone (0.4-0.6), high pressure zone (first pressure range, 0.6-0.8), and overload zone (second pressure range, >0.8).

[0043] When the decision engine detects that the cognitive stress index remains in the high-pressure zone (0.6-0.8) for a first preset duration (e.g., 60 seconds), it triggers a level 1 response, generating instructions to adjust the subsequent question selection strategy. For example, it notifies the dynamic question scheduling module 50 to prioritize questions with lower cognitive load levels.

[0044] When the index is detected to remain in the overload zone (>0.8) for a second preset duration (e.g., 30 seconds), a secondary response is triggered, generating instructions to adjust the question presentation sequence, pausing the current question and inserting a simple relaxation task unrelated to the exam content (e.g., finding the differences in a picture).

[0045] In one specific embodiment of the present invention, the processing device is further configured to perform at least one of the following operations: When the rate of increase of the cognitive stress index per unit time exceeds a preset threshold, an instruction to trigger the playback of mindfulness-guided content is automatically generated. In response to a stress reduction request initiated by a candidate through the candidate's interactive terminal, stress relief intervention content of corresponding intensity is matched and triggered from a preset intervention content library based on the real-time cognitive stress index at the time the stress reduction request is triggered.

[0046] Specifically, when the rate of increase of the index within a unit of time (e.g., 10 seconds) exceeds a threshold (e.g., the slope is greater than 0.05), indicating a sharp increase in the test taker's stress, mindfulness guidance is automatically triggered, for example, by instructing the test taker's interactive terminal 10 to play an animation guiding the user to take deep breaths in full screen.

[0047] In one specific embodiment of the present invention, the online examination system with adaptive adjustment of cognitive pressure further includes a dynamic question scheduling module 50, which is linked to the processing device and configured to: Access a question bank database pre-labeled with question cognitive load levels and knowledge domain tags; Based on the real-time cognitive stress index and the knowledge domain status of the questions already answered by the examinee, an optimization algorithm is used to dynamically select the next question to be presented.

[0048] Specifically, the dynamic question scheduling module 50 manages a vast question bank database. Each question in the database is pre-labeled with a knowledge domain tag and a cognitive load level (which can be multi-dimensionally labeled by education experts based on question complexity, required knowledge depth, etc.). This module receives strategy instructions from the adaptive decision engine 42 and simultaneously tracks the knowledge point coverage of the examinees. Its core is an optimization algorithm used to balance the following two objectives: 1) complying with the pressure control requirements of the decision engine; 2) maximizing the knowledge point coverage of the exam.

[0049] In one specific embodiment of the present invention, the cognitive stress adaptive adjustment online examination system further includes a security and privacy protection module 60, which is configured to perform at least one of the following operations: End-to-end encrypted transmission is performed on the raw physiological signals acquired from the physiological signal acquisition device; The original physiological signals are subjected to feature extraction or desensitization processing in a trusted execution environment at the processing device end; A preset data retention strategy is applied to the raw physiological signals to automatically destroy or anonymize them after stress assessment is completed.

[0050] Specifically, the security and privacy protection module 60 is responsible for data security. Raw physiological signals are discarded after feature extraction at the edge computing node 30, and the feature vectors are end-to-end encrypted using the TLS 1.3 protocol during transmission to the processing device 40. At the processing device, all sensitive computations (such as model inference) can be performed in a Trusted Execution Environment (TEE, such as Intel SGX). The processing device implements strict retention policies for all examinee data, such as automatically anonymizing or completely destroying the data 24 hours after the exam.

[0051] Example 2: Still refer to Figure 1In the specific implementation of this embodiment, the overall architecture of the cognitive stress adaptive adjustment online examination system includes four collaborative units: the candidate interaction terminal is responsible for the presentation of examination content and human-computer interaction, the physiological monitoring device realizes multimodal signal acquisition, the processing equipment for central processing undertakes the intelligent analysis and decision-making function, and the secure storage module ensures the integrity of data and privacy compliance throughout the process.

[0052] At the physiological signal acquisition level, the system employs a fusion sensing strategy. Contact sensing units continuously capture cardiovascular activity characteristics and skin conductance changes via wearable devices; these bioelectrical signals directly reflect the stress level of the autonomic nervous system. Non-contact sensing units utilize optical imaging technology to capture facial micro-expressions, eye movement trajectories, and body posture features, extracting behavioral indicators related to cognitive load through computer vision algorithms. The two types of signals achieve millisecond-level time synchronization at the hardware level, ensuring spatiotemporal consistency for subsequent analysis. A layered encryption mechanism is used during signal transmission; original data packets are accompanied by unique identifiers and timestamps, laying the foundation for subsequent correlation analysis.

[0053] The signal preprocessing stage is deployed on edge computing nodes. The raw physiological data first undergoes noise reduction filtering to eliminate distortion caused by environmental electromagnetic interference and motion artifacts. For cardiovascular signals, the system identifies key waveform features and calculates rhythm variability indices; for electrodermal signals, it focuses on conductivity change trends and transient response intensity; for visual behavioral signals, a deep learning model identifies specific facial expression patterns and gaze focus drift characteristics.

[0054] The preprocessing results are encoded into standard feature vectors and uploaded to the central processing equipment through a dedicated channel, significantly reducing network transmission load.

[0055] The stress assessment model is the intelligent hub of the system. It adopts a multi-source feature fusion architecture. First, it performs spatiotemporal alignment and normalization on the feature vectors of different physiological modes to eliminate the bias caused by individual differences in sensors.

[0056] The core algorithm establishes a mapping relationship between feature combinations and cognitive stress levels through a machine learning framework. This framework uses large-scale stress scenario experimental data during the training phase, covering typical physiological response patterns of different age groups and cultural backgrounds in cognitive tasks.

[0057] The model outputs a standardized pressure level value, which is dynamically updated at a fixed frequency, and includes a confidence level assessment parameter for reliability verification.

[0058] The workflow of the adaptive decision engine is as follows: Figure 2 As shown, the exam parameter adjustment implements a tiered strategy: moderate pressure triggers a first-level response (difficulty / time adjustment), and high pressure activates a second-level response (question type change).

[0059] The adaptive decision engine constructs a two-layer control mechanism: The first layer is a stress threshold response strategy: when the system detects that the stress level is consistently in the medium range, it automatically triggers the primary regulation mode, such as appropriately reducing the abstract thinking requirements of subsequent questions or fine-tuning the answering time for a single question; if the stress level enters the high-risk range, the secondary regulation mode is activated, temporarily inserting basic question types with lower cognitive load to stabilize the emotional state.

[0060] The second layer is an active intervention mechanism: the system monitors the rate of pressure change in real time and automatically pushes a breathing rhythm guidance program when it detects a sharp increase in pressure level; at the same time, it retains a self-regulation channel for candidates, allowing them to request structured relaxation training based on their subjective feelings, and the system will intelligently match the intervention intensity based on the real-time pressure status. Intervention can be implemented on candidates through linkage between the intervention actuator and the candidate's interactive terminal.

[0061] The system is equipped with an intelligent question scheduling mechanism, and the exam content scheduling system enables personalized test path management: Each question in the question bank database is pre-labeled with multi-dimensional attribute tags, including knowledge domain affiliation, cognitive processing depth, and problem-solving strategy category. The dynamic scheduling algorithm uses real-time stress trends as its core input and combines them with the test takers' historical performance characteristics to make collaborative decisions: during high-pressure phases, questions with shallower cognitive processing depth and higher knowledge relevance are prioritized; when the stress returns to a normal threshold, a gradual difficulty recovery strategy is adopted to avoid a secondary impact on test takers caused by sudden changes in cognitive load.

[0062] The entire process automatically records the correlation between stress fluctuations and test-taking performance, forming a traceable cognitive state evolution map to identify knowledge weaknesses that are susceptible to stress.

[0063] Security and privacy protection systems are implemented throughout the entire data processing lifecycle: Physiological signals are encrypted from the moment they are acquired, and key management employs a dynamic rotation mechanism. During processing, the raw physiological data undergoes feature extraction and desensitization in an isolated, trusted computing environment, retaining only non-sensitive feature vectors after desensitization. The storage system implements a tiered retention strategy: raw signal data is automatically destroyed after stress assessment, feature data is retained for no more than the entire examination cycle, and only aggregated analysis-level data can be stored long-term for teaching and research. All data access operations are logged.

[0064] The ethics control module defines strict boundaries for data use and establishes a principle of physical separation between stress data and exam assessment: cognitive stress indicators serve only the real-time adaptive adjustment function and are completely excluded from the input parameters of the scoring algorithm; any derivative reports involving stress analysis must be generated with the explicit authorization of the examinee; stress reduction intervention operation records adopt a two-way confirmation mechanism, and examinees can terminate the intervention process or delete relevant records at any time. This design ensures that the application of technology is always within the ethical framework.

[0065] System performance is confirmed through a multi-dimensional verification mechanism: From a cognitive science perspective, the system maintains stable performance of test takers within the optimal stress performance range, preventing anxiety from encroaching on working memory resources. In terms of educational measurement, it significantly reduces random errors and problem-solving strategy collapse caused by stress, making test results more accurately reflect the level of knowledge mastery. At the technical implementation level, the closed-loop control system has real-time response capability, and all control operations are completed within a delay range that is imperceptible to the examinees, ensuring the continuity of the examination process.

[0066] Practical applications show that the system can balance the fairness of assessment for candidates with different stress sensitivity traits, and in particular optimize the performance stability of high-anxiety groups in key exams.

[0067] The system's scalability is reflected in its three-part architectural design: the hardware layer reserves interfaces for multiple types of sensors, supporting seamless integration with future new physiological monitoring technologies; the algorithm layer adopts a modular stress assessment framework, allowing feature extractors or classifiers to be replaced according to specific examination scenarios; and the strategy layer provides a visual rule configuration interface, enabling educational institutions to customize stress thresholds and intervention plans. This open architecture ensures that the system can continuously adapt to the evolving needs of educational measurement technologies.

[0068] More specifically, the cognitive stress adaptive online examination system is implemented through a distributed hardware architecture. The candidate's interactive terminal uses a computing device with an integrated touch screen. This device has a built-in dynamic interface rendering engine that uses a responsive design framework and can adjust the density of questions, font size, and color contrast in real time according to instructions.

[0069] The front-facing camera is equipped with an infrared fill light module for pupil tracking in low-light environments; the microphone array supports directional noise reduction to separate voice content from ambient noise.

[0070] The contact-type physiological sensing unit includes flexible fabric ECG electrodes that are directly attached to the chest to collect ECG signals. The signal sampling rate is set to meet the minimum requirements for heart rate variability analysis. The wrist-worn skin conductivity sensor uses a dual-electrode constant voltage design to continuously measure conductivity changes caused by sweat gland activity. Data transmission uses a low-power wireless protocol.

[0071] The non-contact monitoring unit processes the video stream using computer vision algorithms: the pupil tracking subsystem locates the iris contour and calculates the rate of change in diameter; the facial blood volume analysis subsystem extracts the RGB color fluctuation features of the cheekbone region and estimates the changes in microvascular blood volume.

[0072] The signal synchronization mechanism is implemented collaboratively by the hardware layer. All sensing units have built-in high-precision clock chips, and the time reference is aligned through a time synchronization protocol when powered on. The contact sensor data packet includes a millisecond-level timestamp, and each frame of the video stream is marked with the start time of acquisition. Edge computing nodes are deployed on local routers and equipped with dedicated signal processing chips; After receiving the raw data, the node first performs time alignment: the ECG signal is aligned with the peak of the R wave, the skin conductance signal is aligned with the start point of the rising edge, and the video features are aligned with the midpoint of the frame exposure. After alignment, a multimodal data block with a unified time axis is generated, and each 200 milliseconds is encapsulated as a transmission unit.

[0073] ECG signal preprocessing includes four steps: Step 1: The raw signal is first passed through a 0.5 to 40 Hz bandpass filter to eliminate baseline drift and electromyographic interference; Step 2: Next, the adaptive threshold method is used to detect the peak value of the R wave and calculate the interval sequence between adjacent R waves; Step 3: Then, outlier removal is performed on the interval series: if a certain interval deviates from the moving average by more than 20%, it is replaced by linear interpolation of the preceding and following intervals; Step 4: Finally, calculate the time-domain indices, including the standard deviation of the interval and the root mean square of the difference between adjacent intervals. For the frequency-domain indices, obtain the power spectral density ratio of low frequency to high frequency using Fast Fourier Transform.

[0074] Furthermore, phase segmentation algorithm is used for electrodermal signal processing: After the raw signal is high-pass filtered by 0.05 Hz, a steep rising edge with an amplitude exceeding 0.03 microSiemens is defined as the reaction initiation point. Three features are extracted: reaction rise slope, peak amplitude, and half-life recovery time. The average trigger frequency is calculated for consecutive reaction events, and individual differences are compensated for by incorporating readings from environmental temperature and humidity sensors.

[0075] Video signal processing implements three-level analysis: Level 1 facial localization: A convolutional neural network is used to detect the coordinates of 68 key points in each frame of the image; Second-level dynamic feature extraction: Tracking the trajectory of pupil diameter changes and calculating the coefficient of variation of diameter within the sliding window; analyzing the frequency of eyelid opening and closing and counting the number of incomplete blinks per minute; measuring the displacement of the corrugator supercilii muscle using optical flow method. The third-level time-series coding: downsamples the feature sequence to a frequency of 10 Hz to generate a standardized feature vector.

[0076] The feature fusion module runs at the edge nodes. It integrates five ECG indicators, four EEG parameters, and six video features into a fifteen-dimensional vector according to time windows. A multimodal feature fusion formula is used: (1), Among them, superscript and It is a timestamp or index label; This represents the time index, indicating the [number]th [time index]. Each sampling time; Represents the modal index, indicating the first modal index. Physiological signal modes, These correspond to electrocardiogram, skin conductance, and video visual signals, respectively. Indicates the first The calibration delay for each modality is used to align all signals on the time axis because the hardware and algorithm processing speeds of different sensors are different. This represents the feature vector after time delay compensation and normalization; it is the eigenvector of the th element. Types of modes in The original feature vector at time step; It is the first The weight matrix of each modality is obtained through training and is used to quantify the contribution of different physiological modalities to stress assessment. The resulting feature vectors are fused. Principal component analysis is used to compress them into an eight-dimensional feature space, eliminating intermodal correlations. Each feature packet is appended with a signal quality label: if the missing data rate of a certain sensor exceeds 30%, the confidence level of that dimension is marked as low.

[0077] For example, the stress assessment model uses a gradient boosting decision tree architecture: During the model training phase, a dedicated dataset was constructed: subjects completed a standardized cognitive stress test in the laboratory, physiological signals were recorded simultaneously, and a subjective stress scale was filled out every 30 seconds. The data labeling rules are as follows: when the subjective score is 7 or above for 10 seconds, this period is marked as a high-pressure sample; resting state data is marked as a low-pressure sample. In feature engineering, synthetic minority class oversampling techniques are used to balance the sample distribution; The model input is an eight-dimensional fused feature, and the output is a standardized stress value from 0 to 1. The training objective is to minimize the mean square error between the predicted value and the subjective rating. Hyperparameter optimization is achieved through grid search: the tree depth is fixed at six levels, the learning rate is set to 0.1, and the number of iterations is determined to be 120 rounds using the early stopping method.

[0078] The model deployment adopts a dual-track system of online and offline deployment: The online version loads pre-trained model weights, receives feature packets in real time to perform predictions, and outputs stress values ​​and confidence scores every 0.5 seconds. The offline version performs model drift detection every hour: it compares the real-time data distribution with the KL divergence of the training set, and triggers model retraining when the deviation exceeds the threshold. The retraining data comes from high-quality signal segments from recent exams, and the incremental training process is completed within an isolated sandbox.

[0079] The adaptive decision engine includes a state machine engine: The system defines five pressure states: calm state (pressure value less than 0.4); warning state (0.4 to 0.6); moderate pressure (0.6 to 0.8); high pressure (greater than 0.8); and sudden rise state (increase of more than 40% within 10 seconds).

[0080] The state transition rule is: when a certain state continues for a set duration, a response is triggered. The duration threshold is set through a configuration file.

[0081] The exam parameter adjustments employ a tiered strategy. Upon entering an alert state, the system records the baseline difficulty level of the current questions. If the system transitions to a moderate-pressure state for 15 seconds, a Level 1 response is activated: it retrieves similar questions from the question bank database with a difficulty level reduced by 0.15 to 0.25 to replace subsequent questions; or it extends the time limit for each question by 10% to 15%. If the system enters a high-pressure state for 10 seconds, a Level 2 response is activated: it pauses the current question type and inserts three basic questions with a cognitive load level lower than Level 2, reducing the scoring weight of these basic questions to 80% of the original question type. All adjustments are executed through a dynamic scheduling interface.

[0082] The stress reduction intervention module features dual-channel triggering. Automatic triggering of the channel monitoring state transitions: when a sudden increase is detected, the exam interface is immediately frozen, and a full-screen breathing guidance animation plays. The animation consists of four seconds of inhalation with a circular diffusion effect, seven seconds of breath-holding with a progress ring, and eight seconds of exhalation with a contraction animation, cycling twice before automatically resuming. A manual trigger channel provides a persistent stress reduction button; clicking it sends an encrypted request. The decision engine verifies real-time stress values: if greater than 0.5, a 60-second natural scene video with guided audio plays; if less than 0.5, a five-second miniature breathing prompt pops up. An independent timer is activated during intervention, pausing the exam countdown.

[0083] The dynamic scheduling system manages the question sequence, and each question in the question bank database is labeled with metadata: the cognitive load level adopts a five-level system, with level one being factual recall questions and level five being comprehensive reasoning questions; knowledge domain tags are linked to the subject knowledge graph; problem-solving strategies are divided into three categories: memory-based, application-based, and analysis-based. The scheduling algorithm maintains a state matrix and uses an optimization objective function: (2), in, Indicate whether to select a question ( Indicates selection. (Indicates not to select); This refers to finding the combination of problems that maximizes the objective function while satisfying all constraints (such as total time). ; This is a real-time pressure value, representing the pressure at... The cognitive stress level of the examinee is calculated by the stress assessment model at any time, and the value ranges from [0,1]. For the title The cognitive load level represents the level of cognitive effort required to answer the question (e.g., defined as levels 1 to 5), with higher levels indicating more difficult questions. For the title The knowledge coverage gain is used to measure the contribution of selecting this question to improving the coverage of the knowledge structure of this exam. The pressure-adaptive weighting coefficient is an adjustable parameter between 0 and 1. The pressure-load conversion factor (scaling factor) is used to convert the pressure value... Mapped to the cognitive load level of the question Within the range of comparable values.

[0084] The decision-making process is as follows: First, select the load level range based on the pressure status, limiting it to no more than level three under moderate pressure and no more than level two under high pressure; second, calculate the knowledge coverage weight in the candidate set; finally, select the optimal question number based on the remaining time. The knowledge coverage heatmap is updated after each scheduling to ensure the validity of the exam content.

[0085] The stress performance graph is constructed using temporal alignment technology, recording stress value trajectories at five-second intervals and simultaneously capturing answer events: question presentation time, number of option modifications, final submission time, and result. The analysis system establishes a mapping between stress ranges and answer performance: average response time is statistically analyzed within the 0.7 to 0.8 stress range; error type distribution is analyzed above 0.8. Clustering algorithms are used to identify stress-sensitive knowledge points, marking knowledge units where error rates surge under high pressure.

[0086] Data transmission security is implemented through layered encryption: The raw physiological signals are encrypted at the sensor end using the AES-256 algorithm, and the key is stored in a secure chip. A dedicated VPN tunnel is established for transmission between edge nodes and processing devices, and the data is encrypted twice using the national standard SM4 algorithm. A hash checksum is appended to the encrypted messages to prevent tampering during transmission. The receiving channel of the processing device is equipped with a traffic shaping mechanism to defend against denial-of-service attacks.

[0087] The data processing environment constructs a trusted execution space, with the central processing unit divided into secure enclaves equipped with hardware-level memory encryption modules. The physiological feature extraction process is completed within these enclaves: the raw signal is processed immediately after decryption, and the generated feature vector immediately overwrites the original memory area. The desensitization operation removes individual identification information, retaining only standardized feature values ​​and timestamps.

[0088] Data storage implements lifecycle management. Raw signal streams are written to a time-series database, with a retention policy of automatic deletion every 24 hours. Anonymized feature data is retained for 31 days for model optimization. Aggregated analysis data, after anonymization, is stored in a relational database, including stress interval statistical reports and knowledge point vulnerability assessments. Database access is subject to role-based control: question setters can only see question metadata; invigilators can see student status markers; system administrators can access anonymized aggregated data.

[0089] Ethical control mechanisms are implemented throughout the entire process. Stress data is stored in a physically isolated area, disconnected from the scoring system network. A confirmation dialog box pops up before the stress reduction intervention is initiated, allowing candidates to choose to skip or delay. All stress-related logs are stored independently, and general assessment reports are not accessible. The system provides a one-click exit option: after candidates disable physiological monitoring, the system switches to a fixed difficulty mode.

[0090] Sensor calibration follows a standardized process, with a three-minute benchmark test initiated before each exam: candidates sit still and watch a video of natural scenery while the system collects basic physiological parameters. The system calculates the resting mean of heart rate variability, baseline skin conductance, and normal pupil diameter as individualized calibration benchmarks. Calibration data is encrypted and stored, used only for signal interpretation during the specific exam session.

[0091] The model iteratively establishes a closed-loop optimization mechanism, extracting 5% of the exam data weekly to generate a stress pattern report, labeling three types of key events: accurate stress warning cases, false alarm events (low pressure misjudged as high pressure), and missed alarm events (high pressure not detected). A manual review team labels the nature of each event, adding confirmed samples to the training set. Incremental training is performed offline, and the new model is switched to online mode after one week of parallel validation in shadow mode.

[0092] Pre-defined rules are stored in a rule database, which provides a graphical configuration interface. After logging in, administrators of educational institutions can adjust stress threshold boundaries: modify the starting value for moderate stress, and set the trigger duration for high stress. They can also configure the range of question adjustments: define the maximum reduction in difficulty and the upper limit for time extension. Furthermore, they can manage the intervention content library: upload guided breathing videos and replace mindfulness audio resources. All configuration changes are logged, and version rollback is supported.

[0093] Typical application scenarios demonstrate the system's resilience. In a math competition scenario, when a candidate's stress level rose to 0.83 for 12 seconds during a geometry proof question, the system automatically skipped the next two comprehensive questions and inserted three basic calculation questions. After the stress level dropped to 0.62, the scheduling algorithm gradually recovered at 80% of the original difficulty curve. In a recruitment exam scenario, when a candidate triggered the stress relief button and their stress level reached 0.52, the system played a forest stream video accompanied by breathing guidance. After the intervention, the time limit for the next question was dynamically extended by 15%, and the evaluation report noted "The stress regulation strategy was effective."

[0094] The efficacy verification was conducted through multi-dimensional testing. In the cognitive efficacy test, monitoring using a dual-task paradigm with working memory showed that the response speed of secondary tasks improved by 28% in adaptive mode, confirming a reduction in cognitive resource consumption. Quantitative analysis showed that the standard deviation of performance in the high-anxiety group was 15.7 in the traditional mode, decreasing to 9.2 in the adaptive mode. In the technical performance test, the average delay from exceeding the stress threshold to completing the question change was 1.2 seconds, with no perceived interruption by the test takers.

[0095] The system's scalability design includes reserved interfaces, and the hardware layer defines a standard sensor protocol to support certified access for third-party devices. The algorithm layer encapsulates feature extraction interfaces, allowing for the replacement of novel stress assessment models. The strategy layer provides an open state machine rule engine, enabling educational institutions to import custom state transition logic. This architecture ensures the system continuously adapts to innovations in educational measurement technologies.

[0096] The cognitive stress adaptive adjustment online examination system provided in this embodiment dynamically senses the cognitive stress state of examinees and adaptively adjusts the examination environment, constructing a complete technical closed-loop process including signal acquisition, stress assessment, decision control, and feedback execution. This optimizes the accuracy of assessment results and effectively solves the problem of ability assessment deviation caused by stress fluctuations in traditional examinations.

[0097] Example 3: Please see Figure 3 This embodiment provides an online examination method for adaptive adjustment of cognitive stress. The method runs on the system described in Embodiment 1 and specifically includes the following steps: Step 301: Real-time acquisition of multimodal physiological signals from examinees; Step 302: Analyze the multimodal physiological signals based on a stress assessment model to obtain the cognitive stress index, which represents the cognitive stress level of the examinee, in real time. Step 303: Dynamically generate control instructions based on the cognitive stress index and preset adaptive rules; Step 304: Based on the control instructions, adjust the examination parameters presented to the examinee and / or trigger stress relief intervention measures.

[0098] The online examination method for adaptive adjustment of cognitive stress provided in this invention can monitor the cognitive stress level of examinees in real time and dynamically and adaptively adjust examination parameters or provide intervention measures based on the stress level to optimize examinee performance and improve the accuracy of examination assessment.

[0099] Example 4: Please see Figure 1 , 4 This embodiment provides a complete process for an online examination method that adaptively adjusts cognitive stress, specifically including the following steps: Step S401: System initialization and calibration; After logging into the testing system, candidates will undergo a brief baseline calibration phase before the official start (e.g., having candidates sit quietly and watch a neutral landscape video for 3 minutes). During this period, the system collects the candidate's basic physiological signals to calculate their personalized resting-state stress level baseline.

[0100] Step S402: Real-time acquisition and preprocessing; After the exam begins, the physiological signal acquisition device 20 continues to operate. The edge computing node 30 performs noise reduction, filtering, and feature extraction on the acquired raw signals in real time, generates a standardized multimodal feature vector stream, and sends it to the processing device 40.

[0101] Step S403: Calculate the cognitive stress index in real time; The stress level assessment module 41 of the processing device 40 receives the feature vector stream and inputs it into a pre-trained stress assessment model (such as an LSTM model) to calculate the cognitive stress index at the current moment in real time.

[0102] Step S404: Adaptive decision-making and instruction generation; The adaptive decision engine 42 receives the cognitive stress index and matches it against preset rules and state ranges for judgment: If the index is in the optimal zone, the system will proceed normally, with the dynamic question scheduling module 50 selecting the next question.

[0103] If the index triggers the duration rules of the high-pressure zone or overload zone, or triggers the pressure surge rule, the decision engine generates corresponding control instructions based on the specific rules.

[0104] If the system receives a stress reduction request initiated by the examinee through the interactive terminal 10, the decision engine will combine the current stress index and match and generate corresponding intervention instructions from the preset intervention content library (which includes music, videos, etc. of different intensities).

[0105] Step S405: Implement regulation and intervention; The processing device 40 sends the generated control instructions to the candidate interactive terminal 10 and / or the dynamic question scheduling module 50.

[0106] If the instruction is to adjust exam parameters, such as reducing the difficulty of subsequent questions or extending the answering time, the dynamic question scheduling module 50 will execute the strategy when selecting the next question.

[0107] If the instruction is a stress relief intervention, such as playing a breathing guidance animation or inserting a relaxation mini-game, it will be executed immediately by the candidate's interactive terminal 10, and the corresponding content will be displayed on the interface.

[0108] Step S406: Loop and End; The system continuously executes steps S402 to S405 in a loop until the exam ends. Throughout the process, the system dynamically and in a closed loop adjusts the examinee's stress level, striving to maintain them in an optimal state.

[0109] The present invention provides an adaptive cognitive stress online examination system and method. The system includes a candidate-end device (i.e., a candidate interaction terminal), a processing device, a physiological signal acquisition device, and a database. The physiological signal acquisition device is used to collect the candidate's physiological signals in real time. The processing device has a built-in stress level assessment model. By analyzing the physiological signals, it calculates the candidate's cognitive stress index in real time. Based on the cognitive stress index and preset adaptive rules, the online examination system dynamically adjusts the difficulty and number of questions presented to the candidate during the examination, as well as the time limit for answering questions, or provides stress relief interventions. This achieves personalized adaptation of the examination content and pace, thereby reducing the negative impact of stress on the candidate's true ability and improving the accuracy and fairness of the examination assessment.

[0110] In some specific embodiments of the present invention, this solution provides a server, including: a memory 530 and a processor 510; the memory 530 and the processor 510 communicate with each other via a communication bus 540; the memory 530 stores computer instructions that can be executed on the processor 510; when the processor 510 invokes the computer instructions, it can perform the following steps: Receive multimodal physiological signals from test takers; Based on the aforementioned multimodal physiological signals, and processed through a pre-trained stress assessment model, a cognitive stress index representing the cognitive stress level of examinees is obtained in real time. Based on the cognitive stress index and preset adaptive rules, dynamic control instructions are generated; Based on the aforementioned control instructions, the adjustment of examination parameters and / or the triggering of stress relief interventions are controlled.

[0111] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, communications interface 520, and memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute an online examination method that adaptively adjusts cognitive stress.

[0112] It should be noted that the electronic device in this embodiment can be a server, a PC, or other devices, as long as its structure includes the following: Figure 5 The processor 510, communication interface 520, memory 530, and communication bus 840 shown are interconnected via the communication bus 540. The processor 510 can call logical instructions stored in the memory 530 to execute the aforementioned method. This embodiment does not limit the specific implementation of the electronic device.

[0113] The server can be a single server or a group of servers. The server group can be centralized or distributed (for example, the servers can be a distributed system).

[0114] In possible embodiments, the server can be local or remote relative to the terminal. For example, the server can access information stored in the user terminal, a database, or any combination thereof via a network.

[0115] As another example, the server can directly connect to at least one of the user terminal and the database to access the information and / or data stored therein.

[0116] In possible embodiments, the server can be implemented on a cloud platform; by way of example only, the cloud platform may include private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, inter-cloud, multi-cloud, or any combination thereof.

[0117] In possible embodiments, the server and user terminal can be implemented on an electronic device having one or more components as described in the embodiments of the present invention.

[0118] Furthermore, networks can be used for the exchange of information and / or data.

[0119] In possible embodiments, one or more components in the interaction scenario (e.g., server, user terminal, and database) may send information and / or data to other components.

[0120] In possible embodiments, the network can be any type of wired or wireless network, or a combination thereof. By way of example only, the network may include a wired network, a wireless network, a fiber optic network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), wireless local area networks (WLANs), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, or a near field communication (NFC) network, or any combination thereof.

[0121] In possible embodiments, the network may include one or more network access points. For example, the network may include wired or wireless network access points, such as base stations and / or network switching nodes, through which one or more components of the interaction scenario can connect to the network to exchange data and / or information.

[0122] Furthermore, the logic instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0123] Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] In a possible embodiment, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the online examination method for adaptive adjustment of cognitive stress provided in the above embodiments.

[0125] In a possible embodiment, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to perform the methods provided in the above-described method embodiments.

[0126] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online examination system with cognitive stress self-adaptive adjustment, characterized in that, The system comprises: an examinee interaction terminal configured to present examination content to an examinee and receive interactive input of the examinee; a physiological signal acquisition device configured to acquire multi-modal physiological signals of the examinee in real time; a processing device communicatively connected with the examinee interaction terminal and the physiological signal acquisition device; wherein the processing device is configured to: obtain the multi-modal physiological signals received by the physiological signal acquisition device, process the multi-modal physiological signals through a stress evaluation model to obtain a cognitive stress index representing cognitive stress level of the examinee in real time; generate a regulation instruction dynamically according to the cognitive stress index and a preset adaptive rule; send the regulation instruction to the examinee interaction terminal to adjust the examination parameters presented by the examinee interaction terminal, and / or trigger a stress relief intervention measure.

2. The online examination system for cognitive stress self-adaptive adjustment according to claim 1, wherein, The multi-modal physiological signals comprise at least one of: cardiovascular electrophysiological signals or galvanic skin response signals acquired through a contact sensing unit; visual behavior signals containing at least one of facial micro-expression, eye movement trajectory or body posture features of the examinee acquired through a non-contact sensing unit.

3. The online examination system of adaptive adjustment of cognitive stress according to claim 1 or 2, characterized in that, The system further comprises an edge computing node located between the physiological signal acquisition device and the processing device, and configured to: perform at least one of time synchronization alignment, noise reduction filtering and feature extraction on the multi-modal physiological signals before sending the multi-modal physiological signals to the processing device to generate a multi-modal feature vector; send the multi-modal feature vector to the processing device.

4. The online examination system for cognitive stress self-adaptive adjustment according to claim 1, wherein, When generating the regulation instruction dynamically according to the cognitive stress index and the preset adaptive rule, the processing device is specifically configured to: generate a first-level response for adjusting subsequent question selection strategy when the cognitive stress index is within a first preset stress interval for a duration reaching a first preset time length, the first-level response comprising: reducing the average cognitive difficulty level of subsequent presented questions, and / or extending the average answering time limit thereof; generate a second-level response for adjusting question presentation sequence when the cognitive stress index is within a second preset stress interval for a duration reaching a second preset time length, the second-level response comprising: pausing the current question type and inserting a basic question with a cognitive load level lower than the current question; wherein the second preset stress interval is higher than the first preset stress interval.

5. The online examination system for cognitive stress self-adaptive adjustment according to claim 1 or 4, characterized in that, The processing device is further configured to perform at least one of the following operations: automatically generate an instruction for triggering the playing of mindfulness guidance content when it is monitored that the growth rate of the cognitive stress index within a unit time exceeds a preset threshold; in response to a stress relief request initiated by the examinee through the examinee interaction terminal, match and trigger stress relief intervention content with corresponding intensity from a preset intervention content library according to the real-time cognitive stress index at the time of the stress relief request.

6. The online examination system for cognitive stress self-adaptive adjustment according to claim 1, wherein, The system further comprises a dynamic question scheduling module, which is linked with the processing device and configured to: access a question database pre-labeled with question cognitive load level and knowledge field tag; dynamically select the next presented question through an optimization algorithm based on the real-time cognitive stress index and the knowledge field state covered by the recorded questions answered by the examinee.

7. The online examination system for cognitive stress self-adaptive adjustment according to claim 1, characterized in that, The system further comprises a security and privacy protection module configured to perform at least one of the following operations: implementing end-to-end encrypted transmission of raw physiological signals collected from the physiological signal collection device; performing feature extraction or desensitization processing on the raw physiological signals in a trusted execution environment at the processing device end; applying a preset data retention policy to the raw physiological signals to automatically destroy or anonymize them after completing stress assessment.

8. An online examination method of cognitive stress self-adaptive adjustment, characterized in that, The online examination system of any one of claims 1-7, comprising the following steps: real-time collection of multi-modal physiological signals of examinees; analysis of the multi-modal physiological signals based on a pre-trained stress assessment model to obtain a cognitive stress index representing cognitive stress level of the examinees in real time; dynamic generation of control instructions according to the cognitive stress index and preset adaptive rules; adjustment of examination parameters presented to the examinees and / or triggering of stress relief intervention measures based on the control instructions.

9. A server applied to the online examination system of any one of claims 1-7, comprising a memory and a processor, and a computer program is stored on the memory, characterized in that, The processor executes the computer program to implement the following steps: receiving multi-modal physiological signals of examinees; processing the multi-modal physiological signals based on a pre-trained stress assessment model to obtain a cognitive stress index representing cognitive stress level of the examinees in real time; dynamic generation of control instructions according to the cognitive stress index and preset adaptive rules; control of adjustment of examination parameters and / or triggering of stress relief intervention measures based on the control instructions.

10. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the online examination method of adaptive adjustment of cognitive stress according to claim 8.

Images