Course behavior simulation education system fusing physiological signal perception

Abstract

The application provides a course behavior simulation education system fusing physiological signal perception, comprising: a physiological signal acquisition and synchronization module, which acquires electroencephalogram signals, electrocardiogram signals, skin electrical activity and eye movement tracking data of learners in a simulation education environment in real time through a non-invasive device, and accurately aligns and fuses the acquired physiological signals with time codes of the simulation environment and interactive behaviors of the learners; a data preprocessing and feature extraction module, which pre-processes the acquired multi-modal synchronous data stream to eliminate noise and interference, and extracts feature vectors with representative and diagnostic significance from the pre-processed data; and a physiological state evaluation module, which inputs the extracted multi-dimensional feature vectors into trained classification algorithms and regression algorithms, and maps the feature vectors into physiological state evaluation indexes of the learners through calculation and analysis of the algorithms.

Description

Technical Field

[0001] This invention relates to the field of simulation education technology, and in particular to a curriculum behavior simulation education system that integrates physiological signal perception. Background Technology

[0002] Simulation education, as a highly immersive learning method, allows learners to repeatedly practice complex skills in virtual environments, playing an increasingly crucial role in fields such as medicine, aviation, and emergency training. Its safety, repeatability, and controllability make it an important pathway to improve learning outcomes. However, existing simulation education systems primarily rely on learners' external actions and task completion to determine their learning status, making it difficult to accurately capture learners' real-time physiological changes. This often results in delayed or inaccurate adjustments to course content, hindering the optimization of immersive learning experiences.

[0003] During the learning process, attention, cognitive load, and emotional state fluctuate continuously. These changes are directly manifested through physiological signals, such as the intensity of brain waves, changes in heart rate, fluctuations in skin conductance, and eye movement patterns. If the system cannot perceive and understand these physiological signals in a timely manner, it cannot make corresponding adjustments when learners experience fatigue, distraction, or excessive stress. For example, when a learner is handling a complex surgical simulation task, their cognitive load may suddenly increase, their heart rate variability may decrease, and their fixation point may begin to jump frequently. However, if the system continues to push more information or increase the difficulty of the task as planned, it will lead to greater stress for the learner, and their learning efficiency will actually decrease. This disconnect between physiological state and the pace of the course limits the personalization and adaptability of simulation education.

[0004] A deeper problem lies in the fact that physiological signals are multimodal, continuous, and highly time-varying, with various signals interconnected and changing rapidly over time. Analyzing these signals in real-time within an immersive simulation environment and transforming them into a reliable assessment of the learning state presents a significant challenge. This is especially true when precise alignment of physiological signals with events and actions within the virtual environment is required; any time deviation or signal interference can distort the assessment results, thereby affecting subsequent course adjustments and feedback.

[0005] How to continuously and synchronously collect and analyze multimodal physiological signals while learners are fully immersed in simulation courses, and closely integrate them with environmental interaction data to achieve real-time understanding of the learning state, has become a key issue in improving the effectiveness of simulation education. Summary of the Invention

[0006] This invention provides a curriculum behavior simulation education system that integrates physiological signal perception, mainly comprising: The physiological signal acquisition and synchronization module uses non-invasive devices to collect learners' EEG signals, ECG signals, skin conductance activity, and eye-tracking data in a simulated educational environment in real time. It then precisely aligns and fuses the collected physiological signals with the timecode of the simulated environment and the learners' interactive behaviors. The data preprocessing and feature extraction module preprocesses the acquired multimodal synchronous data stream to eliminate noise and interference, and extracts representative and diagnostic feature vectors from the preprocessed data. The physiological state assessment module inputs the extracted multidimensional feature vectors into the trained classification and regression algorithms. Through the calculation and analysis of the algorithms, the feature vectors are mapped to the learner's physiological state assessment indicators. The simulation content optimization module, based on physiological state assessment indicators, applies a pre-set set of control strategy rules to drive the simulation engine to adjust the simulation education content in real time according to the control strategy rule set, thereby obtaining optimized simulation content.

[0007] Furthermore, the physiological signal acquisition and synchronization module acquires learners' electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, skin conductance, and eye-tracking data using non-invasive devices to obtain raw multimodal signal sets; The original multimodal signal group is filtered, and Gaussian filtering is used to remove noise to obtain a clean signal group. For the cleaning signal group, if an abnormal value is detected and exceeds a preset threshold, the corresponding part is removed to obtain a valid signal group. Timecode and interaction behavior data are obtained from the simulation environment, and the synchronization signal sequence is obtained by matching and aligning the effective signal group with the timecode. The synchronization signal sequence fusion interaction data is integrated using a weighted average method to obtain a multimodal synchronization data stream.

[0008] Furthermore, the data preprocessing and feature extraction module acquires a multimodal synchronous data stream through an acquisition device, and uses a low-pass filtering method to eliminate high-frequency noise in the data stream to obtain a clean signal sequence; The electroencephalogram power spectrum is extracted from the clean signal sequence, and the signal frequency domain representation is converted by fast Fourier transform. The heart rate variability index is calculated, and the standard deviation of the interval between adjacent heartbeats is obtained by time domain analysis to determine the physiological response baseline. The amplitude of skin conductance response is measured against the physiological response baseline to obtain the peak change of skin conductance level, a gaze path vector is constructed, and the eye movement trajectory coordinate sequence is tracked to obtain a preliminary feature set. The preliminary feature set is fused with real-time stream management attributes extracted from the data stream. The real-time stream management attributes are obtained by synchronous timestamp annotation. A multi-dimensional vector is integrated. If the vector dimension exceeds a preset threshold, the fusion weight is adjusted. The adjustment of the fusion weight is achieved by weighted average operation to obtain an optimized feature vector. Attention pattern analysis is performed based on the optimized feature vectors. Principal component analysis is used to reduce the dimensionality of the vector components, and the correlation of emotional states is inferred. The dependency between vectors is calculated through the correlation coefficient to obtain the final multidimensional feature vectors.

[0009] Furthermore, the physiological state assessment module obtains the raw input from the physiological signal data and generates the multidimensional feature vector through filtering and normalization preprocessing; The multidimensional feature vectors are input into a support vector machine classification algorithm to obtain attention concentration and cognitive load levels. The multidimensional feature vector is used as input to a linear regression algorithm to determine the emotional valence and arousal index; Based on the indicators of attention concentration, cognitive load level, emotional valence and arousal, a physiological abnormality warning is generated by weighted averaging and fusion, and a physiological state assessment index is obtained.

[0010] Furthermore, the simulation content optimization module collects physiological signals such as the user's heart rate and skin conductance through monitoring equipment, obtains the quantitative values ​​of the physiological state indicators, and determines the current user's fatigue level; Based on the quantitative values ​​of the physiological state indicators, a preset set of control strategy rules is matched. If the fatigue level exceeds the threshold, a strategy to reduce complexity is selected first to obtain a set of adjustment parameters. The task complexity level and interference event triggering frequency are updated using the aforementioned adjustment parameter set to construct a preliminary simulation scenario framework and determine the basic structure of the scenario. The virtual character interaction tone module is integrated into the preliminary simulation scene framework. The tone softness parameter is obtained through a preset tone mapping table. If the frequency of interference events is high, the tone is softened synchronously to determine the interaction optimization scheme. The interactive optimization scheme integrates physiological fatigue threshold feedback data obtained through real-time monitoring to drive iterative adjustments of the simulation content, resulting in optimized simulation content.

[0011] Furthermore, it also includes: The collaboration and self-regulation optimization module compares the physiological state assessment indicators in the optimized simulation content with the preset optimization range to determine the improvement of learners' self-regulation ability; in a multi-person collaborative simulation education scenario, it obtains the self-regulation skill indicators of multiple collaborators and calculates the coherence of physiological indicators. The personalized path and intervention module optimizes the collaborative learning map by linking and storing physiological data and learning outcomes, and then building a predictive model based on the linked data.

[0012] Furthermore, the simulated personalized path and intervention module integrates the respiratory regulation changes corresponding to the exceeded indicators through a biofeedback training task to obtain real-time display data; The training task content is dynamically adjusted based on the real-time displayed data to determine the feedback effect of the breathing regulation changes; The self-regulation skill index is calculated based on the feedback effect, providing a basis for subsequent physiological state optimization.

[0013] Furthermore, the collaboration and self-regulation optimization module obtains the self-regulation skill indicators from multiple collaborators and determines the quantitative basis of team tacit understanding by comparing them with physiological indicators. By integrating the team's tacit understanding into the quantitative basis of breathing regulation changes, and using correlation analysis to calculate the emotional synchronization assessment value, the basis for adjusting the task structure is obtained. Based on the task structure adjustment and real-time display data integration, if the emotion synchronization assessment value exceeds the preset optimization range, roles are dynamically assigned and a collaborative optimization graph framework is determined. By integrating feedback effect calculations through the aforementioned collaborative optimization graph framework, skill index calculation results are obtained, and the relevance improvement of collaborator data acquisition is determined. By comparing and optimizing the coherence enhancement interval, a graph containing dynamic role allocation is generated, resulting in a collaborative optimization graph.

[0014] Furthermore, the personalized path and intervention module obtains physiological data and learning outcomes from the collaborative optimization map, and determines the fusion relationship through data integration and analysis; For the aforementioned fusion relationship, physiological results are fused and stored using multi-source data synchronization. Through the synchronization of the multi-source data, a prediction model is established, trained, and optimized. The prediction model adopts the random forest algorithm, which obtains the prediction result by constructing multiple decision trees and voting.

[0015] Furthermore, the personalized path and intervention module obtains the output of the prediction model, generates a path plan and dynamically adjusts the personalized path, obtains an intervention plan based on the personalized path, and integrates educational data feedback to form a forward-looking educational intervention plan.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention uses non-invasive devices to collect learners' electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, electrodermal activity, and eye-tracking data in real time. It then precisely aligns and fuses these physiological signals with the timecode of the simulation environment and the learners' interactive behaviors, ensuring that the physiological signals closely correspond to the events and operations in the virtual environment, thus solving the problem of physiological signals being out of sync with the pace of the course. The acquired multimodal synchronous data stream is preprocessed to eliminate noise and interference, thereby improving the quality and reliability of the data. Furthermore, representative and diagnostic feature vectors, such as EEG power spectrum, heart rate variability, skin conductance response amplitude, and gaze path feature vector, are extracted from the preprocessed data. This solves the problem mentioned in the background technology that physiological signals are multimodal, continuous, and highly time-varying, and that various signals are interconnected and change rapidly over time. By inputting the extracted multidimensional feature vectors into trained classification and regression algorithms, the feature vectors are mapped to learners' physiological state assessment indicators through algorithm calculation and analysis. The machine learning-based assessment method can reflect learners' internal physiological state in real time and accurately, solving the problem that existing simulation education systems mainly rely on learners' external operational behaviors and task completion to judge their learning status, making it difficult to accurately capture learners' internal real-time physiological changes. Based on physiological state assessment indicators, a pre-set set of control strategy rules is applied to drive the simulation engine to adjust the simulation education content in real time; the interactive tone of the virtual character is adjusted according to the emotional state, etc. The dynamic adjustment mechanism can optimize the course content in a timely manner according to the learner's real-time physiological state, which solves the problem of the course content adjustment often being lagging or inaccurate mentioned in the background technology, and improves the personalization and adaptability of simulation education. Attached Figure Description

[0017] Figure 1 This is a framework diagram of a curriculum behavior simulation education system that integrates physiological signal perception according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0019] like Figure 1 As shown in the figure, this embodiment of a curriculum behavior simulation education system integrating physiological signal perception may specifically include: The physiological signal acquisition and synchronization module collects learners' electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, skin conductance activity, and eye-tracking data through non-invasive devices, and aligns and fuses them with the simulation environment's timecode and interactive behavior to obtain a multimodal synchronized data stream. The learner's electroencephalogram (EEG), electrocardiogram (ECG), electroskin activity, and eye-tracking data were collected using non-invasive devices to obtain the raw multimodal signal set. The original multimodal signal group is filtered, and Gaussian filtering is used to remove noise to obtain a clean signal group. For the cleaning signal group, if an abnormal value is detected and exceeds a preset threshold, the corresponding part is removed to obtain a valid signal group. Timecode and interaction behavior data are obtained from the simulation environment, and the synchronization signal sequence is obtained by matching and aligning the effective signal group with the timecode. The synchronization signal sequence fusion interaction data is integrated using a weighted average method to obtain a multimodal synchronization data stream; In one implementation, a non-invasive device is used to collect physiological signals from the learner; Specifically, EEG signals are captured by wearing an electrode cap to capture electrical activity in the cerebral cortex, reflecting the learner's level of attention and cognitive load; The electrocardiogram signal is monitored using a chest strap sensor to detect changes in heart rate and indicate emotional arousal levels; Electrodermal activity sensors are placed on the fingers or wrist to measure changes in skin resistance to assess stress or arousal. Eye-tracking devices, such as infrared cameras, track pupil movement and fixation points to reveal the distribution of learners' visual attention to simulated content. These devices are all designed to be wirelessly connected, ensuring that learners can move freely in the simulation environment without interference; Furthermore, the data acquisition process needs to be aligned with the timecode of the simulation environment; For example, simulation environments such as virtual classroom simulation software generate real-time timestamps to record each interaction event, such as the time of clicking a button or switching scenes; The frequency for acquiring EEG signals is typically 250Hz, ECG and electrodermal activity is 100Hz, and eye tracking is 60Hz. To achieve synchronization, the time of all devices is first calibrated using a unified clock source such as the NTP protocol, and then environmental time codes are embedded in the data stream to ensure that physiological signals and interactive behaviors are aligned at the millisecond level. This alignment avoids analysis errors caused by data offset; In one possible implementation, multimodal data fusion employs a time-series alignment method; Specifically, each signal is resampled to a common frequency, such as 50Hz, and then the dynamic time warping algorithm is used to match the signal sequence with the interaction behavior sequence. This algorithm finds the path with the minimum distortion by calculating the distance matrix between signals, thus achieving nonlinear alignment. For example, in a simulation task where learners solve mathematical problems, when eye-tracking data shows fixation on the formula area, it aligns with the increase in theta waves of the EEG signal, and after fusion, forms a synchronous stream representing cognitive effort. Preferably, the fusion process also includes feature-level integration; It should be noted that alpha and beta wave power spectra can be extracted from EEG signals as attention features, heart rate variability can be calculated from ECG signals as an emotion indicator, peak response of skin conductance can be obtained as arousal level, and fixation duration and jump frequency can be statistically analyzed using eye tracking. These feature vectors are superimposed on the time axis to form a multidimensional data stream; In a virtual language learning environment, when learners interact with the audio module, the fused data stream can capture the synchronization between electrocardiogram acceleration and eye movement fixation on words, revealing points of learning interest. For example, in a reading comprehension simulation scenario, when the collected EEG signals show that delta waves dominate, it indicates a state of fatigue. At this time, it is aligned with the environmental timecode and integrated with interactive behaviors such as page dwell time to generate a synchronous data stream for adjusting the content difficulty. This integration enhances the integrity of the data stream, enabling comprehensive monitoring of the learning process; In one embodiment, the challenge of alignment fusion lies in handling signal noise and delay; Specifically, EEG signals are susceptible to interference from muscle artifacts, and noise components can be separated through independent component analysis. For electrocardiogram signals, respiratory artifacts need to be filtered out, and a bandpass filter is used to retain the 0.5-40Hz frequency band; Electrodermal activity data is smoothed to remove sudden spikes, and eye-tracking is used to correct for frames lost during blinks. After these preprocessing steps, the simulation interaction behaviors, such as mouse click events, are aligned and merged into a stream. This method is applicable to various learning tasks such as programming simulations. When learners debug code, the fused data captures the synchronization between peak skin conductance and eye movement fixation errors, indicating frustration and thus supporting the adaptive learning system to adjust prompts. Furthermore, the obtained synchronous data stream can be stored in a time-series database format; For example, in scientific experiment simulations, the data stream integrates EEG beta waves, ECG RR intervals, skin conductivity, and eye-tracking thermograms, and fuses them with environmental timecodes to generate a unified stream for subsequent analysis; This multimodal synchronization improves the accuracy of learning assessment; Understandably, the choice of non-invasive devices can be flexibly adjusted according to the scenario; For example, in group learning simulations, multiple learners wear devices and merge data streams to reflect the dynamics of collective attention, but this is still limited to the field of education; In one embodiment, the fusion algorithm may employ machine learning-assisted alignment; Specifically, an LSTM network is used to learn the temporal dependencies between signals, taking a multimodal sequence as input and outputting a synchronization stream. In historical document reading tasks, the network processes EEG and eye-tracking data, integrates interactions such as page-turning time, and generates streams to identify comprehension difficulties; Preferably, this technical solution ensures data privacy during implementation by anonymizing physiological signals and retaining only the fusion stream used for learning and analysis. The data preprocessing and feature extraction module uses a preprocessing algorithm to eliminate noise in the multimodal synchronous data stream and extracts EEG power spectrum, heart rate variability, skin conductance response amplitude and gaze path feature vectors to obtain multidimensional feature vectors. A multimodal synchronous data stream is acquired by an acquisition device, and a low-pass filtering method is used to eliminate high-frequency noise in the data stream to obtain a clean signal sequence. The electroencephalogram power spectrum is extracted from the clean signal sequence, and the signal frequency domain representation is converted by fast Fourier transform. The heart rate variability index is calculated, and the standard deviation of the interval between adjacent heartbeats is obtained by time domain analysis to determine the physiological response baseline. The amplitude of skin conductance response is measured against the physiological response baseline to obtain the peak change of skin conductance level, a gaze path vector is constructed, and the eye movement trajectory coordinate sequence is tracked to obtain a preliminary feature set. The preliminary feature set is fused with real-time stream management attributes extracted from the data stream. The real-time stream management attributes are obtained by synchronous timestamp annotation. A multi-dimensional vector is integrated. If the vector dimension exceeds a preset threshold, the fusion weight is adjusted. The adjustment of the fusion weight is achieved by weighted average operation to obtain an optimized feature vector. Attention pattern analysis is performed based on the optimized feature vectors. Principal component analysis is used to reduce the dimensionality of the vector components, and the correlation of emotional states is inferred. The dependency between vectors is calculated through the correlation coefficient to obtain the final multidimensional feature vectors. In one implementation, the multimodal synchronous data stream is first preprocessed to eliminate noise; Specifically, the multimodal data stream includes electroencephalogram (EEG) signals, heart rate signals, skin conductance signals, and eye-tracking data, which are acquired through synchronous acquisition devices; Preprocessing algorithms can use wavelet transform or low-pass filters to filter out high-frequency noise; For example, wavelet denoising can be used to decompose the signal into different frequency bands and remove irrelevant components, thereby preserving effective physiological information. This method ensures the quality of the data stream and improves the accuracy of subsequent feature extraction; In practical applications, such as attention monitoring scenarios, this preprocessing step helps to handle noise caused by environmental interference. Furthermore, the power spectrum features of the electroencephalogram (EEG) were extracted; The power spectrum of electroencephalography (EEG) reflects the energy distribution of EEG signals at different frequencies. The principle is to convert the time-domain signal into the frequency domain through fast Fourier transform and calculate the power values ​​of each frequency band, such as alpha waves (8-12Hz) and beta waves (12-30Hz). The specific process includes windowing the preprocessed EEG signal, applying a Hamming window function to each window to reduce spectral leakage, and then calculating the power spectral density. In this way, a vector representing the intensity of brain activity is obtained; for example, in cognitive tasks, the power spectrum can indicate the degree of attention concentration. The key to this extraction step is to select appropriate window length and frequency band division to adapt to the physiological differences of different users, thereby forming reliable feature components; In one possible implementation, delta-wave and theta-wave power can be further integrated to expand the characteristic representation; It should be noted that the extraction of heart rate variability features focuses on the fluctuation analysis of heart rate intervals; Heart rate variability represents the level of activity of the autonomic nervous system. The calculation process first detects the RR interval in the heart rate signal, that is, the time difference between consecutive heartbeat peaks, and then calculates time-domain indicators such as standard deviation (SDNN) and frequency-domain indicators such as low-frequency power (LF). For example, in a synchronous data stream, a peak detection algorithm is used to locate the R peak, which is then used to generate an interval sequence and apply power spectrum estimation. This feature helps assess stress or mood changes and, when combined with EEG in multimodal analysis, enhances overall insight. Preferably, a high-frequency power (HF) ratio can be introduced as a supplement to reflect the influence of the parasympathetic nervous system; In one embodiment, the extraction of the skin conductance response amplitude is relatively straightforward; The amplitude of skin conductance response measures the peak value of changes in skin conductance, and its principle is based on the increased sweat gland activity caused by sympathetic nerve activation; Specifically, the peaks and apexes are identified from the preprocessed skin conductance signals, and the amplitude difference is calculated as a feature. Significant responses are typically detected using a threshold method, where an amplitude exceeding 0.05 μS is considered a valid event. This extraction method is useful in emotion response monitoring, as it can capture instantaneous arousal changes; By accumulating multiple response amplitudes, a vector representation is formed, which is then further fused with other modalities; The advantage of this process is its high real-time performance, making it suitable for continuous monitoring scenarios. For example, the extraction of gaze path feature vectors involves the quantization of eye movement trajectories; A fixation path refers to a sequence of eye movements, including fixation point, saccades, and jumps. The extraction process first segments the path from the eye-tracking data, and then calculates features such as path length, average fixation duration, and direction change rate. Specifically, clustering algorithms are used to group fixations and generate path vectors. For example, in reading tasks, path features can reveal attention allocation patterns. This method enhances the behavioral dimension of multimodal data by dynamically encoding paths using vectors. In practical implementation, path entropy can be added as a complexity indicator to enhance feature robustness. Furthermore, the extracted features are integrated into a multi-dimensional feature vector; The specific process includes standardizing each feature value to unify the scale, and then concatenating them into a single vector, such as concatenating the EEG power spectrum vector with the heart rate variability vector. This multidimensional vector can have tens of dimensions, representing a comprehensive physiological state; In one implementation, principal component analysis can be applied to reduce dimensionality and retain major variations, but the core is to ensure vector integrity to support downstream tasks such as classification. The technical advantage of this integration step is that it enables multimodal collaboration, improves analysis accuracy, and does not introduce additional noise. In another embodiment, the preprocessing parameters are adjusted for different acquisition devices; For example, when using a portable EEG cap, the noise level is high, so the filtering intensity can be increased, while it is simplified when using fixed laboratory equipment; This variant demonstrates the flexibility of the protocol, making it suitable for monitoring in clinical or educational settings; Preferably, the entire process is implemented in a real-time system, accelerating feature extraction through parallel computing and ensuring low-latency output of multi-dimensional vectors; This versatility supports a variety of synchronous data streaming applications, and is not limited to specific hardware; The physiological state assessment module inputs the multidimensional feature vector into the trained classification and regression algorithms, mapping it to attention concentration, cognitive load level, and emotional valence arousal, thus obtaining physiological state assessment indicators. The raw input is obtained from physiological signal data, and the multidimensional feature vector is generated through filtering and normalization preprocessing. The multidimensional feature vectors are input into a support vector machine classification algorithm to obtain attention concentration and cognitive load levels. The multidimensional feature vector is used as input to a linear regression algorithm to determine the emotional valence and arousal index; Based on the indicators of attention concentration, cognitive load level, emotional valence and arousal, a physiological abnormality warning is generated by weighted average fusion, and a physiological state assessment index is obtained. In one implementation, the process of inputting the multidimensional feature vector into the trained classification and regression algorithms first requires understanding the composition of the multidimensional feature vector. These vectors are typically extracted from physiological signals such as brain waves, heart rate, or skin conductance. Their dimensions may include time-domain features such as average value, frequency-domain features such as power spectral density, and statistical features such as variance, thus forming a vector that comprehensively represents the user's physiological state. For example, in health monitoring scenarios, this vector can capture physiological changes in users during task execution, providing basic data for subsequent mapping; In this way, the comprehensiveness of the input data is ensured, supporting the accurate processing of the algorithm; Specifically, the classification algorithm is used here to handle the classification of discrete physiological states, such as mapping attention concentration into three levels: high, medium, and low. Classification algorithms can employ methods such as support vector machines or random forests. These algorithms learn patterns in the training data and project the input vectors into a predefined class space. The training process involves collecting labeled datasets of physiological signals, such as data collected from volunteers during simulated driving tasks, where attention concentration is labeled by experts; The algorithm optimizes the decision boundary during training to minimize classification error, thereby achieving effective classification of new input vectors; This mapping mechanism helps to quickly identify the user's attention level in real-time monitoring, avoiding complex calculations; Furthermore, regression algorithms are used for predicting continuous values, such as cognitive load levels and emotional valence arousal. Regression algorithms, such as linear regression or neural network regression models, map input features to output indicators by fitting a functional relationship between them. For example, in cognitive load assessment, the model learns the correspondence between feature vectors and load scores. The load level can be represented as a continuous value from 0 to 1, where a high value corresponds to a high load state. During training, optimization methods such as gradient descent are used to adjust model parameters to ensure that the error between the predicted values ​​and the actual physiological indicators is minimized. In one possible implementation, the outputs of classification and regression are combined to form a comprehensive physiological state assessment index. For example, attention concentration is used as the classification result, and cognitive load and emotional valence arousal are used as the regression results. After fusion, a multi-dimensional index vector is generated. Preferably, in the implementation scenario of human-computer interaction, this process can be applied to virtual reality training systems; For example, when a user wears a physiological sensor device, the system extracts multi-dimensional feature vectors in real time and inputs them into an algorithm to map attention concentration, which is then used to adjust the training difficulty. If attention span is low, the system can provide additional prompts to improve user engagement; This application demonstrates the versatility of the technology, while extending to different tasks within the same domain, such as learning assistance systems, where cognitive load level mapping is used to detect user fatigue, and emotional valence arousal assesses positive or negative emotions, thereby optimizing the interactive experience. It should be noted that the mapping of emotional valence arousal involves a two-dimensional model of valence (positive to negative) and arousal (low to high); The regression algorithm learns the patterns of feature vectors, such as high-frequency EEG components corresponding to high arousal, and calculates specific values. In one embodiment, the training data comes from an emotion-inducing experiment, in which physiological responses are recorded when users watch videos, and the algorithm establishes a mapping function based on these responses. This detailed mapping process ensures the reliability of the assessment indicators and provides accurate feedback in physiological state monitoring; For example, in fatigue detection scenarios, after inputting multidimensional feature vectors, the classification algorithm first determines whether the attention concentration is below the threshold. If so, the regression algorithm further quantifies the cognitive load level to avoid the limitations of a single algorithm. This combination yields comprehensive physiological state assessment indicators, supporting subsequent decisions such as alarm triggering; In another implementation, the algorithm can be trained using cross-validation to ensure the model's generalization ability on different user data; For example, the dataset is divided into a training set and a test set, and after iterative optimization, it is applied to actual evaluation; This strategy enhances the flexibility of the technology and ensures consistent application across the field of physiological health; Furthermore, the technical advantage of this mapping process lies in its ability to achieve a quantitative description of physiological states, such as in medical assistance, helping doctors monitor changes in patients' attention without introducing subjective judgment. By providing objective algorithmic outputs, the accuracy and efficiency of the evaluation are improved. In one embodiment, a comprehensive score can be generated by combining indicators of attention concentration, cognitive load level, and emotional valence arousal, which can be used for overall physiological state reporting. This implementation covers the core features of the claims, ensuring applicability in a variety of monitoring scenarios; The simulation content optimization module applies a set of control strategies based on the physiological state assessment indicators to drive the simulation engine to adjust the task complexity, interference event triggering, and virtual character interaction tone, thereby obtaining optimized simulation content. By collecting physiological signals such as the user's heart rate and skin conductance through monitoring equipment, the quantitative values ​​of the physiological state indicators are obtained to determine the current user's fatigue level. Based on the quantitative values ​​of the physiological state indicators, a preset set of control strategy rules is matched. If the fatigue level exceeds the threshold, a strategy to reduce complexity is selected first to obtain a set of adjustment parameters. The task complexity level and interference event triggering frequency are updated using the aforementioned adjustment parameter set to construct a preliminary simulation scenario framework and determine the basic structure of the scenario. The virtual character interaction tone module is integrated into the preliminary simulation scene framework. The tone softness parameter is obtained through a preset tone mapping table. If the frequency of interference events is high, the tone is softened synchronously to determine the interaction optimization scheme. The interactive optimization scheme integrates physiological fatigue threshold feedback data obtained through real-time monitoring to drive iterative adjustments of the simulation content, resulting in optimized simulation content. In one implementation, physiological state assessment indicators are obtained through real-time monitoring equipment; Specifically, users wear sensors, such as heart rate monitors and skin conductance meters, in virtual simulation training environments, which collect physiological signal data. Subsequently, signal processing algorithms are used to filter and extract features from the data to form evaluation indicators, such as heart rate variability index and stress level score; These indicators quantify the user's physiological state, such as fatigue or stress levels, providing a basis for subsequent regulation; In this way, the evaluation process is ensured to accurately reflect the user's real-time situation and support the dynamic optimization of the simulation content; Furthermore, the set of regulatory strategy rules is formulated based on physiological state assessment indicators; This rule set includes preset thresholds and corresponding strategies; For example, when the heart rate variability index is below a certain threshold, it indicates that the user is in a state of high stress, and the rule set activates strategies to reduce task complexity. It should be noted that the rule set adopts an if-then structure design; For example, if the stress level score exceeds a moderate threshold, the reduction of disruptive events is preferentially triggered; This set of rules is designed based on principles of physiological psychology to ensure that adjustments conform to the body's adaptive mechanisms, thereby maintaining the user's optimal engagement during virtual training. For example, in a flight simulation training scenario, the simulation engine receives instructions from the set of control strategy rules and adjusts the task complexity. The specific process includes the engine parsing the rule output and dynamically modifying the simulation parameters, such as switching from a simple flight path to a navigation task under complex weather conditions; Through the engine's modular architecture, the complexity adjustment module calculates the current task load and gradually increases or decreases the difficulty value according to rules. For example, the initial complexity is set to the basic level, and as physiological indicators improve, the engine gradually introduces more variables. This adjustment mechanism ensures a gradual progression of the training process, avoiding user overload; In one possible implementation, the interference event is triggered by the simulation engine according to a set of rules; For example, when evaluation metrics show that the user's attention is distracted, the engine triggers random events such as sudden alarms or system malfunctions; The triggering logic of these events is based on a probability model and linked to physiological state to ensure that the frequency of events is moderate; Preferably, after triggering, the engine monitors the user's response and further feeds it back into the rule set to form a closed-loop control, thereby simulating real interference during training and improving the user's coping ability; Understandably, adjusting the interactive tone of virtual characters is a key aspect of optimizing simulation content; The simulation engine drives virtual characters, such as instructors, to change tone parameters based on physiological states, for example, from stern to encouraging. Specifically, if metrics indicate user fatigue, the engine switches to milder expressions from its tone library, such as using "Stay calm and keep trying" instead of "Correct immediately"; This adjustment is achieved through the natural language processing module, ensuring a more human-like interaction and supporting users to maintain motivation in the virtual environment; In one embodiment, the above process is integrated into medical emergency simulation training; The physiological state assessment indicators first collect the user's pulse and respiration data to form an indicator vector; The regulation strategy rule set applies threshold judgment, such as activating the simplified task strategy when the respiratory rate is too high; The simulation engine adjusts the complexity accordingly, reducing complex surgical steps, lowering the probability of triggering disruptive events such as sudden patient complications, and enabling the virtual mentor character to adopt a supportive tone. Through these steps, the optimized simulation content is more closely aligned with the user's state, thus improving training efficiency. Furthermore, in the driving simulation training embodiment, the rule set prioritizes safety thresholds; For example, when the heart rate spikes, the engine immediately reduces the complexity of the task, such as switching from highway driving to low-speed city driving, and suppresses interfering events such as vehicle malfunctions. Virtual characters, such as navigation voices, will switch to calm guidance; This diverse application demonstrates the versatility of the technical solution within the same field, ensuring effective optimization across different training sub-scenarios. It should be noted that the logical sequence of the entire process begins with data acquisition and ends with the output of optimized content; The modules are connected through data flow; for example, evaluation metrics are directly input into the rule set, and the rule output drives the engine to adjust. This consistency ensures the stability of the simulation system, and in actual deployment, it can be expanded to more physiological sensors, improving overall adaptability; When the physiological state assessment index in the optimized simulation content exceeds the preset optimization range, the personalized path and intervention module integrates the changes in respiratory regulation through biofeedback training tasks and displays them in real time to obtain self-regulation skill indicators. Specifically, the physiological state evaluation index in the optimized simulation content is obtained, and it is determined whether the index exceeds the preset optimization interval by comparing it with the preset optimization interval. Through biofeedback training tasks, the respiratory regulation changes corresponding to the exceeded indicators are integrated to obtain real-time display data; The training task content is dynamically adjusted based on the real-time displayed data to determine the feedback effect of the breathing regulation changes; The self-regulation skill index is calculated based on the feedback effect to obtain the basis for subsequent physiological state optimization. In one implementation, the system first performs a physiological state assessment on the optimized simulation content; Specifically, optimized simulation content refers to scenarios generated through virtual environment simulation. For example, in the field of psychological relaxation training, simulation content may include calm natural landscape simulations that have been adjusted to adapt to the user's initial physiological data. Physiological state assessment indicators include parameters such as heart rate, skin conductivity, or electroencephalogram (EEG), which are collected and calculated in real time by sensors. The preset optimization range is a threshold range set based on the user's historical data or standard physiological range, such as a heart rate range of 60 to 80 beats per minute. If the evaluation indicators exceed this range, it indicates that the user's physiological state has not achieved the expected optimization effect and further intervention is required. Furthermore, when the physiological state assessment index exceeds the preset optimization range, the system initiates a biofeedback training task; This task integrates a real-time display mechanism for changes in respiratory regulation; For example, a biofeedback training task can be designed as a task module that guides users to perform deep breathing exercises, wherein changes in breathing regulation are displayed in real time through dynamic graphics on the screen, such as waveforms showing curves of changes in breathing depth and frequency. The integration process involves synchronizing the user's real-time breathing data with the simulation content. For example, if the user breathes too fast, the system will adjust elements in the simulation scene, such as wind speed simulation, to prompt adjustment, thereby helping the user to perceive and correct their physiological state. It should be noted that the self-regulation skill index is generated by analyzing users' performance in biofeedback training tasks; Specifically, this metric assesses a user’s ability to respond to changes in breathing regulation, such as calculating the adjustment time and success rate of a user from exceeding the range to returning to the normal range. In one possible implementation, the system records the user's breathing data from multiple training sessions and generates a quantitative indicator, such as a regulation efficiency score, which is calculated based on the magnitude and stability of regulation changes, thereby reflecting the user's level of self-regulation skills. Preferably, in a psychological stress management scenario, this process can be applied to routine training procedures; For example, when a user experiences stress simulation in a virtual reality environment, if physiological indicators such as heart rate exceed the optimized range of 70 to 90, the system immediately switches to biofeedback mode, displays breathing regulation guidance, such as prompts like "inhale for 5 seconds, exhale for 5 seconds," and displays animated feedback of lung expansion in real time. Through multiple iterations, users gradually improved their self-regulation skills, from initial low scores to stable values. Understandably, the versatility of this technical solution is reflected under different training intensities; For example, in low-intensity relaxation training, the preset optimization range is more lenient, while in high-intensity stress training, the range is more stringent. Biofeedback tasks adjust the display method according to the scenario, such as using color changes to indicate the effect of breathing regulation, with red indicating exceeding the range and green indicating recovery, thereby enhancing the user's intuitive perception; In one embodiment, the system further integrates multimodal data to optimize the process; Specifically, in addition to breathing regulation, it can also be combined with the display of posture adjustment changes, such as capturing the user's body position through a camera and feeding it back to the screen in real time; If the physiological state assessment indicators continue to exceed the range, the task will gradually increase in difficulty, such as extending the breathing regulation cycle, in order to train more advanced self-regulation skills. The resulting skill metrics can be used to generate reports and record user progress curves; For example, in group training scenarios, this method can be extended to synchronous feedback for multiple people; Each participant's physiological indicators are evaluated independently. If someone exceeds the optimization range, the entire simulation is adjusted to a collective biofeedback task, integrating everyone's breathing changes and displaying them in real time on a shared interface, thereby cultivating the team's self-regulation skills. Furthermore, this implementation method ensures the continuity of the process; Through the aforementioned cycle of assessment and feedback, users can gradually master regulation skills in a simulation environment, and the improvement in skill indicators reflects the improvement in physiological state. In one embodiment, the system may use cloud computing to process data; Specifically, the optimized simulation content is uploaded to the cloud server for batch calculation of physiological state assessment indicators. If the range is exceeded, the cloud pushes a biofeedback task to the user device to remotely display changes in respiratory regulation in real time, thereby obtaining self-regulation skill indicators under distributed training. The collaboration and self-regulation optimization module obtains the self-regulation skill indicators of multiple collaborators, calculates the coherence of physiological indicators, quantifies team tacit understanding, adjusts the task structure, and obtains a collaboration optimization map. Specifically, the self-regulation skill indicators are obtained from multiple collaborators and compared with physiological indicators to determine the quantitative basis of team tacit understanding; By integrating the team's tacit understanding into the quantitative basis of breathing regulation changes, and using correlation analysis to calculate the emotional synchronization assessment value, the basis for adjusting the task structure is obtained. Based on the task structure adjustment and real-time display data integration, if the emotion synchronization assessment value exceeds the preset optimization range, roles are dynamically assigned and a collaborative optimization graph framework is determined. By integrating feedback effect calculations through the aforementioned collaborative optimization graph framework, skill index calculation results are obtained, and the relevance improvement of collaborator data acquisition is determined. By judging the optimization interval through the coherence enhancement comparison, a graph containing dynamic role allocation is generated, and a collaborative optimization graph is obtained. In one implementation, the system first acquires the self-regulation skill indicators of multiple collaborators; These metrics are derived from the quantitative scores generated in the aforementioned biofeedback training task, such as the regulation efficiency score of each collaborator under the real-time display of changes in respiratory regulation. It should be noted that self-regulation skills indicators reflect an individual's ability to control their physiological state, such as calculated values ​​based on regulation time and success rate. Specifically, the system collects heart rate and skin conductance data from each collaborator through sensors and aggregates these indicators on a cloud server to form a multi-user dataset; This process ensures data synchronization for subsequent team-level analysis; Furthermore, based on the acquired self-regulation skill indicators, the system calculates the coherence of physiological indicators; Physiological coherence refers to the degree of synchronization of physiological signals among multiple collaborators, such as the correlation of heart rate waveforms; For example, during the calculation process, the system uses the correlation coefficient method to pair and compare the heart rate sequences of each collaborator to quantify the similarity of the signals in the time domain; If the heart rate changes of two collaborators are highly consistent, the coherence value is high. This coherence calculation helps identify the overall level of coordination in the team's physiological state; In one possible implementation, the system first normalizes the metrics, then applies the Pearson correlation coefficient formula to calculate the coherence value between each pair of collaborators, and calculates the team average coherence. It should be noted that this calculation process takes into account time windows, such as sliding windows in 10-minute increments, to avoid noise interference and thus obtain reliable coherence indicators. In this way, the system transforms individual skill indicators into a quantitative basis for team physiological coordination; Preferably, the system uses the coherence of physiological indicators to quantify team tacit understanding; Team cohesion is a comprehensive score based on coherence value, reflecting the synergy of physiological responses among collaborators; For example, in a psychological stress management scenario, if the average coherence exceeds 0.8, the level of tacit understanding is rated as high. Specifically, the quantification process involves weighted summation, which combines relevance with individual skill indicators to generate a compatibility score from 0 to 100. This score takes into account weights, such as giving higher weight to the metrics of core collaborators; In one embodiment, the system uses a threshold to determine whether the team's tacit understanding is below 50, indicating insufficient physiological synchronization and requiring intervention, thus providing a basis for adjusting the task structure. Understandably, the system adjusts the task structure based on the quantified team synergy. Task structure refers to the module configuration of a biofeedback training task, such as the difficulty and duration of a breathing regulation task; For example, if the level of teamwork is low, the system adds a collective synchronized breathing module to extend the task cycle to 15 minutes to enhance team coordination; Specifically, the adjustment process includes analyzing the tacit understanding score and mapping it to preset rules, such as switching the task structure from individual mode to group interaction mode when the tacit understanding score is between 30 and 50. This adjustment aims to improve overall physiological coherence; In group training scenarios, this mechanism is applied to virtual reality environments to ensure that tasks dynamically adapt to the team's state. Furthermore, the system generates a collaborative optimization graph through the adjusted task structure; Collaboration optimization graph is a visual representation of a network graph that shows the team's synergy. For example, nodes represent collaborators, and the thickness of the edges represents the strength of coherence. In one embodiment, the system uses a graphics algorithm to draw a map, and colors are used to encode the level of tacit understanding, such as green representing an optimized area; It should be noted that the comparison of data before and after integration and adjustment during the map generation process highlights the optimization effect; For example, after multiple training iterations, the graph shows that the coherence borders are thicker, reflecting an improvement in team synergy. This map can be exported as a report and can be viewed remotely. In another implementation, the system is extended to applications with different training intensities; For example, in low-intensity relaxation training, the threshold for measuring tacit understanding is more lenient and the coherence calculation window is larger, while in high-intensity stress training, the threshold is stricter and the task is adjusted more frequently. Specifically, this universality is reflected in the dynamic updating of the graph, such as refreshing the edge lines in real time to provide feedback on the adjustment results, thereby helping users understand the physiological basis of teamwork. Preferably, the system further integrates the multimodal data optimization process; In addition to physiological indicators, coherence can also be calculated by combining voice synchronization data, such as the correlation of dialogue rhythms between collaborators. In one possible implementation, if the level of understanding remains low, a voice guidance module is added to the task structure to enhance interaction and thus improve the overall connectivity in the graph. This integration ensures the continuity of the process, allowing users to gradually master team coordination skills through the graph; The personalized path and intervention module associates physiological data and learning outcomes with the collaborative optimization map, establishes a predictive model to generate personalized path planning, and obtains a prospective educational intervention plan. Specifically, physiological data and learning outcomes are obtained from collaborative optimization maps, and fusion relationships are determined through data integration and analysis; For the aforementioned fusion relationship, physiological results are fused and stored using multi-source data synchronization. Through the synchronization of the multi-source data, a prediction model is established, trained, and optimized. The prediction model adopts the random forest algorithm, which obtains the prediction result by constructing multiple decision trees and voting. Obtain the output of the prediction model, generate a path plan, and dynamically adjust the personalized path; Based on the personalized path described above, an intervention plan is obtained and educational data feedback is integrated to form a proactive educational intervention plan. In one implementation, collaborative optimization graphs are used to process multi-source data in the field of education; Specifically, this graph is a structured framework based on graph theory, where nodes represent education-related entities, such as individual students, learning tasks, or physiological indicators, and edges represent the relationships between these entities. For example, in the student learning process, the graph can connect physiological data nodes with learning outcome nodes, and reflect the collaborative influence between data through weight optimization; The optimization process of this graph involves iteratively adjusting edge weights to minimize data inconsistency, thereby achieving efficient associative storage. It should be noted that the core of the collaborative optimization graph lies in its dynamic adjustment mechanism, which can adapt to data changes in different educational scenarios. For example, in online learning platforms, it can update the correlation between student attention indicators and grade records in real time to ensure the accuracy and applicability of the graph. In this way, the map provides a reliable data foundation for subsequent predictive models and plays a key role in educational interventions; Furthermore, the associated storage of physiological data and learning outcomes is a storage strategy achieved through collaborative optimization of the graph.

[0020] In one possible implementation, students' physiological data is first collected, such as heart rate variability and electroencephalogram (EEG) patterns recorded via wearable devices. This data reflects the students' learning status, such as attention span or stress level. Then, this physiological data is correlated with learning outcomes, such as test scores or homework completion rates. Specifically, the process involves converting the physiological data into feature vectors and matching them with quantitative indicators of learning outcomes, storing these vectors in graph nodes.

[0021] For example, in a classroom setting, if a student's heart rate indicates high stress, their learning outcome node will be associated with and stored as a corresponding inefficiency marker. This associative storage not only preserves the original data but also records causal relationships through graph edges, facilitating querying and analysis.

[0022] Preferably, this storage strategy employs a distributed database to support large-scale educational data, ensuring data security and rapid access. In the education field, this approach helps track students' long-term developmental trajectories, providing data support for personalized interventions.

[0023] For example, the process of building a predictive model focuses on generating forward-looking outputs using data stored in relation to the data.

[0024] Specifically, the model is a machine learning-based framework, such as a neural network structure, in which the input layer receives physiological data and learned features from the graph, the hidden layer performs feature fusion and pattern recognition, and the output layer generates prediction results.

[0025] It should be noted that model training involves a supervised learning strategy. First, samples are extracted from historical data, such as the correspondence between past students' physiological indicators and learning paths. Then, parameters are optimized through gradient descent to improve prediction accuracy.

[0026] For example, in primary and secondary school education environments, the model can learn patterns in high-attention physiological data corresponding to efficient learning outcomes, thereby predicting future learning bottlenecks. This modeling process emphasizes data preprocessing steps, such as normalizing physiological indicators, to avoid noise interference. Furthermore, the model framework integrates cross-validation mechanisms to ensure generalization ability across different student groups. Through this detailed training and optimization, the predictive model can provide a reliable basis for educational pathway planning and achieve forward-looking assessment of student potential in practical applications.

[0027] In one embodiment, personalized route planning is generated based on the output of a prediction model.

[0028] Specifically, this planning is a sequential decision-making process. First, it extracts students' potential learning needs from model predictions, such as identifying a student's physiological fatigue pattern in mathematics that corresponds to low performance. Then, it uses path algorithms, such as the graph-based shortestpath method, to generate a sequence of steps from the current state to the target state.

[0029] For example, in a higher education setting, if a model predicts that students will have difficulty concentrating in a programming course, the planned learning path might include adjusting the order of learning modules, introducing basic conceptual modules first to alleviate stress, and then gradually progressing to more complex tasks. This generative process takes into account individual factors, such as students' age and interests, to ensure the feasibility of the path.

[0030] Preferably, path planning also incorporates feedback loops, allowing for dynamic adjustments to the plan based on real-time physiological data, thereby enhancing adaptability. In education, this approach helps optimize resource allocation and achieve efficient learning guidance.

[0031] It is understandable that obtaining a forward-looking educational intervention plan is a comprehensive solution formed by integrating personalized pathway planning.

[0032] In one possible implementation, the scheme includes specific interventions, such as recommending tutoring resources or adjusting the pace of instruction for anticipated learning disabilities.

[0033] For example, if the pathway planning indicates that students experience peak physiological stress during language learning, the intervention plan might suggest incorporating rest intervals or the use of assistive tools. This approach emphasizes proactivity, meaning that plans are developed in advance based on model predictions to prevent problems from occurring.

[0034] Specifically, the solution generation involves a rule engine that maps path planning output to an educational strategy library, selecting matching intervention options. Furthermore, in school management scenarios, this solution can be extended to the group level, such as planning intervention activities for the entire class to ensure collective learning effectiveness. In this way, proactive educational intervention solutions can improve overall effectiveness in educational practice. Further extending this approach, in another implementation, the optimization of the collaborative optimization graph can be aided by a genetic algorithm. The specific process includes initializing the graph edge weights and then iteratively optimizing them through crossover and mutation operations to maximize the accuracy of data association.

[0035] For example, in vocational education and training, optimized graphs can better link trainees' physiological fatigue data with their skill mastery outcomes, thereby improving storage efficiency. This method enhances the robustness of the graphs and makes them suitable for educational environments with significant data fluctuations.

[0036] For example, for variants of the prediction model, an ensemble learning framework, such as a random forest combined with a neural network, can be introduced.

[0037] Specifically, this framework addresses the uncertainty of physiological data through a voting mechanism involving multiple sub-models. For example, in special education scenarios, when predicting the learning paths of autistic students, the ensemble method can integrate multi-dimensional features to improve prediction reliability. This implementation demonstrates the model's flexibility and supports applications in different educational subfields.

[0038] In one embodiment, personalized route planning can be further subdivided into short-term and long-term routes.

[0039] Specifically, the short-term path focuses on immediate intervention, such as daily learning adjustments, while the long-term path plans semester goals.

[0040] For example, in online education platforms, short-term paths, based on physiological data predictions, may recommend specific practice modules to quickly improve results. This differentiation enhances the targeted nature of the planning.

[0041] It should be noted that the extension of associated storage includes time-series data processing. In educational tracking systems, storage strategies can record time-series sequences of physiological data and correlate them with timestamps of learning outcomes. For example, analyzing the impact of students' stress changes over a week on their grades provides richer input to the model. Furthermore, the effectiveness of prospective educational interventions is reflected in their practical deployment.

[0042] For example, after being implemented in pilot schools, the solution was able to reduce student dropout rates through route planning, objectively achieving optimized allocation of educational resources. This description highlights the practical value of the solution in the field of education.

[0043] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A curriculum behavior simulation education system integrating physiological signal perception, characterized in that, include: The physiological signal acquisition and synchronization module uses non-invasive devices to collect learners' EEG signals, ECG signals, skin conductance activity, and eye-tracking data in a simulated educational environment in real time. It then precisely aligns and fuses the collected physiological signals with the timecode of the simulated environment and the learners' interactive behaviors. The data preprocessing and feature extraction module preprocesses the acquired multimodal synchronous data stream to eliminate noise and interference, and extracts representative and diagnostic feature vectors from the preprocessed data. The physiological state assessment module inputs the extracted multidimensional feature vectors into the trained classification and regression algorithms. Through the calculation and analysis of the algorithms, the feature vectors are mapped to the learner's physiological state assessment indicators. The simulation content optimization module, based on physiological state assessment indicators, applies a pre-set set of control strategy rules to drive the simulation engine to adjust the simulation education content in real time according to the control strategy rule set, thereby obtaining optimized simulation content.

2. The curriculum behavior simulation education system integrating physiological signal perception according to claim 1, characterized in that, The physiological signal acquisition and synchronization module acquires learners' electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, skin conductance, and eye-tracking data through non-invasive devices to obtain raw multimodal signal sets; The original multimodal signal group is filtered, and Gaussian filtering is used to remove noise to obtain a clean signal group. For the cleaning signal group, if an abnormal value is detected and exceeds a preset threshold, the corresponding part is removed to obtain a valid signal group. Timecode and interaction behavior data are obtained from the simulation environment, and the synchronization signal sequence is obtained by matching and aligning the effective signal group with the timecode. The synchronization signal sequence fusion interaction data is integrated using a weighted average method to obtain a multimodal synchronization data stream.

3. The curriculum behavior simulation education system integrating physiological signal perception according to claim 1, characterized in that, The data preprocessing and feature extraction module acquires a multimodal synchronous data stream through the acquisition device, and uses a low-pass filtering method to eliminate high-frequency noise in the data stream to obtain a clean signal sequence; The electroencephalogram power spectrum is extracted from the clean signal sequence, and the signal frequency domain representation is converted by fast Fourier transform. The heart rate variability index is calculated, and the standard deviation of the interval between adjacent heartbeats is obtained by time domain analysis to determine the physiological response baseline. The amplitude of skin conductance response is measured against the physiological response baseline to obtain the peak change of skin conductance level, a gaze path vector is constructed, and the eye movement trajectory coordinate sequence is tracked to obtain a preliminary feature set. The preliminary feature set is fused with real-time stream management attributes extracted from the data stream. The real-time stream management attributes are obtained by synchronous timestamp annotation. A multi-dimensional vector is integrated. If the vector dimension exceeds a preset threshold, the fusion weight is adjusted. The adjustment of the fusion weight is achieved by weighted average operation to obtain an optimized feature vector. Attention pattern analysis is performed based on the optimized feature vectors. Principal component analysis is used to reduce the dimensionality of the vector components, and the correlation of emotional states is inferred. The dependency between vectors is calculated through the correlation coefficient to obtain the final multidimensional feature vectors.

4. The curriculum behavior simulation education system integrating physiological signal perception according to claim 1, characterized in that, The physiological state assessment module obtains the raw input from physiological signal data and generates the multidimensional feature vector through filtering and normalization preprocessing. The multidimensional feature vectors are input into a support vector machine classification algorithm to obtain attention concentration and cognitive load levels. The multidimensional feature vector is used as input to a linear regression algorithm to determine the emotional valence and arousal index; Based on the indicators of attention concentration, cognitive load level, emotional valence and arousal, a physiological abnormality warning is generated by weighted averaging and fusion, and a physiological state assessment index is obtained.

5. The curriculum behavior simulation education system integrating physiological signal perception according to claim 1, characterized in that, The simulation content optimization module collects physiological signals such as user heart rate and skin conductance through monitoring equipment, obtains the quantitative values ​​of the physiological state indicators, and determines the current user fatigue level. Based on the quantitative values ​​of the physiological state indicators, a preset set of control strategy rules is matched. If the fatigue level exceeds the threshold, a strategy to reduce complexity is selected first to obtain a set of adjustment parameters. The task complexity level and interference event triggering frequency are updated using the aforementioned adjustment parameter set to construct a preliminary simulation scenario framework and determine the basic structure of the scenario. The virtual character interaction tone module is integrated into the preliminary simulation scene framework. The tone softness parameter is obtained through a preset tone mapping table. If the frequency of interference events is high, the tone is softened synchronously to determine the interaction optimization scheme. The interactive optimization scheme integrates physiological fatigue threshold feedback data obtained through real-time monitoring to drive iterative adjustments of the simulation content, resulting in optimized simulation content.

6. The curriculum behavior simulation education system integrating physiological signal perception according to claim 1, characterized in that, Also includes: The collaboration and self-regulation optimization module compares the physiological state assessment indicators in the optimized simulation content with the preset optimization range to determine the improvement of learners' self-regulation ability; in a multi-person collaborative simulation education scenario, it obtains the self-regulation skill indicators of multiple collaborators and calculates the coherence of physiological indicators. The personalized path and intervention module optimizes the collaborative learning map by linking and storing physiological data and learning outcomes, and then building a predictive model based on the linked data.

7. A curriculum behavior simulation education system integrating physiological signal perception according to claim 6, characterized in that, The personalized path and intervention module integrates the respiratory regulation changes corresponding to the exceeded indicators through biofeedback training tasks and obtains real-time display data. The training task content is dynamically adjusted based on the real-time displayed data to determine the feedback effect of the breathing regulation changes; The self-regulation skill index is calculated based on the feedback effect, providing a basis for subsequent physiological state optimization.

8. The curriculum behavior simulation education system integrating physiological signal perception according to claim 6, characterized in that, The collaboration and self-regulation optimization module obtains the self-regulation skill indicators from multiple collaborators and compares them with physiological indicators to determine the quantitative basis of team tacit understanding. By integrating the team's tacit understanding into the quantitative basis of breathing regulation changes, and using correlation analysis to calculate the emotional synchronization assessment value, the basis for adjusting the task structure is obtained. Based on the task structure adjustment and real-time display data integration, if the emotion synchronization assessment value exceeds the preset optimization range, roles are dynamically assigned and a collaborative optimization graph framework is determined. By integrating feedback effect calculations through the aforementioned collaborative optimization graph framework, skill index calculation results are obtained, and the relevance improvement of collaborator data acquisition is determined. By comparing and optimizing the coherence enhancement interval, a graph containing dynamic role allocation is generated, resulting in a collaborative optimization graph.

9. A curriculum behavior simulation education system integrating physiological signal perception according to claim 1, characterized in that, The personalized path and intervention module obtains physiological data and learning outcomes from the collaborative optimization map, and determines the fusion relationship through data integration and analysis. For the aforementioned fusion relationship, physiological results are fused and stored using multi-source data synchronization. Through the synchronization of the multi-source data, a prediction model is established, trained, and optimized. The prediction model adopts the random forest algorithm, which obtains the prediction result by constructing multiple decision trees and voting.

10. A curriculum behavior simulation education system integrating physiological signal perception according to claim 9, characterized in that, The personalized path and intervention module obtains the output of the prediction model, generates path planning and dynamically adjusts the personalized path, obtains an intervention plan based on the personalized path, and integrates educational data feedback to form a forward-looking educational intervention plan.

Images