A method and system for evaluating mental health of primary and secondary school students based on multi-modal physiological data

Abstract

This invention discloses a method and system for assessing the mental health of primary and secondary school students based on multimodal physiological data, relating to the field of mental health technology. The method includes the following steps: S1, data acquisition, which includes synchronously collecting multimodal physiological data of primary and secondary school students during the completion of a mental health scale test. The multimodal physiological data includes at least electroencephalogram (EEG) signals, heart rate signals, blood oxygen saturation signals, and facial video streams; S2, data synchronization, which includes segmenting the multimodal physiological data using the items in the mental health scale as time nodes, and using the time axis of the EEG signals as a reference. This method and system for assessing the mental health of primary and secondary school students based on multimodal physiological data, by integrating four modalities of data—EEG, heart rate, blood oxygen, and micro-expressions—comprehensively assesses mental health status from multiple dimensions, significantly improving the comprehensiveness and accuracy of the assessment.

Description

Technical Field

[0001] This invention relates to the field of mental health technology, specifically to a method and system for assessing the mental health of primary and secondary school students based on multimodal physiological data. Background Technology

[0002] In the current social context, the mental health of primary and secondary school students is becoming increasingly prominent, rising to the level of a focal point of widespread concern and attention across all sectors of society. Primary and secondary school students are at a critical stage of physical and mental development, facing pressure from multiple aspects such as learning, interpersonal relationships, and family, making them prone to various mental health problems. According to relevant research, the detection rate of mental health problems among primary and secondary school students in my country is on the rise, with issues such as anxiety, depression, and low self-esteem being relatively common. These problems not only affect students' learning and lives but may also have a profound negative impact on their future development.

[0003] Traditional methods for assessing mental health primarily rely on questionnaires and interviews, which are inherently subjective and limited. Questionnaires can be influenced by students' subjective opinions, leading to inaccurate results; interviews require professional psychologists, are costly, and are difficult to implement on a large scale. Therefore, finding a more objective, accurate, and efficient method for assessing mental health is of significant practical importance.

[0004] Multimodal physiological data integrates brainwave signals collected by an electroencephalogram (EEG), heart rate, blood oxygen saturation, and micro-expressions captured by a high-definition camera, reflecting the psychological state of primary and secondary school students from multiple perspectives. EEG data directly reflects brain activity, with different frequency bands closely related to different psychological states. Physiological indicators such as heart rate and blood oxygen saturation reflect the body's stress response, indirectly reflecting psychological pressure. Micro-expressions are temporary external manifestations of inner emotions, revealing subconscious emotional states. Through comprehensive analysis of this multimodal physiological data, the mental health of primary and secondary school students can be assessed more comprehensively and accurately, providing a scientific basis for mental health education and intervention.

[0005] Based on the above, the following problems exist: the methods for fusion between different modal physiological data are not yet perfect, and how to effectively integrate multimodal data and give full play to the advantages of each modality remains an urgent problem to be solved; most existing studies focus on specific mental illnesses or emotional states, and there are relatively few studies on the mental health assessment of primary and secondary school students, a special group, lacking targeted assessment indicators and models; traditional methods require offline analysis and cannot provide real-time feedback on psychological fluctuations during the testing process; and the use of fixed thresholds to assess mental health does not take into account individual differences among primary and secondary school students, such as age and gender. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for assessing the mental health of primary and secondary school students based on multimodal physiological data, so as to overcome the shortcomings of the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for assessing the mental health of primary and secondary school students based on multimodal physiological data, comprising the following steps: S1, data acquisition, which includes synchronously acquiring multimodal physiological data of primary and secondary school students during the completion of a mental health scale test, wherein the multimodal physiological data includes at least EEG signals, heart rate signals, blood oxygen saturation signals, and facial video streams; S2, data synchronization, which includes segmenting the multimodal physiological data using the items in the mental health scale as time nodes, and aligning the physiological data of other modalities in the time dimension using the time axis of the EEG signals as a reference, thereby obtaining time-synchronized multimodal data segments; S3, features The process includes: S4, Fusion and Model Processing, which involves extracting modal features from data segments synchronized at different times, fusing the extracted multimodal features, and inputting them into a multimodal Transformer classification model to obtain preliminary mental health status classification probabilities; S5, Personalized Dynamic Assessment, which involves acquiring individual difference information of the students being assessed, calculating a personalized dynamic assessment threshold based on the individual difference information and the students' historical mental health data, calibrating the preliminary mental health status classification probabilities using the dynamic assessment threshold, and generating a final quantitative assessment result for mental health; and S6, Report Generation, which involves generating a visual assessment report and personalized intervention suggestions based on the final quantitative assessment result for mental health.

[0008] Preferably, the electroencephalogram (EEG) signal is acquired by a portable EEG device, the heart rate signal and blood oxygen saturation signal are acquired by a photoplethysmography (PPG) sensor, and the facial video stream is acquired by a high-definition camera; the acquisition of the multimodal physiological data and the presentation of the psychological health scale test are triggered synchronously in time.

[0009] Preferably, the data synchronization step specifically includes: dividing the continuously collected multimodal physiological data into data segments corresponding to each item according to the start and end times of each item in the mental health scale; using the EEG signal data segment as a reference sequence, calculating the dynamic time warping path between the heart rate signal data segment, the blood oxygen saturation signal data segment, and the micro-expression feature sequence extracted from the facial video stream and the reference sequence; and adjusting the heart rate signal data segment, the blood oxygen saturation signal data segment, and the micro-expression feature sequence on the time axis according to the dynamic time warping path to align them with the EEG signal data segment at specific times.

[0010] Preferably, in the feature fusion and model processing step, the extracted modal features include: frequency domain features and time domain nonlinear features extracted from the EEG signal; heart rate variability time domain features extracted from the heart rate signal; statistical features extracted from the blood oxygen saturation signal; and action units and their intensity features extracted from the facial video stream through facial expression coding system analysis.

[0011] Preferably, the feature fusion adopts a weight-based feature-level fusion method, specifically: principal component analysis is used to determine the weight of each modal feature in the fusion vector, and the weighted modal feature vectors are concatenated to form a unified fusion feature vector.

[0012] Preferably, in the personalized dynamic assessment step, the individual difference information includes at least age and gender; the personalized dynamic assessment threshold... It is obtained by calculation using the following formula:

[0013]

[0014] in, μ and σ are the mean and standard deviation of the target mental health index of the reference group matched with the student's age and gender, respectively, and k is a preset multiplier. This is the rate of change of the current value relative to the historical mean, calculated based on the students' historical mental health data. The value is the ratio of individual data volatility to reference group data volatility, calculated based on the students' historical mental health data; α and β are the trend influence factor and stability influence factor, respectively.

[0015] Preferably, the final quantitative assessment result of mental health is obtained by comparing the initial mental health status classification probability with the dynamic assessment threshold. The comparison yields the following result: if the classification probability is greater than or equal to... If the condition is abnormal, it is determined that there is a corresponding psychological health risk tendency, and a quantitative score is given according to the degree of abnormality; otherwise, it is determined to be normal.

[0016] The present invention provides a method for assessing the mental health of primary and secondary school students based on multimodal physiological data, which has the following beneficial effects:

[0017] 1. By integrating data from four modalities—electroencephalogram (EEG), heart rate, blood oxygen, and micro-expression—psychological health status is comprehensively assessed from multiple dimensions, significantly improving the comprehensiveness and accuracy of the evaluation.

[0018] 2. Dynamic thresholds are set based on the developmental characteristics and individual differences (such as age, gender, and historical data) of primary and secondary school students, making the evaluation results more objective and personalized, and overcoming the problem of rigid fixed threshold standards.

[0019] 3. It realizes a real-time processing flow from data collection, processing, analysis to feedback, enabling dynamic monitoring and early warning of psychological state, and providing the possibility for timely intervention.

[0020] 4. The system supports long-term data tracking and can optimize intervention plans based on assessment history, forming a closed-loop dynamic management of "assessment-intervention-reassessment", which effectively promotes the sustainable development of the mental health of primary and secondary school students.

[0021] A system using the aforementioned method for assessing the mental health of primary and secondary school students based on multimodal physiological data includes: a synchronous data acquisition module configured to synchronously trigger and acquire EEG signals, heart rate signals, blood oxygen saturation signals, and facial video streams during a mental health scale test; a multimodal data processing module configured to segment and time-synchronized align the acquired data, and extract features from each modality; a fusion classification module configured to fuse multimodal features, process them using a multimodal Transformer classification model, and output preliminary classification probabilities; a personalized assessment module configured to calculate dynamic assessment thresholds based on individual differences and historical data, calibrate the preliminary classification probabilities, and generate quantitative assessment results; and an intelligent reporting module configured to visualize the quantitative assessment results and generate structured assessment reports and intervention recommendations.

[0022] Preferably, the synchronous data acquisition module includes: a test terminal for presenting the test content of the mental health scale and sending a synchronous trigger signal; and an electroencephalogram (EEG), a photoelectric sensor, and a camera that are communicatively connected to the test terminal, receive the synchronous trigger signal, and begin data acquisition. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0024] Figure 1 A flowchart of a method for assessing the mental health of primary and secondary school students based on multimodal physiological data;

[0025] Figure 2 This is an architecture diagram of a psychological health assessment system for primary and secondary school students based on multimodal physiological data.

[0026] Figure 3 This is a flowchart of dynamic threshold calculation for a method for assessing the mental health of primary and secondary school students based on multimodal physiological data. Detailed Implementation

[0027] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0028] Please see Figure 1-3 This invention provides a method and system for assessing the mental health of primary and secondary school students based on multimodal physiological data, comprising the following steps: S1, data acquisition, which includes synchronously acquiring multimodal physiological data of primary and secondary school students during the completion of a mental health scale test. The multimodal physiological data includes at least electroencephalogram (EEG) signals, heart rate signals, blood oxygen saturation signals, and facial video streams; S2, data synchronization, which includes segmenting the multimodal physiological data using items in the mental health scale as time nodes, and aligning the remaining modal physiological data in the time dimension using a dynamic time warping algorithm based on the time axis of the EEG signals to obtain time-synchronized multimodal data segments; S3, feature fusion and... The model processing includes extracting modal features from data segments synchronized at different times, fusing the extracted multimodal features, and inputting them into a multimodal Transformer classification model to obtain preliminary mental health status classification probabilities; S4, personalized dynamic assessment, which includes obtaining individual difference information of the students being assessed, calculating personalized dynamic assessment thresholds based on individual difference information and students' historical mental health data, calibrating the preliminary mental health status classification probabilities using the dynamic assessment thresholds, and generating the final quantitative assessment results of mental health; S5, report generation, which includes generating a visual assessment report and personalized intervention suggestions based on the final quantitative assessment results of mental health.

[0029] Based on the above, specifically, preprocessing operations are performed on the collected multimodal physiological datasets. The sampling frequency of the currently collected EEG signals is 250 Hz, and a bandpass filter is used to extract the alpha waves (8-13 Hz) from the EEG signal data. Wave (14-20Hz), The test employs target frequency bands such as delta waves (21-30Hz), delta waves (0.5-4Hz), theta waves (4-8Hz), and gamma waves (30-48Hz) to highlight EEG components closely related to mental health. Simultaneously, a notch filter is superimposed to suppress 50Hz or 60Hz power frequency interference, eliminating interference from specific noises on the EEG signal. Secondly, wavelet transform is used to remove artifacts and transient noise from the EEG signal, such as electrooculogram (EOG) and electromyogram (EMG) artifacts, improving the signal-to-noise ratio and making the EEG signal more clearly and accurately reflect the subject's brain activity. For the collected micro-expression video data, the video dataset is first decomposed frame by frame into static images. A facial feature point detection model based on a convolutional neural network (CNN) is used to detect key facial points, including feature points in the eyebrows, eyes, and mouth. Changes in the position of these key points can reflect changes in the student's facial expressions during the test, thus identifying emotional changes. Simultaneously, the acquired micro-expression images undergo grayscale adjustment, contrast enhancement, and histogram equalization to improve image clarity and recognizability. Image normalization is also required to unify pixel values ​​to a uniform range for subsequent processing and model training. For heart rate and blood oxygen saturation, data anomalies or incompleteness may occur during acquisition due to equipment malfunction or improper equipment usage. Outliers are corrected based on context or preceding / following data points. When missing values ​​exist, the missing rate is assessed; if the missing rate is significant, the record is deleted. Otherwise, interpolation methods such as linear interpolation and spline interpolation are used to fill in missing values ​​in the heart rate and blood oxygen data, ensuring data integrity. Furthermore, the heart rate and blood oxygen data are standardized to a standard normal distribution with a mean of 0 and a standard deviation of 1, eliminating the influence of different units and ranges and ensuring data consistency and comparability.

[0030] Among them, the brainwave signals were collected by a portable EEG device, the heart rate signals and blood oxygen saturation signals were collected by a photoplethysmography (PPG) sensor, and the facial video stream was collected by a high-definition camera; the collection of multimodal physiological data and the presentation of the mental health scale test were triggered synchronously in time.

[0031] Furthermore, based on the items in the mental health scale, the data for each modality is segmented to ensure that the data within each time period corresponds to the specified items. The segmented multimodal data is then synchronized on the time axis to ensure temporal consistency between different modalities. Specifically, Dynamic Time Warping (DTW) is used to synchronize four modalities of data (EEG signals, heart rate, blood oxygen saturation, and micro-expressions) on the time axis. For the four modalities of data, we have: EEG signals include: Use it as a reference signal; heart rate Blood oxygen saturation Micro-expression data Brainwave signals are aligned with heart rate to calculate brainwave signals. and heart rate (blood oxygen saturation) Micro-expression data DTW alignment path (using the same calculation method):

[0032]

[0033] in, By backtracking the cumulative distance matrix, Find the optimal path P from (1,1) to (N,M). Path P consists of a series of points... Composition, that is, representing the alignment relationship between two time series. For each signal , as well as Find its relationship with the reference signal Alignment path , as well as Adjust the time axis of each modality data according to the alignment path to synchronize it with the EEG signal.

[0034] Specifically, for the frequency domain characteristics of EEG signals, the relative power of each frequency band in the EEG signal is calculated, that is, the proportion of the power of each frequency band to the total power, reflecting the intensity of brain activity at different frequencies. Simultaneously, features such as average power and power spectral density of each frequency band are extracted to more comprehensively describe the frequency domain characteristics of EEG. Time-domain statistics such as mean, standard deviation, variance, and peak value of the EEG signal are calculated to reflect the fluctuation of the signal over time; waveform complexity and nonlinear characteristics of the EEG signal, such as sample entropy, are extracted to describe the complexity and regularity of the EEG signal. The extracted frequency domain features and time-domain features of the EEG are merged into a multi-dimensional feature vector, i.e., the EEG signal features. HRV time-domain features, including NN50, pNN50, and RMSSD, are extracted from heart rate data to reflect heart rate fluctuations. Time-domain features, including mean, variance, extrema, and median, are extracted from blood oxygen saturation data. Micro-expression data feature extraction involves analyzing changes in the position of facial feature points and calculating the amplitude of facial muscle movements to extract emotional intensity features, such as the degree of eyebrow raising, eye opening and closing, and the angle of mouth corner raising, reflecting the intensity of emotions. Based on the Facial Expression Coding System (FACS), facial muscle movements are decomposed into different action units (AUs). Each AU corresponds to a specific facial muscle movement and is associated with a specific emotional expression. By detecting the AUs appearing in micro-expressions and their intensity, the type and intensity of emotions can be quantified.

[0035] The data synchronization steps specifically include: dividing the continuously collected multimodal physiological data into data segments corresponding to each item based on the start and end times of each item in the mental health scale; using the EEG signal data segment as a reference sequence, calculating the dynamic time warping path between the heart rate signal data segment, the blood oxygen saturation signal data segment, and the micro-expression feature sequence extracted from the facial video stream and the reference sequence; and adjusting the heart rate signal data segment, the blood oxygen saturation signal data segment, and the micro-expression feature sequence on the time axis according to the dynamic time warping path to align them with the EEG signal data segment at specific times.

[0036] In the feature fusion and model processing steps, the extracted modal features include: frequency domain features and time domain nonlinear features extracted from EEG signals; time domain features of heart rate variability extracted from heart rate signals; statistical features extracted from blood oxygen saturation signals; and action units and their intensity features extracted from facial video streams through facial expression coding system analysis.

[0037] The feature fusion adopts a weight-based feature-level fusion method, which is as follows: principal component analysis is used to determine the weight of each modal feature in the fusion vector, and the weighted modal feature vectors are concatenated to form a unified fusion feature vector.

[0038] Specifically, the extracted features from each modality are fused. Following the steps above, feature vectors for each modality are obtained, their feature dimensions are unified, and then, based on feature concatenation, EEG signal features are concatenated with heart rate, blood oxygen, and micro-expression data features in two modalities. This feature vector is then input into the sub-models, resulting in three sub-models. Based on the prediction results output by each sub-model, a weighted voting method is used to obtain the fused features. Different weights are assigned to the features of different modalities based on their importance or relevance to mental health assessment; these weights are determined using principal component analysis (PCA).

[0039] Furthermore, a multimodal Transformer model is constructed, using the fused multimodal features as input. A linear transformation maps these features to a fixed-dimensional embedding space, forming an input sequence. The Transformer model employs a multi-head self-attention mechanism, enabling it to simultaneously focus on the relationships between different positions in the input sequence. This captures not only local adjacency relationships but also global dependencies between multimodal features. Following the self-attention mechanism, positional encoding is added to preserve the sequence information of the features. Then, a multi-layer feedforward neural network is used to perform a non-linear transformation on the features, further extracting and learning high-level semantic information. Finally, a fully connected layer is added as a classification head to the output layer of the Transformer model, mapping the model's output feature vector to category labels for mental health states, such as normal, anxious, and depressed.

[0040] This approach incorporates dynamic thresholds to assess mental health status, taking into account individual differences. Age and gender are used as basic individual differences for primary and secondary school students. Historical mental health data (such as previous psychological assessment results and intervention records) is also considered as an individual difference feature, taking into account its impact on mental health status. Dynamic thresholds are set based on individual differences such as age, gender, and historical data to identify students' mental health states (prone to anxiety / depression, etc.), and these states are then quantitatively assessed to generate a mental health score. Specifically,

[0041] Basic threshold calculation:

[0042]

[0043] in, This represents the age-gender group mean. represents the standard deviation of the age-gender group, and k represents the standard deviation factor (ranging from 1.5 to 3).

[0044] Furthermore, in the personalized dynamic assessment step, individual difference information includes at least age and gender; personalized dynamic assessment thresholds. It is obtained by calculation using the following formula:

[0045]

[0046] in, μ and σ are the mean and standard deviation of the target mental health index for a reference group matched with students' age and gender, respectively, and k is a preset multiplier. This is the rate of change of the current value relative to the historical mean, calculated based on students' historical mental health data. The value represents the ratio of individual data volatility to reference group data volatility, calculated based on students' historical mental health data; α and β are the trend influence factor and stability influence factor, respectively.

[0047] The final quantitative assessment of mental health results is achieved by combining the initial mental health status classification probability with a dynamic assessment threshold. The comparison yields the following: if the classification probability is greater than or equal to... If the condition is abnormal, it is determined that there is a corresponding psychological health risk tendency, and a quantitative score is given according to the degree of abnormality; otherwise, it is determined to be normal.

[0048] A system for assessing the mental health of primary and secondary school students based on multimodal physiological data is characterized by comprising: a synchronous data acquisition module, configured to simultaneously trigger and acquire EEG signals, heart rate signals, blood oxygen saturation signals, and facial video streams during the mental health scale test; specifically, responsible for collecting multimodal physiological data of primary and secondary school students during the mental health scale test, using USB transmission, collecting EEG signal data through an EEG analyzer, recording micro-expression changes through a high-definition camera, and recording heart rate and blood oxygen saturation data during the acquisition process through a heart rate and blood oxygen monitoring device; a multimodal data processing module, configured to segment and time-synchronized align the acquired data, and extract features of each modality; a fusion classification module, configured to fuse multimodal features and process them using a multimodal Transformer classification model to output preliminary classification probabilities; a personalized assessment module, configured to calculate dynamic assessment thresholds based on individual differences and historical data, calibrate the preliminary classification probabilities, and generate quantitative assessment results; and an intelligent reporting module, configured to visualize the quantitative assessment results and generate structured assessment reports and intervention suggestions.

[0049] The synchronous data acquisition module includes: a test terminal for presenting the test content of the mental health scale and sending synchronous trigger signals; and an electroencephalogram (EEG), photoelectric sensor, and camera that are connected to the test terminal to receive synchronous trigger signals and start data acquisition.

[0050] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for assessing the mental health of primary and secondary school students based on multimodal physiological data, characterized in that, Includes the following steps: S1. Data acquisition, which includes the synchronous acquisition of multimodal physiological data of primary and secondary school students during the process of completing the mental health scale test. The multimodal physiological data includes at least brain wave signals, heart rate signals, blood oxygen saturation signals and facial video streams. S2. Data synchronization includes segmenting the multimodal physiological data using the items in the mental health scale as time nodes, and aligning the physiological data of the other modalities in the time dimension using the time axis of the EEG signal as a reference, thereby obtaining time-synchronized multimodal data segments. S3, Feature Fusion and Model Processing, which includes extracting modal features from the data segments synchronized at different times, fusing the extracted multimodal features, and inputting them into the multimodal Transformer classification model to obtain the preliminary mental health status classification probability; S4. Personalized dynamic assessment, which includes obtaining individual difference information of the students being assessed, calculating a personalized dynamic assessment threshold based on the individual difference information and the students' historical mental health data, calibrating the preliminary mental health status classification probability using the dynamic assessment threshold, and generating a final quantitative assessment result of mental health. S5. Report generation, which includes generating a visual assessment report and personalized intervention recommendations based on the final quantitative assessment results of mental health.

2. The method for assessing the mental health of primary and secondary school students based on multimodal physiological data according to claim 1, characterized in that, The brainwave signals were acquired by a portable EEG device, the heart rate signals and blood oxygen saturation signals were acquired by a photoplethysmography (PPG) sensor, and the facial video stream was acquired by a high-definition camera. The collection of the multimodal physiological data and the presentation of the mental health scale test are triggered synchronously in time.

3. The method for assessing the mental health of primary and secondary school students based on multimodal physiological data according to claim 1, characterized in that, The data synchronization steps specifically include: Based on the start and end times of each item in the mental health scale, the continuously collected multimodal physiological data is segmented into data segments corresponding to each item. Using the EEG signal data segment as a reference sequence, the dynamic time warping path between the heart rate signal data segment, the blood oxygen saturation signal data segment, and the micro-expression feature sequence extracted from the facial video stream and the reference sequence is calculated respectively. Based on the dynamic time warping path, the heart rate signal data segment, the blood oxygen saturation signal data segment, and the micro-expression feature sequence are scaled and adjusted on the time axis to align them with the EEG signal data segment at specific times.

4. The method for assessing the mental health of primary and secondary school students based on multimodal physiological data according to claim 1, characterized in that, In the feature fusion and model processing steps, the extracted modal features include: Frequency domain features and time domain nonlinear features extracted from the electroencephalogram (EEG) signals; Temporal features of heart rate variability extracted from the heart rate signal; Statistical features extracted from the blood oxygen saturation signal; Action units and their intensity features are extracted from the facial video stream using a facial expression coding system.

5. The method for assessing the mental health of primary and secondary school students based on multimodal physiological data according to claim 1, characterized in that, The feature fusion adopts a weight-based feature-level fusion method, which is as follows: principal component analysis is used to determine the weight of each modal feature in the fusion vector, and the weighted modal feature vectors are concatenated to form a unified fusion feature vector.

6. The method for assessing the mental health of primary and secondary school students based on multimodal physiological data according to claim 1, characterized in that, In the personalized dynamic assessment step, the individual difference information includes at least age and gender; the personalized dynamic assessment threshold... It is obtained by calculation using the following formula:

7. Among them, μ and σ are the mean and standard deviation of the target mental health index of the reference group matched with the student's age and gender, respectively, and k is a preset multiplier. This is the rate of change of the current value relative to the historical mean, calculated based on the students' historical mental health data. The value is the ratio of individual data volatility to reference group data volatility, calculated based on the students' historical mental health data; α and β are the trend influence factor and stability influence factor, respectively.

8. The method for assessing the mental health of primary and secondary school students based on multimodal physiological data according to claim 1, characterized in that, The final quantitative assessment of mental health results is obtained by comparing the initial mental health status classification probability with the dynamic assessment threshold. The comparison yields the following result: if the classification probability is greater than or equal to... If the condition is abnormal, it is determined that there is a corresponding psychological health risk tendency, and a quantitative score is given according to the degree of abnormality; otherwise, it is determined to be normal.

9. A system for assessing the mental health of primary and secondary school students based on multimodal physiological data as described in any one of claims 1 to 7, characterized in that, include: The synchronous data acquisition module is configured to synchronously trigger and acquire EEG signals, heart rate signals, blood oxygen saturation signals, and facial video streams during mental health scale testing. The multimodal data processing module is configured to segment the collected data, synchronize and align the data over time, and extract features for each modality. The fusion classification module is configured to fuse multimodal features and process them using a multimodal Transformer classification model, outputting preliminary classification probabilities. The personalized assessment module is configured to calculate dynamic assessment thresholds based on individual differences and historical data, calibrate the initial classification probability, and generate quantitative assessment results. The intelligent reporting module is configured to visualize the quantitative assessment results and generate structured evaluation reports and intervention recommendations.

10. A psychological health assessment system for primary and secondary school students based on multimodal physiological data according to claim 8, characterized in that, The synchronous data acquisition module includes: a test terminal for presenting the test content of the mental health scale and sending a synchronous trigger signal; and an electroencephalogram (EEG), a photoelectric sensor, and a camera that are communicatively connected to the test terminal, receive the synchronous trigger signal, and begin data acquisition.

Images