Intelligent abnormal psychological behavior screening method for college student groups

Abstract

The invention discloses an intelligent abnormal psychological behavior screening method for college student groups, and relates to the technical field of psychological screening and artificial intelligence, and the method comprises the following specific steps: inducing a subject to respond through a standardized cognitive task; synchronously acquiring and processing eye movement and electroencephalogram signals; extracting multi-modal physiological features and carrying out fusion calculation to generate a time sequence cognitive coupling feature value; training a time sequence anomaly discrimination model based on the normal group data; and finally, the features of the person to be tested are input into the model, a discrimination coefficient is calculated, a grading screening result is output, and noninvasive, objective and efficient psychological state assessment and early warning are realized. According to the method, a non-invasive screening process suitable for a campus scene is constructed, multi-modal fusion and time sequence dynamic analysis are utilized, the accuracy and objectivity of college student psychological state assessment are improved, graded and personalized screening results and intervention suggestions can be output, and an efficient and quantifiable scientific tool is provided for campus psychological health management.

Description

Technical Field

[0001] This invention relates to the fields of psychological screening and artificial intelligence technology, specifically to an intelligent screening method for abnormal psychological behaviors in college students. Background Technology

[0002] The psychological state of current university students is affected by multiple factors such as academic pressure, social relationships, and employment anxiety, leading to an increasing probability of abnormal psychological behaviors. The importance and urgency of campus mental health management are becoming increasingly prominent. Accurate and efficient screening for psychological abnormalities has become a core component of the campus mental health service system. The external manifestations of abnormal psychological behaviors are closely related to internal cognition and physiological state. Judging psychological state by capturing changes in physiological signals during the cognitive process has become an important research direction in the field of mental health screening. With the development of intelligent sensing technology and artificial intelligence algorithms, the portability of physiological signal acquisition devices is constantly improving, and multimodal signal analysis and time-series model discrimination technologies are gradually maturing, providing technical support for intelligent screening of abnormal psychological behaviors. The widespread use of intelligent terminals and portable sensing devices in campus settings has also made routine psychological screening for university students feasible. Intelligent screening methods based on cognitive task induction and physiological signal analysis have become a research and application direction that meets the actual needs of campuses.

[0003] Traditional methods for screening college students for psychological abnormalities mainly rely on questionnaires and scales, depending on the subjects' subjective self-reports for result determination. This approach is easily influenced by factors such as emotions and cognitive biases, making it difficult to reflect the true psychological and physiological state. Furthermore, the screening results lack objective physiological data support, resulting in significant judgment bias and insufficient accuracy. Some screening methods based on physiological signals only use a single signal type for analysis, failing to comprehensively capture the psychological and physiological correlation characteristics in the cognitive process. The completeness and accuracy of feature representation are limited. At the same time, the signal collection and processing lack standardized procedures, making it difficult to uniformly compare the collection results in different scenarios. The discrimination models are mostly based on static features, failing to capture the dynamic changes in psychological states during the cognitive process. This leads to insufficient timeliness and targeting of the screening, making it difficult to adapt to the large-scale and routine screening needs on campus, and also failing to achieve graded determination and precise intervention of psychological abnormal tendencies. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent screening method for abnormal psychological behaviors among university students. This method induces the subject's response through three cognitive tasks: emotion, attention, and memory. It simultaneously collects eye movement and electroencephalogram (EEG) signals and performs temporal alignment and noise reduction processing. It uses a multimodal cognitive coupling algorithm to fuse eye movement and EEG features, constructs a temporal dynamic discrimination model, and outputs graded screening results by calculating the degree of deviation between the test subject and the characteristics of the normal group, thereby achieving objective and efficient identification and early warning of abnormal psychological tendencies.

[0005] To solve the above-mentioned technical problems, this invention provides the following technical solution: an intelligent screening method for abnormal psychological behaviors among university students, the specific steps of which are as follows: S100, cognitive task induction: Using a campus smart tablet computer terminal, three types of cognitive task sequences, namely emotional stimulus browsing, attention test and memory retrieval, are presented to college student subjects in sequence, and cognitive response is formed through standardized task stimuli. S200, synchronous signal acquisition: During the entire process of three types of cognitive tasks, eye movement trajectory signals and electroencephalogram (EEG) signals of the subjects are synchronously acquired using a unified timestamp benchmark. The acquired signals are then processed by artifact removal and smoothing to align the temporal sequence of the two types of signals, providing high-quality and highly synchronous data support for subsequent feature extraction and algorithm calculation. S300, Multimodal Coupling Processing: Feature extraction is performed on the time-aligned eye movement trajectory signal and EEG physiological signal to obtain the comprehensive value of eye movement micro-features and the comprehensive value of EEG rhythm features. The weight coefficients of eye movement features, EEG features, and time-series decay factors are determined. The multimodal cognitive coupling algorithm is used to calculate the above features and parameters to obtain the cognitive coupling feature values ​​at each sampling time. S400, Discrimination Model Training: Build a dynamic discrimination model for psychological anomalies based on the time-series anomaly discrimination algorithm. Use cognitive coupling feature values ​​as model input, determine the total number of time-series sampling points, the mean value of cognitive coupling features in the normal college student group, the maximum value of cognitive coupling features, the minimum value of cognitive coupling features, and the physiological adaptation correction coefficient of the college student group. Use normal college student sample data to iteratively calibrate and train the model to complete the optimization of model parameters. S500, Screening Result Output: The above cognitive task induction, signal synchronous acquisition, and multimodal coupling processing steps are repeated for the college student subjects to be tested to obtain their full-time cognitive coupling feature value. This feature value is input into the trained psychological abnormality dynamic discrimination model, and the psychological abnormality discrimination coefficient is calculated by the time-series abnormality discrimination algorithm. The corresponding psychological abnormality screening classification result is output according to the preset threshold, forming a complete screening process.

[0006] Furthermore, the total duration of the three cognitive tasks is controlled at 8-12 minutes, with the emotional stimulus browsing task lasting 3-4 minutes, the attention test task lasting 2-3 minutes, and the memory retrieval task lasting 3-5 minutes. A 10-15 second rest interval is set between each type of task. The emotional stimulus browsing task presents stimuli including positive, neutral, and negative emotional face images and scene images, with 30-40 images for each type of emotion. The image size is uniformly 1024×768 pixels, and the presentation rate is 2-3 seconds per image. The images are presented in a randomly shuffled order. The attention test task uses a number symbol matching paradigm, presenting numbers 0-9, each number corresponding to... The task uses a unique geometric symbol. Before starting the task, a 30-second explanation of the corresponding number symbol is presented. During the test, numbers are presented randomly, and subjects are required to click on the corresponding symbol within a specified time. If the click response time exceeds 2 seconds, the response result is recorded. The memory retrieval task uses the n-back paradigm, with n set to 2. A string of random numbers of 15-20 digits is presented first, at a rate of 1 digit per second. After the first string is presented, another string of numbers is presented, and subjects are required to identify the digits that are the same as the first two digits. The recognition accuracy and response time are recorded simultaneously. After the recognition is completed, the subject clicks to confirm. This ensures that the presentation process, stimulus specifications, and execution requirements of the three types of tasks are consistent and standardized.

[0007] Furthermore, in S200, during the synchronous signal acquisition, the eye-tracking signal reflects the dynamic changes in the subject's eyes during the cognitive task, including fixation point position, saccade amplitude, saccade direction, fixation duration, pupil diameter, and pupil dilation rate. The electroencephalogram (EEG) signal reflects the electrical activity state of relevant brain regions, primarily capturing electrical activity data from the prefrontal, temporal, parietal, and occipital lobes, which are related to psychological state and cognitive response. The eye-tracking signal is acquired via a campus smart tablet terminal with a sampling rate of 60Hz, an acquisition accuracy of 0.5° of visual angle, and a acquisition range covering the entire tablet screen. During the acquisition process, the subject's eye movements are captured in real time and converted into electrical signals for storage. Electroencephalogram (EEG) signals are acquired using a portable dry electrode EEG headband, which contains eight dry electrodes placed at preset locations in the subject's frontal, temporal, parietal, and occipital lobes. The sampling rate is set to 250 Hz, and the acquisition bandwidth is 0.5-30 Hz. During acquisition, the dry electrodes are in contact with the subject's scalp to capture the electrical activity of brain regions in real time and convert it into processable electrical signals. Both the eye-tracking acquisition module and the EEG headband are synchronized with the unified timestamp of the tablet terminal and start acquisition synchronously. During the acquisition process, artifact removal and smoothing are performed on both types of signals simultaneously to ensure that the timing of the two types of signals is aligned.

[0008] Furthermore, in the synchronous acquisition of the S200 signal, artifact removal and smoothing are performed according to the following steps: First, artifact removal was performed on the acquired eye movement trajectory signals and electroencephalogram (EEG) physiological signals, followed by smoothing. Artifact removal employed corresponding methods for the two types of signals. For EEG physiological signals, artifacts mainly included electromyography (EMG) artifacts and electrooculography (EOG) artifacts. A preset amplitude threshold of ±75μV was used. Signal segments with amplitudes exceeding this threshold were identified as artifacts and directly removed. After removal, missing signal segments were filled in to ensure the continuity of the EEG signals. For eye movement trajectory signals, artifacts mainly included blink interference artifacts and fixation point drift artifacts. Signal segments with fixation point stability below 0.3° / s and pupil diameter abrupt changes exceeding 0.5mm were identified as artifacts and removed. After removal, adjacent valid data were interpolated to fill in the missing data. For EEG physiological signals, the smoothing process sets a sliding window size of 5 sampling points and performs a point-by-point moving average calculation on the EEG signals after artifact removal. For eye movement trajectory signals, the sliding window size is set to 3 sampling points and performs a point-by-point moving average calculation on the eye movement signals after artifact removal. During the moving average calculation, the signal data at each time-series sampling moment is retained to ensure that the two types of signals can still maintain accurate temporal alignment after processing, which meets the needs of subsequent feature extraction and algorithm calculation.

[0009] The point-by-point moving average calculation is as follows: For EEG physiological signals, a sliding window of 5 sampling points is used, and the arithmetic mean of the signal values ​​of the current sampling point and the two sampling points before and after it is calculated as the smoothed signal value at the current time. For eye movement trajectory signals, a sliding window of 3 sampling points is used, and the arithmetic mean of the signal values ​​of the current sampling point and the one sampling point before and after it is calculated as the smoothed signal value at the current time. The first and last sampling points of the sequence are filled with boundary mirroring to supplement the missing sampling points, ensuring that the smoothing calculation of the entire sequence is completed, and that the two types of signals after smoothing correspond one-to-one with the original time sequence sampling time, maintaining accurate time sequence alignment.

[0010] Furthermore, in the S300 multimodal coupling processing, the comprehensive value of eye movement micro-features is obtained by extracting fixation point density, saccade amplitude, pupil dilation rate, fixation duration, saccade latency, and pupil diameter variation coefficient features from the time-aligned eye movement trajectory signal, and then calculating them by weighting after normalization. The comprehensive value of EEG rhythm features is obtained by performing wavelet decomposition on the time-aligned EEG physiological signal to extract alpha, theta, delta, and gamma wave related rhythm indicators, including alpha wave inhibition rate, theta wave absolute power, the theta wave to beta wave power ratio, delta wave relative power, and gamma wave event-related synchronization features. Both types of comprehensive values ​​correspond to the time-series sampling times of the three types of cognitive tasks and are directly used as the basic input data of the multimodal cognitive coupling algorithm.

[0011] Furthermore, in the S300 multimodal coupling process, the multimodal cognitive coupling algorithm adopts the following mathematical expression:

[0012] in, Let be the cognitive coupling feature value at the t-th time-series sampling time; t is the index of the time-series sampling time; i is the cognitive task type index, with values ​​of 1, 2, and 3, corresponding to emotional stimulus browsing task, attention test task, and memory retrieval task, respectively. Let be the comprehensive value of eye movement micro-features at the t-th temporal sampling time under the i-th cognitive task; The comprehensive value of EEG rhythm features at the t-th time sequence sampling time under the i-th cognitive task; These are the weighting coefficients for eye-tracking features; The weighting coefficients of EEG features and satisfying + =1; Let be the temporal decay factor at the t-th temporal sampling time under the i-th cognitive task.

[0013] Furthermore, in the training of the S400 discrimination model, the dynamic discrimination model for psychological abnormalities uses a time-series sequence composed of cognitive coupling feature values ​​at each time-series sampling moment as input. The input dimension is consistent with the total number of time-series sampling points, corresponding to the number of feature samples throughout the entire cognitive task cycle. The model adopts a deep neural network structure based on a time-series attention mechanism, including an input layer, multiple time-series encoding layers, and an abnormality discrimination output layer. The time-series encoding layer is used to capture the dynamic pattern of feature changes over time during the cognitive process, and the abnormality discrimination output layer outputs the abnormal scores at each sampling moment. During the training phase, the time-series sequence of cognitive coupling features of a normal university student group is used as the training set. The reconstruction loss is used as the optimization objective. First, the input features are preprocessed based on the mean, extreme values, and physiological adaptation correction coefficients of the cognitive coupling features of the normal group to eliminate the influence of group differences and dimensions. Then, the model parameters are iteratively updated until the reconstruction error converges to the preset threshold of 0.001, completing the model calibration and training. After training, the model can identify abnormal features that deviate from the normal cognitive time-series pattern by calculating the abnormal scores of the feature sequences of the test samples, thereby achieving dynamic discrimination of psychological abnormalities.

[0014] Furthermore, the mathematical expression for the time-series anomaly detection algorithm in the S500 screening result output is:

[0015] in, The psychological abnormality discrimination coefficient is used to characterize the degree of psychological abnormality tendency of the college student subjects being tested. The higher the value, the more significant the psychological abnormality tendency. The total number of temporal sampling points throughout the entire cognitive task cycle; Let be the cognitive coupling feature value of the college student subject at the t-th time-series sampling time; is the mean value of the cognitive coupling characteristics of the normal college student group at the t-th time series sampling moment; is the maximum value of the cognitive coupling characteristics of the normal college student group; is the minimum value of the cognitive coupling characteristics of the normal college student group; is the physiological adaptation correction coefficient of the college student group, which is used to correct the influence of group physiological differences on the discrimination result.

[0016] Furthermore, in the S500 screening result output, the determination rules for the psychological abnormality screening classification results are as follows: When the psychological abnormality discrimination coefficient M ≤ 0.3, it is determined that the psychological state is normal, and the screening result of no obvious psychological abnormality tendency is output; When 0.3 < M ≤ 0.6, it is determined that there is a mild psychological abnormality tendency, and the screening result of having a mild psychological abnormality tendency and suggesting daily psychological adjustment and attention is output; When 0.6 < M ≤ 0.8, it is determined that there is a moderate psychological abnormality tendency, and the screening result of having a moderate psychological abnormality tendency and suggesting professional psychological counseling and evaluation is output; When M > 0.8, it is determined that there is a severe psychological abnormality tendency, and the screening result of having a severe psychological abnormality tendency and suggesting immediate professional psychological intervention and medical diagnosis and treatment is output. The above classification results are output in the form of a standardized report, including the cognitive coupling characteristic sequence of the subject, the abnormality discrimination coefficient, and the corresponding classification suggestions, forming a complete psychological abnormality screening process.

[0017] Compared with the prior art, the intelligent screening method for abnormal psychological behaviors of college student groups has the following beneficial effects: First, by building a dedicated intelligent screening system for abnormal psychological behaviors of college student groups, relying on campus intelligent terminals to induce cognitive tasks, synchronously collecting eye movement and electroencephalogram dual-modal physiological signals and completing time series alignment and refined processing, comprehensively capturing multi-dimensional cognitive response data, fusing eye movement micro-features and electroencephalogram rhythm features through a multi-modal coupling algorithm, constructing a coupling eigenvalue that fits the cognitive process in combination with a time series attenuation factor, breaking through the limitations of single physiological signal analysis, making feature extraction more in line with the dynamic law of college students' cognitive activities, building a dynamic discrimination model based on a time series abnormality discrimination algorithm, and completing model training with the normal group feature benchmark, enabling the discrimination process to accurately capture abnormal patterns in the cognitive time series, greatly improving the matching degree of the screening result and the actual psychological state of college students, and realizing the upgrade from static detection to dynamic discrimination.

[0018] Second, this invention constructs a standardized intelligent screening process that can be implemented in campus settings through standardized cognitive task sequences, signal acquisition procedures, and data processing methods. It eliminates the dependence on professional laboratory equipment and manual operation, adapts to the campus life of university students, and has the advantages of convenience and universality. By setting multi-gradient psychological abnormality screening grading standards, it outputs targeted intervention suggestions based on the discrimination coefficient, realizing the transformation from single result judgment to graded and personalized screening feedback. It can quickly identify severe psychological abnormality tendencies and prompt emergency intervention, and also provide appropriate adjustment and counseling suggestions for mild and moderate tendencies. This makes the screening results have practical application value and provides accurate quantitative reference and graded treatment basis for campus mental health management.

[0019] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0021] Figure 1 A flowchart for an intelligent screening method for abnormal psychological behaviors in college students; Figure 2 This is a data transmission diagram of an intelligent screening method for abnormal psychological behaviors among university students. Figure 3 A schematic diagram of data transmission in the signal synchronization acquisition step of this invention. Detailed Implementation

[0022] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0023] Example 1: Implementation of a comprehensive screening program for abnormal psychological behaviors among new college students This embodiment is applied to a university-wide screening for abnormal psychological behaviors among freshmen during their enrollment. The screening targets are all freshmen enrolled that year at a specific university. The screening venues are centrally located in multimedia classrooms within the university's teaching buildings. Each classroom is staffed with professional personnel responsible for equipment setup, operation guidance, and process control. Each participant is individually equipped with a campus smart tablet terminal and a portable dry electrode EEG headband. All devices undergo unified timestamp calibration beforehand to ensure the synchronization and accuracy of signal acquisition from a hardware perspective. The entire screening process is strictly implemented according to the method and steps of this invention. Relying on standardized operating procedures, it achieves efficient and accurate assessment of the freshmen's psychological state, providing objective and scientific physiological and cognitive data support for the establishment of campus mental health records. Figure 1 As shown, the specific operation is as follows: Cognitive task induction: Staff guided new participants to sit in order according to their classes, and each participant was properly fitted with a portable dry electrode EEG headband, ensuring a tight and secure fit between the electrodes and the scalp. The equipment was adjusted to normal working condition to avoid affecting subsequent signal acquisition due to contact issues. Simultaneously, participants were provided with campus smart tablet terminals and guided to complete basic terminal operation confirmations, maintaining a stable testing posture while holding the terminal. The terminal automatically presented a sequence of three cognitive tasks according to a preset program: emotional stimulus browsing, attention testing, and memory retrieval. The three tasks were seamlessly connected, ensuring continuous cognitive response. The emotional stimulation browsing task presents three categories of emotional materials: positive, neutral, and negative. These include an equal number of images of emotional faces and everyday scenes, all uniformly sized and presented in random order with a fixed presentation duration. This diverse range of emotional materials comprehensively induces the subjects' emotional cognitive responses, covering psychological state feedback across different emotional dimensions. The attention test task first displays a fixed-duration explanation of numerical symbols to ensure the subjects fully understand the task rules. Then, randomly presented numbers are presented, requiring subjects to quickly click on the corresponding geometric symbols. The terminal simultaneously records the subjects' response speed and accuracy, accurately capturing their level of attention and cognitive reaction ability through a numerical symbol matching paradigm. The memory retrieval task employs a number n-back paradigm, presenting a string of numbers strictly according to the paradigm requirements and requiring subjects to identify the numbers matching the first two digits. The terminal simultaneously records the identification results, exploring the subjects' working memory and cognitive processing abilities through this task. These three cognitive tasks form a standardized stimulus sequence, comprehensively and systematically inducing the subjects' cognitive responses. This provides a sufficient and effective cognitive stimulus foundation for subsequent physiological signal collection, ensuring that the collected signals accurately reflect the subjects' psychological and physiological states.

[0024] Synchronous Signal Acquisition: Throughout the three cognitive tasks performed by the subjects, the eye-tracking acquisition module and the portable dry electrode EEG headband were simultaneously activated using a unified timestamp from the campus smart tablet terminal as a benchmark. This enabled the synchronous acquisition of eye movement trajectory signals and EEG physiological signals. The unified timestamp benchmark ensured complete temporal matching between the two types of signals, laying the foundation for subsequent temporal alignment. The eye-tracking acquisition module captured the subjects' eye movement trajectory signals in real time, including all dynamic eye-related data such as fixation point position, saccade amplitude, saccade direction, fixation duration, pupil diameter, and pupil dilation rate. This accurately recorded the subjects' subtle eye movements during the cognitive task execution, and the dynamic eye data directly reflected the subjects' attention allocation, emotional responses, and cognitive processing. The portable dry electrode EEG headband captured the electrical activity data of the prefrontal, temporal, parietal, and occipital lobes, as well as brain regions related to psychological state and cognitive response, forming EEG physiological signals. The electrical activity data of these brain regions physiologically reflected the subjects' cognitive activities and changes in psychological state. The two types of signals complemented each other, achieving multi-dimensional capture of the subjects' psychological and physiological states. During the acquisition process, artifact removal and smoothing were performed on both types of signals simultaneously to avoid various interference factors affecting signal quality. Specifically, electromyography (EMG) and electrooculography (EOG) artifacts exceeding amplitude thresholds were specifically removed from the EEG signals. After removal, missing signal segments were filled in to ensure the integrity and effectiveness of the EEG signals. For eye movement (EMG) signals, blink interference and fixation drift artifacts were removed. After removal, adjacent valid data interpolation was used to fill in the gaps, ensuring that the eye movement signals accurately reflect the subject's eye movements. After artifact removal, sliding windows were set for each type of signal to perform moving average calculations, further reducing noise interference, making the signal curves smoother, and the data characteristics more obvious. Ultimately, precise temporal alignment of the eye movement signals and EEG signals was achieved, ensuring a one-to-one matching relationship between the two types of signals at each sampling time. This provides high-quality, highly matched signal data for subsequent multimodal coupling processing. Figure 3 As shown.

[0025] Multimodal coupling processing: Systematic feature extraction is performed on the time-aligned eye movement trajectory signal and EEG signal respectively. The feature extraction process strictly follows signal analysis standards, comprehensively extracting micro-features such as fixation density, saccade amplitude, pupil dilation rate, fixation duration, saccade latency, and pupil diameter variation coefficient from the eye movement trajectory signal. The extracted micro-features are normalized to eliminate dimensional differences between different features, and then a weighted calculation is performed to obtain a comprehensive value of eye movement micro-features. The comprehensive value can centrally reflect the core features in the eye movement trajectory signal, making the complex eye movement signal data into a concise and representative feature index. From the EEG signal, wavelet decomposition is used to extract alpha, theta, delta, and gamma wave related rhythm indicators, including alpha wave inhibition rate, theta wave absolute power, the theta wave to beta wave power ratio, delta wave relative power, and gamma wave event-related synchronization features. Based on the extracted rhythm indicators, a comprehensive value of EEG rhythm features is calculated. The comprehensive value can accurately reflect the rhythmic change pattern in the EEG signal and capture the electrical activity characteristics of relevant brain regions. Subsequently, based on the physiological and cognitive characteristics of university students and the testing scenario of this screening, and combined with a large amount of previous experimental data, eye-tracking feature weighting coefficients, electroencephalogram (EEG) feature weighting coefficients, and time-series decay factors were determined to be suitable for the freshman population. The determination of these coefficients and factors fully aligns with the physiological and psychological characteristics of the freshman population, ensuring the accuracy of subsequent calculation results. Based on the determined coefficients, factors, and the extracted combined values ​​of the two types of features, a multimodal cognitive coupling algorithm was used for calculation. The multimodal cognitive coupling algorithm uses the following mathematical expression:

[0026] in, Let be the cognitive coupling feature value at the t-th temporal sampling time; t is the index of the temporal sampling time; i is the index of the cognitive task type. Let be the comprehensive value of eye movement micro-features at the t-th temporal sampling time under the i-th cognitive task; The comprehensive value of EEG rhythm features at the t-th time sequence sampling time under the i-th cognitive task; These are the weighting coefficients for eye-tracking features; These are the weighting coefficients for EEG features; Let t be the temporal decay factor at the t-th time-series sampling time under the i-th cognitive task. The algorithm can effectively explore the intrinsic cognitive correlation between eye movement and EEG signals, achieve deep fusion of the features of the two types of signals, and calculate the cognitive coupling feature value of the subject at each sampling time. This feature value can comprehensively reflect the cognitive state and psychophysiological characteristics of the subject at each sampling time, transforming multi-dimensional signal data into a single and representative feature index, providing concise and effective data support for the input of the subsequent discrimination model.

[0027] Discrimination Model Training: Prior to this screening, a dynamic discrimination model for psychological abnormalities was built and trained. The model is based on a temporal anomaly discrimination algorithm and employs a deep neural network structure based on a temporal attention mechanism. This structure effectively captures the dynamic patterns of feature changes over time during the cognitive process, accurately identifies abnormal changes in temporal features, and adapts to the dynamic characteristics of psychological states. During the model training phase, cognitive coupling feature values ​​collected in advance from normal university students of different grades at our university were used as model input. The input data covered the physiological cognitive characteristics of our university students, making the model more closely aligned with the actual situation of our students. Simultaneously, in conjunction with the cognitive task settings of this screening, key parameters such as the total number of temporal sampling points, the mean and extreme values ​​of cognitive coupling features of normal university students, and the physiological adaptation correction coefficient adapted to our freshmen group were determined. The determination of various parameters fully considered the campus setting and the characteristics of the student group, eliminating the influence of group differences. Model training used reconstruction loss as the optimization objective. After preprocessing the input feature data, the model parameters were iteratively updated multiple times to continuously optimize the model's discrimination ability until the reconstruction error converged to a preset threshold, completing the comprehensive calibration and training of the model. The trained model can accurately identify abnormal features of cognitive temporal patterns and form an accurate discrimination ability for cognitive coupling feature value sequences of different psychological states. It provides scientific and reliable model support for the determination of the results of this newborn screening, ensuring the objectivity and accuracy of the screening results.

[0028] Screening Results Output: After completing the cognitive task induction, synchronous signal acquisition, and multimodal coupling processing for all new students, the system automatically generates full-time cognitive coupling feature values ​​for each new student. This feature value sequence completely records the changes in the subject's psychophysiological characteristics throughout the entire cognitive task execution process. The full-time cognitive coupling feature values ​​of each subject are then input into the trained dynamic discrimination model for psychological abnormalities. The model performs a comprehensive analysis and calculation of the input feature value sequence based on a time-series anomaly discrimination algorithm. The mathematical expression of the time-series anomaly discrimination algorithm is as follows:

[0029] in, The coefficient for detecting psychological abnormalities; The total number of temporal sampling points throughout the entire cognitive task cycle; Let be the cognitive coupling feature value of the college student subject under test at the t-th time-series sampling time, where t is the index of the time-series sampling time; Let be the mean of the cognitive coupling characteristics of a normal college student group at the t-th time-series sampling time. This represents the maximum value of the cognitive coupling characteristics of a normal university student population. This represents the minimum value of cognitive coupling characteristics in a normal university student population. A physiological adaptation correction coefficient was established for university students to accurately compare the characteristics of the test samples with those of the normal group. This resulted in the calculation of a psychological abnormality discrimination coefficient for each subject. This coefficient quantifies the degree of deviation between the subject's psychological state and that of the normal group, providing a quantitative indicator for the classification of psychological abnormality tendencies. Based on preset threshold judgment rules, the psychological abnormality discrimination coefficient for each subject was standardized and classified. Screening results were output according to the coefficient value range, categorizing the psychological state as normal, mild psychological abnormality tendency, moderate psychological abnormality tendency, and severe psychological abnormality tendency. Personalized and standardized psychological screening reports were generated for each freshman. These reports comprehensively included their cognitive coupling characteristic sequence, psychological abnormality discrimination coefficient, and corresponding classification suggestions. The reports were standardized in content and detailed in data, providing a clear reference for subsequent mental health management work. The university's mental health center conducts systematic follow-up processing based on the screening results. Freshmen deemed to have normal mental health are included in the regular mental health management system. For freshmen diagnosed with mild psychological abnormalities, special attention files are created, and counselors are assigned to provide daily psychological adjustment guidance, using methods such as psychological science popularization and emotional counseling to help students adjust their mental state. For freshmen diagnosed with moderate psychological abnormalities, the university's mental health center arranges professional psychological counselors to conduct one-on-one psychological counseling and assessment, deeply analyzing the students' psychological problems and developing personalized adjustment plans. For freshmen diagnosed with severe psychological abnormalities, a home-school collaboration mechanism is immediately activated, with the university's mental health center connecting with professional mental health institutions to provide students with professional psychological intervention and medical treatment services. Counselors and parents are also assigned to provide full-process follow-up to ensure the students' mental health. Through this comprehensive screening and follow-up intervention, the university achieves accurate assessment and tiered management of freshmen's mental health, laying a solid foundation for the construction of a campus mental health service system, effectively preventing campus psychological crises, and helping freshmen quickly adapt to university life.

[0030] This embodiment targets a comprehensive screening scenario for new university students. Utilizing standardized facilities and uniformly calibrated equipment, it implements a full-process screening process, including cognitive task induction, synchronous signal acquisition, multimodal coupling processing, discriminant model training, and screening result output. Each step is tailored to the characteristics of the new student population. A standardized cognitive task induces authentic responses, eye-tracking and EEG signals are simultaneously acquired and processed for temporal alignment, feature values ​​are obtained through a multimodal cognitive coupling algorithm, and discriminant coefficients are calculated and graded results are output based on the trained model. The screening results are then used to implement tiered management and subsequent interventions, achieving accurate assessment of the new students' psychological state, providing data support for establishing campus mental health records, and helping new students adapt to university life.

[0031] Example 2: Implementation of Special Screening for Abnormal Psychological Behaviors among Graduating Students in Higher Education Institutions This embodiment is applied to a targeted screening program for abnormal psychological behaviors among graduating students in a university. The screening targets all senior students at a certain university, aiming to accurately assess their psychological state under the multiple pressures of academic defenses, job hunting, and college entrance exams during graduation season. The goal is to promptly identify any abnormal psychological tendencies and provide targeted intervention. The screening is conducted in a dedicated testing room at the university's Student Mental Health Center. The quiet and comfortable environment minimizes external interference, allowing participants to maintain a relaxed state and ensuring the accuracy of the test results. The testing room is equipped with multiple professional campus smart tablet terminals and portable dry electrode EEG headbands. All devices have undergone professional debugging and are synchronized with a unified timestamp. Professional psychological counselors and staff from the university's Mental Health Center are responsible for the screening, providing professional guidance and psychological support throughout the process. The entire screening process strictly follows the method steps of this invention, relying on scientific and standardized screening methods to accurately detect the psychological state of graduating students, providing a scientific basis for campus mental health management during graduation season. Figure 2 As shown, the specific operation is as follows: Cognitive Task Induction: Graduating students, organized by class, went to the school's mental health center testing room in batches to participate in the screening. Staff guided students into the testing room beforehand and provided brief psychological counseling to alleviate their anxiety and help them participate in the test with a calm mindset. Staff ensured each participant wore a portable dry electrode EEG headband correctly, carefully checking the fit between the electrodes and the scalp to ensure no poor contact or loosening. The headband was also adjusted to normal working condition to ensure effective acquisition of EEG signals. Subsequently, participants were given a campus smart tablet computer and guided to familiarize themselves with its operation. After confirming that all functions were normal, participants were instructed to maintain a comfortable and stable testing posture. The terminal presented participants with three cognitive task sequences according to a preset program: emotional stimulus browsing, attention testing, and memory retrieval. These three tasks were seamlessly connected in a fixed order, forming a standardized cognitive stimulus sequence to comprehensively induce the participants' cognitive responses. The emotional stimulation browsing task presents three categories of emotional images—positive, neutral, and negative—in equal quantities and of uniform size, including images of emotional faces and everyday scenes. The images are presented in random order with a fixed presentation duration per image. By using materials with different emotional dimensions, the task aims to elicit emotional responses from participants, accurately capturing the emotional state and emotional regulation abilities of graduating students and reflecting their emotional changes when facing pressure. The attention test task first displays a fixed-duration explanation of the corresponding numerical symbols to allow participants to fully understand the task rules. Then, numbers are presented randomly, requiring participants to quickly click on the corresponding geometric symbols. The terminal simultaneously records the participants' response speed, accuracy, and error rate, accurately detecting their attention through a numerical symbol matching paradigm. The task assesses students' concentration, cognitive reaction speed, and resistance to interference. Graduation season students' attention is easily distracted by multiple pressures, and this task effectively reflects their attentional state. The memory retrieval task uses a digit n-back paradigm, strictly adhering to the paradigm requirements. A fixed-length random digit string is presented at a fixed rate, followed by another string. Participants are required to identify the digits matching the first two and confirm. The terminal simultaneously records data such as recognition accuracy and reaction time. The memory retrieval task assesses participants' working memory, cognitive processing ability, and psychological load. Graduation season students experience significant psychological burden, and this task accurately reflects their cognitive state and stress levels. These three cognitive tasks are designed to closely align with the psychological characteristics and cognitive state of graduation season students. Standardized task stimuli allow participants to develop genuine and effective cognitive responses, providing a sufficient cognitive stimulus foundation for subsequent physiological signal collection. This ensures that the collected signals accurately reflect the true psychological and physiological state of graduation season students under multiple pressures.

[0032] Synchronous signal acquisition: Throughout the process of subjects performing three types of cognitive tasks, the eye-tracking acquisition module of the campus smart tablet terminal and the portable dry electrode EEG headband were simultaneously activated based on the unified timestamp of the campus smart tablet terminal. This enabled the synchronous and real-time acquisition of eye-tracking trajectory signals and EEG physiological signals. The unified timestamp ensured that the two types of signals were highly synchronized in the time dimension, ensuring that the two types of signals could be accurately matched at each sampling moment, thus providing a foundation for subsequent temporal alignment and multimodal coupling processing. The eye-tracking acquisition module utilizes the high-definition camera and eye-tracking recognition technology of the campus smart tablet terminal to capture the subject's eye movement trajectory signals in real time. It comprehensively records dynamic changes in the eyes, such as fixation point position, saccade amplitude, saccade direction, fixation duration, pupil diameter, and pupil dilation rate. Eye dynamics are a direct external manifestation of psychological state. Data such as pupil diameter changes and fixation duration can accurately reflect students' emotional arousal level, attention allocation, and psychological stress level. Saccade amplitude and direction can reflect students' cognitive processing and information acquisition methods. The portable dry electrode EEG headband collects electrical activity data from brain regions related to psychological state and cognitive response, such as the prefrontal, parietal, occipital, and temporal lobes, in real time, forming EEG physiological signals. The prefrontal lobe is related to emotion regulation and cognitive control, the temporal lobe is related to memory and emotional processing, the parietal lobe is related to attention and spatial cognition, and the occipital lobe is related to visual processing. The electrical activity data of each brain region can reflect the cognitive activities, emotional state, and psychological stress of students during the graduation season from a physiological perspective. The simultaneous acquisition of these two types of signals enables multi-dimensional and comprehensive capture of students' psychological and physiological state, making up for the limitations of single signal analysis. During the acquisition process, artifact removal and smoothing were performed on both types of signals simultaneously to minimize the impact of various interference factors on signal quality. For electromyography (EMG) and electrooculography (EOG) artifacts in the EEG signals, corresponding methods were used for removal. After removal, missing signal segments were filled in to ensure the integrity and continuity of the EEG signals. For eye movement (EMG) trajectory signals, blink interference and fixation drift artifacts were accurately identified and removed. After removal, adjacent valid data interpolation was used to fill in the gaps, ensuring that the eye movement signals accurately reflect the dynamic changes in the subject's eyes. After artifact removal, corresponding sliding windows were set for both the EEG and EEG trajectory signals to perform point-by-point moving average calculations, further filtering noise from the signals and making the signal data smoother and more prominent. After all processing was completed, the temporal alignment of the EEG trajectory signals and EEG signals was achieved, ensuring a one-to-one matching relationship between the two types of signals at each sampling time. This provides high-quality, highly matched signal data for subsequent multimodal coupling processing, ensuring the accuracy of subsequent feature extraction and calculation results.

[0033] Multimodal coupling processing: Comprehensive and systematic feature extraction is performed on the time-aligned eye movement trajectory signals and electroencephalogram (EEG) signals. The feature extraction process strictly follows professional standards for signal analysis and processing to ensure that the extracted features accurately reflect the core patterns of the signals. Multiple micro-features, such as fixation point density, saccade amplitude, pupillary dilation rate, fixation duration, saccade latency, and pupillary diameter variation coefficient, are accurately extracted from the eye movement trajectory signals. These micro-features reflect the subject's eye movement patterns and dynamic characteristics from different dimensions. The extracted micro-features are normalized to eliminate dimensional differences and numerical range influences between different features, allowing for effective fusion of various features. A weighted calculation yields a comprehensive eye movement micro-feature value, which transforms the complex eye movement trajectory signal into a single, concise, and highly representative feature index that centrally reflects the core information in the eye movement signal. Wavelet decomposition was performed on the electroencephalogram (EEG) signals to accurately extract rhythmic indicators related to alpha, theta, delta, and gamma waves. These indicators included alpha wave inhibition rate, absolute theta wave power, the theta wave to beta wave power ratio, delta wave relative power, and gamma wave event-related synchronization characteristics. These rhythmic indicators accurately reflect the electrical activity patterns and changes in relevant brain regions and are important physiological indicators reflecting psychological states and cognitive activities. Based on the extracted rhythmic indicators, a comprehensive value of EEG rhythmic characteristics was calculated. This comprehensive value centrally reflects the core rhythmic characteristics of the EEG signals, transforming complex EEG signal data into representative characteristic indicators. Combining the physiological and cognitive characteristics of graduating students, the psychological stress of graduation season, and the testing scenario of this special screening, extensive preliminary experimental data and analysis were used to determine eye-tracking feature weighting coefficients, EEG feature weighting coefficients, and temporal decay factors suitable for graduating students. The determination of these coefficients and factors fully reflects the actual situation of graduating students, effectively highlighting characteristics related to psychological stress and emotional states, allowing subsequent calculation results to more accurately reflect the psychological state of graduating students. Based on determined eye-tracking feature weighting coefficients, EEG feature weighting coefficients, temporal decay factors, and extracted comprehensive values ​​of eye-tracking micro-features and EEG rhythm features, a multimodal cognitive coupling algorithm is used for deep computation. This algorithm effectively uncovers the intrinsic cognitive correlation between eye-tracking and EEG signals, achieving deep fusion and coupling of the two signal features. It overcomes the limitations of single-signal analysis, fully leveraging the complementary advantages of the two signals to calculate the cognitive coupling feature values ​​of the subject at each sampling time. These feature values ​​comprehensively reflect the subject's cognitive state, emotional changes, and psychophysiological characteristics at each sampling time, transforming multi-dimensional and complex signal data into a concise and effective single feature indicator. It fully records the changes in the subject's psychophysiological state throughout the entire cognitive task execution process, providing high-quality and representative data support for the input of subsequent dynamic discrimination models for psychological abnormalities.

[0034] Model Training: The dynamic discrimination model for psychological abnormalities used in this special screening was built based on a temporal anomaly discrimination algorithm. The model employs a deep neural network structure containing an input layer, multiple temporal coding layers, and an anomaly discrimination output layer. The input layer receives the sequence of cognitive coupling feature values. The multiple temporal coding layers effectively capture the dynamic patterns of feature changes over time during the cognitive process, accurately identifying subtle changes and abnormal patterns in temporal features. The anomaly discrimination output layer outputs the anomaly scores at each sampling time, achieving accurate identification of psychological abnormalities. This structure is well-suited to the dynamic and complex characteristics of psychological states, effectively identifying the psychological abnormalities of graduating students under multiple pressures. During the model training phase, the cognitive coupling feature values ​​of normal undergraduate students (excluding graduating students) from this university were used as the main input data. This was combined with cognitive coupling feature data from previous graduating students, ensuring the model both aligns with the overall physiological and cognitive characteristics of the university's students and accurately distinguishes between normal and abnormal psychological states during graduation season. In advance, based on the cognitive task settings and testing procedures for this special screening, key parameters such as the total number of time-series sampling points, the mean and extreme values ​​of cognitive coupling characteristics in a normal university student group, and the physiological adaptation correction coefficient adapted to the university student group were determined. The determination of these parameters fully considered the differences between the campus setting and the student group, effectively eliminating the influence of group differences and units of measurement. Before model training, the input features were standardized and preprocessed based on the determined parameters to make the input data more consistent with the model's training requirements. Then, using reconstruction loss as the optimization objective, the model parameters were iteratively updated multiple times to continuously optimize the model's discriminative ability and accuracy until the reconstruction error converged to the preset threshold, completing the comprehensive training and calibration of the model. The trained model can accurately capture the dynamic patterns of feature changes over time during the cognitive process, effectively identify abnormal features that deviate from the normal cognitive time-series patterns, and accurately judge the psychological state of graduating students. This provides scientific and reliable model support for the judgment of the results of this special screening, ensuring that the screening results truly reflect the psychological state of graduating students.

[0035] Screening Results Output: After cognitive task induction, synchronous signal acquisition, and multimodal coupling processing for each graduating class subject, the system automatically performs comprehensive calculations on the acquired signal data and extracted features to generate a full-time cognitive coupling feature value for each subject. This feature value sequence completely and continuously records the changes in the subject's psychophysiological characteristics throughout the cognitive task execution process, accurately reflecting the subject's cognitive state and emotional changes under multiple pressures. The full-time cognitive coupling feature values ​​of each subject are input into a trained dynamic discrimination model for psychological abnormalities. Based on a temporal abnormality discrimination algorithm, the model performs comprehensive and detailed analysis and calculation of the input feature value sequence, accurately comparing the feature differences between the test sample and the normal group. Combining the abnormality scores at each sampling time, a comprehensive calculation is performed to obtain the psychological abnormality discrimination coefficient for each subject. This coefficient can quantitatively reflect the degree of deviation between the subject's psychological state and the normal group, providing an objective and quantitative indicator for the classification and determination of psychological abnormality tendencies. Based on pre-defined grading rules and the numerical range of the psychological abnormality discrimination coefficient, the psychological state of each subject was standardized and graded, categorized as normal, mild, moderate, or severe psychological abnormality, with corresponding screening results output. A personalized, standardized psychological screening report was generated for each subject, comprehensively including their cognitive coupling characteristic sequence, psychological abnormality discrimination coefficient, and targeted grading recommendations. The report was professional and detailed, providing clear and specific reference for subsequent psychological interventions. The school's mental health center developed personalized and graded follow-up psychological intervention plans based on the screening results. For graduating students assessed as having normal psychological states, lectures on graduation season psychological adjustment were conducted, providing students with methods and techniques for emotion regulation and stress relief, guiding them to maintain a positive mental state. For graduating students assessed as having mild psychological abnormality tendencies, group counseling was conducted, using methods such as group discussions, emotional guidance, and stress release to help students alleviate graduation season stress, adjust their mental state, and improve their emotion regulation abilities. For students assessed as having moderate psychological abnormality tendencies, specialists from the school's mental health center provided further intervention. Professional psychological counselors arrange one-on-one professional psychological counseling and career / further education guidance, deeply analyze the root causes of students' psychological problems, and develop personalized psychological adjustment plans based on students' employment and further education needs. They also provide relevant guidance on employment and further education to alleviate students' real-world pressures. For students diagnosed with severe psychological abnormalities, an emergency intervention mechanism is immediately activated. The school's mental health center connects with professional psychological treatment institutions to provide students with professional psychological intervention and medical treatment services. Simultaneously, counselors and parents work together to provide full-process follow-up and support, promptly resolving students' psychological crises and ensuring their mental health and safety.Through this special screening and subsequent tiered intervention, we can accurately alleviate the psychological pressure on graduating students, promptly identify and intervene in students' abnormal psychological tendencies, effectively prevent the occurrence of campus psychological crises during the graduation season, and help graduating students smoothly navigate the graduation season and successfully enter society.

[0036] This embodiment targets a specific screening scenario for graduating university students during graduation season. The screening is conducted in a professional assessment room, alleviating students' anxiety throughout the process. Cognitive tasks tailored to the psychological characteristics of graduating students are used to elicit responses. Eye movement and electroencephalogram (EEG) signals are simultaneously collected and processed. A multimodal cognitive coupling algorithm is used to fuse features to obtain cognitive coupling feature values. A specially trained model is then used to calculate discriminant coefficients and classify students into different levels. Based on the screening results, personalized intervention plans are developed for students with different psychological states. This comprehensive intervention, from science lectures to professional diagnosis and treatment, precisely alleviates the stress of graduating students, promptly intervenes in abnormal psychological tendencies, and prevents campus psychological crises.

[0037] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent screening of abnormal psychological behaviors in university students, characterized in that, The specific steps of this method are as follows: S100, cognitive task induction: Using a campus smart tablet computer terminal, three types of cognitive task sequences, namely emotional stimulus browsing, attention test and memory retrieval, are presented to college student subjects in sequence, and cognitive response is formed through standardized task stimuli. S200, synchronous signal acquisition: During the entire process of three types of cognitive tasks, eye movement trajectory signals and electroencephalogram (EEG) signals of the subjects are synchronously acquired using a unified timestamp reference. The acquired signals are then processed for artifact removal and smoothing to align the timing of the two types of signals. S300, Multimodal Coupling Processing: Feature extraction is performed on the time-aligned eye movement trajectory signal and EEG physiological signal to obtain the comprehensive value of eye movement micro-features and the comprehensive value of EEG rhythm features. The weight coefficients and time-series decay factors of eye movement and EEG features are determined, and the cognitive coupling feature value at each sampling time is calculated using a multimodal cognitive coupling algorithm. S400, Discrimination Model Training: Build a dynamic discrimination model for psychological abnormalities based on the temporal anomaly discrimination algorithm, take the cognitive coupling feature value as the model input, and determine the total number of temporal sampling points, the mean and extreme values ​​of cognitive coupling features of normal college students, and the physiological adaptation correction coefficient. S500, Screening Result Output: Repeat the above steps of cognitive task induction, signal synchronous acquisition, and multimodal coupling processing for the college student subject to be tested to obtain its full-time cognitive coupling feature value. Input the feature value into the trained psychological abnormality dynamic discrimination model, calculate the psychological abnormality discrimination coefficient through the time-series abnormality discrimination algorithm, and output the corresponding psychological abnormality screening classification result according to the preset threshold.

2. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In S100, during the cognitive task induction, the main contents of the three types of cognitive tasks are as follows: The emotional stimulus browsing task presents three types of emotional materials: positive, neutral, and negative, including emotional face pictures and daily scene pictures. The number of each type of emotional picture is equal, the picture size is uniform, they are presented in random order, and the presentation time of each picture is fixed. The attention test task uses a number symbol matching paradigm, presenting ten numbers from 0 to 9 and their corresponding unique geometric symbols. Before the task, a fixed-duration explanation of the number symbols is presented. During the test, numbers are presented randomly, and subjects are required to click on the corresponding geometric symbols and record their responses. The memory retrieval task uses a number n-back paradigm, with n fixed at 2. A fixed-length random number string is presented first, at a fixed rate. After the first string is presented, another string of numbers is presented, and subjects are required to identify the numbers that are the same as the first two digits and record their identification. After completing the identification, subjects click to confirm.

3. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the S200 signal synchronous acquisition, the eye movement trajectory signal is a signal that can reflect the dynamic changes of the subject's eyes during the cognitive task execution, including fixation point position, saccade amplitude, saccade direction, fixation duration, pupil diameter and pupil dilation rate. Electroencephalogram (EEG) signals are signals that reflect the electrical activity state of relevant brain regions in a subject's brain. They mainly capture electrical activity data from the prefrontal, temporal, parietal, and occipital lobes, which are related to psychological state and cognitive response. Eye-tracking signals are collected through a campus smart tablet terminal. During the collection process, the subject's eye movements are captured in real time and converted into electrical signals for storage. EEG signals are collected through a portable dry electrode EEG headband. During the collection process, the dry electrodes are attached to the subject's scalp to capture the electrical activity of brain regions in real time and convert it into processable electrical signals. Both the eye-tracking acquisition module and the EEG headband are synchronized with the unified timestamp of the tablet terminal and start the acquisition synchronously. During the acquisition process, artifact removal and smoothing are performed on both types of signals simultaneously.

4. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the synchronous acquisition of the S200 signal, artifact removal and smoothing processes are specifically as follows: Artifact removal employs corresponding methods for the two types of signals. Artifacts in EEG physiological signals mainly include EMG artifacts and EEG artifacts. A preset amplitude threshold of ±75μV is used. Signal segments with amplitudes exceeding this threshold in the acquired EEG signals are identified as artifacts and directly removed. After removal, missing signal segments are filled in. Artifacts in eye movement trajectory signals mainly include blink interference artifacts and gaze drift artifacts. Signal segments with gaze stability below 0.3° / s and pupil diameter abrupt changes exceeding 0.5mm are identified as artifacts and removed. After removal, adjacent valid data are interpolated to fill in the missing data. For EEG physiological signals, the smoothing process sets the sliding window size to 5 sampling points and performs a point-by-point moving average calculation on the EEG signals after artifact removal. For eye movement trajectory signals, the sliding window size is set to 3 sampling points and performs a point-by-point moving average calculation on the eye movement signals after artifact removal. During the moving average calculation, the signal data at each time-series sampling moment is retained.

5. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the S300 multimodal coupling processing, the comprehensive value of eye movement micro-features is obtained by extracting fixation point density, saccade amplitude, pupil dilation rate, fixation duration, saccade latency, and pupil diameter variation coefficient features from the time-aligned eye movement trajectory signal, and then calculating them by weighting after normalization. The comprehensive value of brain electrical rhythm features is obtained by performing wavelet decomposition on the time-aligned brain electrical physiological signal to extract alpha, theta, delta, and gamma wave related rhythm indicators, including alpha wave inhibition rate, theta wave absolute power, the theta wave to beta wave power ratio, delta wave relative power, and gamma wave event-related synchronization features.

6. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the S300 multimodal coupling process, the multimodal cognitive coupling algorithm adopts the following mathematical expression: in, Let be the cognitive coupling feature value at the t-th temporal sampling time; t is the index of the temporal sampling time; i is the index of the cognitive task type. Let be the comprehensive value of eye movement micro-features at the t-th temporal sampling time under the i-th cognitive task; The comprehensive value of EEG rhythm features at the t-th time sequence sampling time under the i-th cognitive task; These are the weighting coefficients for eye-tracking features; These are the weighting coefficients for EEG features; Let be the temporal decay factor at the t-th temporal sampling time under the i-th cognitive task.

7. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the training of the S400 discrimination model, the psychological anomaly dynamic discrimination model takes the temporal sequence composed of cognitive coupling feature values ​​at each temporal sampling time as input, corresponding to the number of feature samples in the entire cycle of the cognitive task; the model adopts a deep neural network structure based on temporal attention mechanism, including an input layer, multiple temporal coding layers and an anomaly discrimination output layer. The temporal coding layer is used to capture the dynamic pattern of feature changes over time during the cognitive process, and the anomaly discrimination output layer outputs the anomaly score at each sampling time. In the training stage, the time series of cognitive coupling characteristics of the normal college student group is used as the training set, and the reconstruction loss is used as the optimization goal. First, the input features are preprocessed based on the mean, extreme values of the cognitive coupling characteristics of the normal group and the physiological adaptation correction coefficient to eliminate the influence of group differences and dimensions. Then, the model parameters are updated iteratively until the reconstruction error converges to the preset threshold of 0.001, completing the model calibration and training. After training, the model can identify abnormal features deviating from the normal cognitive time series pattern by calculating the abnormal scores of the feature sequences of the samples to be tested.

8. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the output of the S500 screening result, the mathematical expression of the time series abnormal discrimination algorithm is as follows: in, The coefficient for detecting psychological abnormalities; The total number of temporal sampling points throughout the entire cognitive task cycle; Let be the cognitive coupling feature value of the college student subject under test at the t-th time-series sampling time, where t is the index of the time-series sampling time; Let be the mean of the cognitive coupling characteristics of a normal college student group at the t-th time-series sampling time. This represents the maximum value of the cognitive coupling characteristics of a normal university student population. This represents the minimum value of cognitive coupling characteristics in a normal university student population. Physiological adaptation correction coefficients for university students.

9. The intelligent screening method for abnormal psychological behaviors of university students according to claim 1, characterized in that, In the output of the S500 screening result, the judgment rules for the classification results of psychological abnormality screening are as follows: When the psychological abnormality discrimination coefficient M ≤ 0.3, it is judged that the psychological state is normal, and the screening result without obvious psychological abnormality tendency is output; When 0.3 < M ≤ 0.6, it is judged that there is a mild psychological abnormality tendency, and the screening result of "there is a mild psychological abnormality tendency, it is recommended to carry out daily psychological adjustment and attention" is output; When 0.6 < M ≤ 0.8, it is judged that there is a moderate psychological abnormality tendency, and the screening result of "there is a moderate psychological abnormality tendency, it is recommended to carry out professional psychological counseling and evaluation" is output; When M > 0.8, it is judged that there is a severe psychological abnormality tendency, and the screening result of "there is a severe psychological abnormality tendency, it is recommended to immediately carry out professional psychological intervention and medical diagnosis and treatment" is output. The above classification results are output in the form of a standardized report, including the cognitive coupling feature sequence of the subject, the abnormal discrimination coefficient and the corresponding classification suggestions, forming a complete psychological abnormality screening process.

Images