A Method and System for Extracting Neuroelectrophysiological Features Based on Personalized Audiovisual Paradigm

Abstract

This invention belongs to the field of neurophysiological detection technology and discloses a method and system for extracting neurophysiological features based on a personalized audiovisual paradigm. The method includes: presenting audiovisual stimuli based on a first preset set of spontaneously generated associations (SOAs) to perform behavioral assessment and obtain the subject's synchronous judgment probability curve; fitting the curve with a function model to calculate the individual temporal integration window (TIW), which characterizes the individual's temporal integration accuracy; adaptively generating a second set of SOAs containing the calculated TIW value based on the calculated TIW; presenting audiovisual stimuli based on the personalized second set of SOAs and simultaneously acquiring electroencephalogram (EEG) signals to extract neurophysiological features related to audiovisual integration. This invention precisely targets stimuli to the individual's perceptual critical zone, significantly improving the signal-to-noise ratio and individual discrimination of neurophysiological signals, and providing an effective tool for the objective and accurate assessment of perceptual abnormalities.

Description

Technical Field

[0001] This invention belongs to the field of neurophysiological detection technology, specifically relating to a method and system for extracting neurophysiological features based on a personalized audiovisual paradigm. Background Technology

[0002] Schizophrenia is a severe mental illness. Before the onset of typical positive symptoms (such as hallucinations and delusions), often in the early or high-risk stages of the illness, underlying perceptual processing abnormalities already exist. These abnormalities typically manifest as a decline in the ability to recognize basic sensory information such as sounds and images, and a decreased ability to judge the chronological order of events. These perceptual changes often precede or are independent of fluctuations in clinical symptoms, and are therefore considered neuropsychological markers with value for early identification and intervention.

[0003] In multisensory integration research, the Temporal Integration Window (TIW) is a core concept. When processing stimuli from different sensory channels (such as vision and hearing), the human brain has a tolerable range for temporal inconsistencies; this range is called the TIW. Within this window, even if there are slight time differences between visual and auditory stimuli, the brain tends to perceive them as a unified or synchronous event. Therefore, the size of the TIW reflects the precision of multisensory temporal integration: a larger TIW means a higher tolerance for temporal asynchrony, lower sensitivity, and poorer temporal processing precision.

[0004] The Time-to-Wait (TIW) of an individual can be measured using the classic Simultaneity Judgment Task (SJ task). In this task, researchers present participants with a series of audiovisual stimuli pairs with different temporal offsets (i.e., asynchronous stimulus onset (SOA)) and ask them to judge whether the stimuli are "synchronous" or "asynchronous." By statistically analyzing the proportion of "synchronous" judgments under different SOAs, a "synchronization rate - SOA" function curve can be plotted, and the individual's TIW can be calculated. Multiple studies and the inventors' previous data have shown that individuals with early-stage mental illness or at high risk have significantly higher TIWs than healthy controls, providing a reliable behavioral basis for identifying perceptual abnormalities.

[0005] Event-related potentials (ERPs), as an objective neurophysiological indicator, can accurately capture changes in brain electrical activity during sensory and cognitive processing with millisecond-level temporal resolution. In audiovisual integration tasks, ERP components related to early sensory processing (such as N1 / P2 waves) and those related to mid-to-late integration / decision-making exhibit quantifiable differences under different SOA conditions. ERP technology has advantages such as standardization, low cost, and good reproducibility, and has been used in numerous studies to objectively assess cognitive function, monitor treatment efficacy, and predict prognosis, possessing a solid foundation for clinical translation.

[0006] However, there are obvious limitations in the existing technology.

[0007] Existing technical solution A: Behavioral assessment with fixed SOA. Traditional methods use a fixed set of SOAs (e.g., 0, ±100, ±200, ..., ±600ms) to test all subjects. This "one-size-fits-all" approach ignores significant differences in TIW among individuals. For some individuals, this fixed set of SOAs may be completely off-limits to their perceptual "critical zone" (i.e., the region most sensitive to the transition from "synchronous" to "asynchronous"), resulting in an indistinct synchronicity rate curve, thus limiting the accuracy and discriminative power of individual TIW calculations.

[0008] Existing technical solution B: Non-individualized parallel assessment of behavioral science and ERP. While some studies have simultaneously collected EEG signals during fixed SOA tasks in an attempt to analyze ERP differences, the SOA settings are not optimized for individuals, resulting in insufficient sensitivity of ERP to changes in neural activity under "critical conditions." Furthermore, these solutions mostly focus on statistical analysis of group-average differences, lacking the ability to accurately discriminate at the individual level.

[0009] In the process of realizing this invention, the inventors discovered at least the following problems in the prior art:

[0010] (1) Non-individualized: Fixed SOA settings cannot accurately target the perceptual critical zone (near TIW) of each subject, resulting in low signal-to-noise ratio and insufficient discrimination of behavioral data and ERP signals.

[0011] (2) Lack of objective neurological support: relying solely on behavioral indicators (such as TIW) for judgment is easily affected by factors such as the subject's subjective reaction preferences and attention fluctuations.

[0012] (3) Low accuracy in identification: Although differences can be observed at the group level, there is a lot of overlap in data points of different groups (such as high-risk individuals and healthy controls) at the individual level, which limits its accuracy and reliability as an early identification tool.

[0013] (4.) Fragmented process: Behavioral testing, TIW calculation, ERP experiment design and data collection are usually separate steps, which are cumbersome and difficult to form a smooth, closed-loop "one-stop" assessment tool.

[0014] (5) Difficulty in clinical deployment: There is a lack of standardized software systems and low integration with EEG acquisition equipment, which is not conducive to standardized deployment and multi-center promotion in clinical environments.

[0015] The above description of the background technology is only for the purpose of facilitating a deeper understanding of the technical solution of the present invention (the technical means used, the technical problems solved, and the technical effects produced, etc.), and should not be regarded as an admission or in any form an implication that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0016] The present invention aims to at least partially solve the aforementioned technical problems. Therefore, the objective of the present invention is to provide a method and system for extracting neurophysiological features based on a personalized audiovisual paradigm.

[0017] The technical solution adopted in this invention is as follows:

[0018] A method for extracting neurophysiological features based on a personalized audiovisual paradigm includes the following steps:

[0019] S1: In the behavioral assessment phase, audiovisual stimuli based on a first pre-defined set of SOAs (Stimulus Onset Asynchrony) are presented to the participants, and the participants' judgments on whether the audiovisual stimuli are synchronized under each SOA are recorded to obtain a synchronization judgment probability curve. The first pre-defined set of SOAs includes multiple non-zero SOA values ​​and zero SOA values ​​to fully sample the participants' synchronous perception range.

[0020] S2: In the individualized parameter calculation stage, the synchronous judgment probability curve is fitted with a function model to calculate the Temporal Integration Window (TIW), which represents the subject's multi-sensory temporal integration ability.

[0021] S3: In the individualized paradigm configuration stage, based on the TIW, a second SOA set containing the TIW values ​​is generated. This second SOA set is specifically customized for subsequent neurophysiological signal acquisition.

[0022] S4: Neurophysiological feature extraction stage. Audiovisual stimuli based on the second SOA set are presented to the subject, and the subject's electroencephalogram (EEG) signals are collected simultaneously. Neurophysiological features related to audiovisual integration are extracted from the EEG signals.

[0023] Further, in step S2, the function model fitting preferably adopts a Gaussian function model. The standard deviation σ of the Gaussian function model is obtained by fitting, and this standard deviation is used as the TIW. The form of the Gaussian function can be: P(sync|SOA)=a*exp[-(SOA-b)² / (2*c²)], where the parameter c represents the TIW.

[0024] Furthermore, to ensure the reliability of the TIW calculation, after step S2 and before step S3, this method also includes a quality control step: verifying the goodness of fit R of the Gaussian function model. 2 If R 2 If the data falls below a preset threshold, the behavioral data is considered of poor quality, and the system prompts the operator to suggest that the participant repeat steps S1 and S2 until R... 2 The standard has been met.

[0025] Furthermore, in step S3, in order to efficiently capture key neural activities in ERP acquisition, the second SOA set preferably includes at least three types of SOA values: zero SOA value (0ms, as a benchmark for complete synchronization), critical SOA value based on the TIW setting (corresponding to the critical state of the brain oscillating between "synchronous-asynchronous", which is most likely to induce integration-related ERP changes), and a distal SOA value whose absolute value is much larger than the TIW (as a clear asynchronous reference).

[0026] In a preferred embodiment, the critical SOA values ​​are +TIW and -TIW; the remote SOA values ​​are +600ms and -600ms. Therefore, the second SOA set specifically comprises {0, +TIW, -TIW, +600ms, -600ms}. This set constitutes a minimum sufficient sampling of the "peak-critical-endpoint" range, ensuring that ERP acquisition covers the critical interval from fully synchronous to strongly asynchronous.

[0027] Furthermore, in step S4, the process of extracting neurophysiological features includes: performing standard preprocessing (such as filtering, segmentation, baseline correction, artifact removal, etc.) on the acquired raw EEG signal, and then extracting the amplitude or latency of specific components of event-related potentials (ERPs) (such as N1 / P2 / N2), the time-frequency power of the EEG signal, or whole-brain functional connectivity indicators calculated based on multi-lead signals, etc., under different SOA conditions.

[0028] Furthermore, this method may also include step S5: constructing a discriminative model. Using the behavioral indicators (such as TIW, PSS, etc.) calculated in step S2 and the neurophysiological features extracted in step S4 as input variables, a classifier model is trained to distinguish the perceptual states of different populations, such as logistic regression, support vector machine (SVM), random forest, etc. This joint behavioral-neurophysiological indicator can improve the accuracy and robustness of the discrimination.

[0029] The present invention also provides a system for implementing the above method, which is a neurophysiological feature extraction system based on a personalized audiovisual paradigm, comprising:

[0030] The behavioral task module has a built-in stimulus presentation program that presents audiovisual stimuli based on a first preset SOA set to the subjects and records the subjects' key press responses (synchronous / asynchronous) in real time to generate a synchronization judgment probability curve.

[0031] The individualized parameter calculation module connects to the output of the behavioral task module. Its core function is to receive synchronous judgment probability curve data and automatically fit it with a function model (e.g., Gaussian fitting) to calculate the subject's Personal Time Integration Window (TIW). This module can also integrate a goodness-of-fit test function to detect poor fit results (e.g., R-squared). 2 If the value is below the threshold, output a prompt message.

[0032] The individualized paradigm configuration module, which is connected to the individualized parameter calculation module, is used to receive the calculated TIW value and automatically generate a stimulus sequence file containing a second SOA set based on the preset rules.

[0033] The neurophysiological task module, connected to the individualized paradigm configuration module, is used to load and present audiovisual stimuli based on the second SOA set. Simultaneously, it sends trigger signals synchronized with each stimulus event to external EEG devices through a trigger port to ensure accurate segmentation and analysis of EEG signals.

[0034] Furthermore, the system may also include a data processing and discrimination module. This module is used to receive and automatically process the acquired EEG signals (performing preprocessing and feature extraction), and combine them with the Theory of Behavioral Weakness (TIW) using a pre-trained classifier model (such as SVM) to finally output an auxiliary discrimination result or risk score regarding the subject's perceptual state.

[0035] All of the above modules can be integrated into a unified, user-friendly software application (App), enabling one-stop operation from behavioral testing to ERP data collection and result reporting, greatly improving the convenience and standardization of clinical applications.

[0036] The beneficial effects of this invention are as follows:

[0037] This invention identifies an individual's TIW through behavioral pretesting and precisely positions the SOA stimuli collected by ERP to the individual's perceptual critical zone, greatly improving the sensitivity and distinguishability of ERP signals to individual differences.

[0038] This invention combines subjective behavioral judgment results (which are susceptible to response bias) with objective ERP neurophysiological indicators to form a behavior-ERP closed-loop joint indicator, thereby offsetting the bias of a single indicator and making the assessment results more comprehensive and reliable.

[0039] Because this invention collects more individual-specific neural signals, it is expected that compared with the traditional method of fixed SOA + single behavioral indicator, this method can significantly improve the sensitivity and specificity of identifying individuals with early-stage mental illness and other perceptual abnormalities, and reduce the overlap of indicators between healthy and at-risk populations.

[0040] The framework of this invention has good scalability. It can be easily integrated with more types of machine learning or deep learning models for discriminative analysis, and can also be combined with follow-up assessment functions for longitudinal monitoring of disease changes or evaluation of treatment effects. Attached Figure Description

[0041] Figure 1 A flowchart of a method provided in an embodiment of the present invention.

[0042] Figure 2 This is a conceptual diagram of an integrated software system provided for an embodiment of the present invention. Detailed Implementation

[0043] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0044] It should be understood that, and also noted, in the embodiments, the functions / actions may appear in a different order than those shown in the figures. For example, depending on the functions / actions involved, they may actually be performed substantially concurrently, or sometimes the two figures shown consecutively may be performed in reverse order.

[0045] This embodiment details a method for extracting neurophysiological features based on a personalized audiovisual paradigm and the complete process of its system.

[0046] Reference Figure 1This method mainly includes four steps S1-S4, and optionally includes a subsequent discriminant model application step S5.

[0047] Step S1: Behavioral Assessment Phase (SJ Task)

[0048] The goal of this stage is to quickly and accurately measure the individual TIW of the subject.

[0049] (1) Participant preparation and practice: Participants sat in a quiet, well-lit room, focusing on the center of a computer screen. They first underwent task practice to ensure they fully understood the task requirements. In the practice sessions, the presented audiovisual stimuli were either extremely synchronous (SOA=0ms) or extremely asynchronous (SOA=±600ms). Since these time intervals were subjectively distinguishable, there was a single correct answer. Participants were required to achieve an accuracy rate of 80% or higher before proceeding to the formal experiment.

[0050] (2) Stimulus parameters: The visual stimulus is a white circle located in the center of the screen, with a presentation time of 13.3ms. The auditory stimulus is a pure tone at 1000Hz, presented through headphones, with a duration of 13.3ms. It should be noted that the choice of presentation time is related to the screen refresh rate of the specific hardware device used. When using different hardware devices, the presentation time can be adjusted reasonably, and the duration of the auditory stimulus should be consistent with the presentation time.

[0051] (3) Formal Experiment: A series of audiovisual stimulus pairs were presented to the participants. The SOA of these stimulus pairs was pseudo-randomly drawn from a pre-defined, broad first SOA set. This set was {0, ±100, ±200, ±300, ±400, ±500, ±600ms}. Each SOA condition was presented 10 times, for a total of 130 trials. In each trial, the participants were required to determine whether the perceived audiovisual stimulus was "synchronous" or "asynchronous" by pressing a button.

[0052] (4.) Data output: The system records the number of "synchronization" judgments under each SOA condition and calculates the "synchronization" judgment probability, thus obtaining a raw data curve of "synchronization judgment probability - SOA".

[0053] Step S2: Individualized Parameter Calculation Stage

[0054] This stage is conducted automatically after the behavioral assessment is completed.

[0055] (1) Model Fitting: The system calls the built-in calculation module to fit the "Synchronization Judgment Probability - SOA" curve obtained in step S1 using a Gaussian function model. The Gaussian function equation is as follows:

[0056] P(sync|SOA)=a*exp[-(SOA-b)² / (2*c²)]

[0057] Where P(sync|SOA) is the probability of being considered synchronized given SOA; a is the peak amplitude of the fitted curve; b is the Point of Subjective Simultaneity (PSS), which is the time point at which the subject feels most synchronized; and c is the standard deviation of the Gaussian function, which is defined as the individual's Time Integration Window (TIW).

[0058] (2) Quality control: After the fitting is completed, the system immediately calculates the goodness of fit R. 2 Set a preset threshold of 0.4. If the calculated R... 2 A value <0.4 indicates poor quality of behavioral data or confused response patterns in the subject, resulting in unreliable fitting results. In this case, the system will prompt the operator to have the subject rest and repeat step S1 twice. If the fitting remains poor after multiple repetitions, this should be recorded, as it may indicate significant difficulties in task comprehension or audiovisual integration, and further clinical examination is recommended.

[0059] Step S3: Individualized Paradigm Configuration Stage

[0060] After successful TIW calculation, the system automatically configures the stimulus parameters for the subsequent ERP acquisition phase.

[0061] (1) Generate the second SOA set: Based on the TIW value calculated in step S2, the system generates a simplified but highly targeted second SOA set. This set is designed as {0,+TIW,-TIW,+600ms,-600ms}.

[0062] 0ms: As the peak condition for fully synchronized integration.

[0063] ±TIW: as the individualized perceptual threshold. At these SOA values, subjects are in a state of oscillation between "synchronous" and "asynchronous" judgments, and the brain's integrative processing is most active, which is expected to induce the most significant ERP differences related to integration.

[0064] ±600ms: This serves as a reference condition for extreme asynchrony. Under this condition, the neural activity of the visual and auditory channels is approximately decoupled, and can be used as a quasi-monosensory reference for comparison with ERP under 0ms or ±TIW conditions to highlight the multisensory integration effect.

[0065] (2) Generate stimulus sequence: Based on this second SOA set, the system generates the stimulus sequence file required for the ERP experiment, in which each SOA condition is repeated 60 times to ensure that ERP data with sufficient signal-to-noise ratio is collected.

[0066] Step S4: Neurophysiological Feature Extraction Stage

[0067] (1) Synchronous EEG Acquisition: Initiate the ERP acquisition task. The system sends a precise time stamp to the connected EEG acquisition device (e.g., a compatible 32-channel or 64-channel EEG machine) simultaneously with each audiovisual stimulus presented via a parallel port or USB trigger port. Subjects still need to complete the synchronous / asynchronous button judgment while viewing the stimulus.

[0068] (2) Data preprocessing: After acquisition, the raw EEG data is processed offline or online. The standard procedure includes: bandpass filtering (e.g., 0.1-30Hz), segmentation with the stimulus presentation time as the zero point (e.g., -200ms to 800ms), baseline correction (e.g., using data from -200ms to 0ms before stimulus), and artifact removal (which can be done by setting an amplitude threshold or using methods such as independent component analysis (ICA) to remove interference from eye movements, electromyography, etc.).

[0069] (3) Feature extraction: For the preprocessed data, ERP stacking and averaging were performed under each SOA condition (0, ±TIW, ±600ms). Then, key neurophysiological features were extracted, such as:

[0070] ERP components: measure the peak amplitude and latency of early sensory components (such as P1 / N1 in the occipital region and P2 / N2 in the central region).

[0071] Time-frequency analysis: Calculate the event-related synchronization / desynchronization (ERS / ERD) power changes in specific frequency bands (such as alpha, beta, and gamma waves) under different SOA conditions.

[0072] Functional connectivity: Analyze connectivity indicators such as phase synchronization between different brain regions.

[0073] Step S5: Constructing and applying the discriminant model

[0074] The indicators obtained from the previous steps are fused together to improve the discrimination performance.

[0075] (1.) Feature fusion: Integrating behavioral indicators (such as TIW, PSS, goodness of fit R) 2 The features are combined with various neurophysiological features extracted from step S4 (such as N1 amplitude, P2 latency, gamma band power, etc.) to form a high-dimensional feature vector.

[0076] (2) Model Training and Application: The feature vector is used to train a machine learning classifier. For example, in the inventors' previous research, data collected using the traditional fixed SOA paradigm was used, with demographic information, cognitive assessment indicators, and ERP features (P1 / N1, P2 / N2) as predictor variables. Six classifiers were trained, including logistic regression, support vector machine (SVM), k-nearest neighbors (KNN), decision tree, random forest, and gradient boosting. The results showed that the SVM model performed best, achieving a binary classification accuracy of 0.85 for high-risk clinical populations and healthy populations, with a sensitivity of 0.875 and a specificity of 0.8333.

[0077] By introducing more individual-specific ±TIW conditions, the extracted ERP features are expected to be more discriminative, thereby further improving the accuracy, sensitivity, and specificity of the classification model.

[0078] The present invention also provides a system for implementing the above method, comprising:

[0079] The behavioral task module has a built-in stimulus presentation program that presents audiovisual stimuli based on a first preset SOA set to the subjects and records the subjects' key press responses (synchronous / asynchronous) in real time to generate a synchronization judgment probability curve.

[0080] The individualized parameter calculation module connects to the output of the behavioral task module. Its core function is to receive synchronous judgment probability curve data and automatically fit it with a function model (e.g., Gaussian fitting) to calculate the subject's Personal Time Integration Window (TIW). This module can also integrate a goodness-of-fit test function to detect poor fit results (e.g., R-squared). 2 Output a prompt message if the value is below the threshold.

[0081] The individualized paradigm configuration module, which is connected to the individualized parameter calculation module, is used to receive the calculated TIW value and automatically generate a stimulus sequence file containing a second SOA set based on the TIW value according to preset rules.

[0082] The neurophysiological task module, connected to the individualized paradigm configuration module, is used to load and present audiovisual stimuli based on the second SOA set. Simultaneously, it sends trigger signals synchronized with each stimulus event to external EEG devices through a trigger port to ensure accurate segmentation and analysis of EEG signals.

[0083] The data processing and discrimination module receives and automatically processes the acquired EEG signals (performing preprocessing and feature extraction), and combines them with behavioral indicators (TIW) through a pre-trained classifier model (such as SVM) to finally output an auxiliary discrimination result or risk score about the subject's perceptual state.

[0084] All of the above modules can be integrated into a unified, user-friendly software application (App), see reference. Figure 2 The software system provided by this invention has the following modular structure:

[0085] Main Interface / Subject Management Module (Data and Permission Management Module): Used for logging in, managing subject information (number, examination date, etc.), encrypting and storing data, and supporting data association for follow-up examinations.

[0086] Behavioral Task Initiation Module: Clicking this will initiate the SJ task described in step S1. The interface is simple, displaying only stimuli and instructions.

[0087] TIW Calculation Module: This module runs automatically after the behavioral task concludes, performing the calculations in step S2. It operates in the background, and the user-visible output includes the calculated TIW and PSS values, as well as the goodness-of-fit R-value. 2 .

[0088] SOA Adaptive Configuration Module: Based on the TIW calculation results, it automatically executes step S3 to generate configuration files for ERP tasks. This process is transparent to end users.

[0089] ERP Task Startup Module: Clicking this module loads the newly generated configuration file and starts the ERP data acquisition task in step S4. This module is responsible for synchronizing with the EEG device's trigger.

[0090] Data Processing and Reporting Module (Data Processing and Judgment Module): After the task is completed, the automated data processing process can be started with one click to extract key features and call the pre-trained judgment model (as described in step S5) to finally generate a comprehensive report containing behavioral indicators, key ERP waveforms and risk assessment scores.

[0091] The core of this invention lies in the closed-loop concept of "behavioral pre-testing - individualized parameter calculation - adaptive neurophysiological acquisition". Its specific implementation details can have various variations, all of which fall within the protection scope of this invention.

[0092] Stimulus material substitution: Visual stimuli can be replaced by flashing dots, Gabor gratings, letters, or inverted T-shaped graphics instead of white circles; auditory stimuli can be replaced by bandpass noise, ticking sounds, or speech syllables instead of 1000Hz pure tones; stimulation duration can be adjusted within the range of 5-30ms.

[0093] SOA ensemble substitution: In the behavioral phase, the step size of SOA can be adjusted as needed, for example, using a finer step size of 50ms or 80ms, or a coarser step size of 120ms. In the ERP phase, in addition to ±TIW, sampling points can be expanded, such as adding ±(0.5×TIW) and ±(1.5×TIW) points, to more finely characterize changes in neural activity around the critical region.

[0094] Thresholding and Model Alternatives: Besides the standard deviation of the Gaussian function, TIW can also be defined as the SOA width corresponding to the synchronization rate dropping to 75% of the peak, or the full width at half maximum (FWHM) on both sides of the curve. Other applicable nonlinear functions can also be tried when fitting the model.

[0095] Alternatives to neurophysiological acquisition methods: The number of EEG electrodes can be selected based on research precision and cost, ranging from portable multiple electrodes to high-density 128- or 256-lead systems. Reference electrodes, sampling rates (500-2000 Hz), and artifact removal methods (such as ASR, SSP) can all employ other methods well-known to those skilled in the art. Furthermore, in addition to EEG, the closed-loop methodology of this invention is also applicable to other neuroimaging techniques with high temporal resolution, such as magnetoencephalography (MEG).

[0096] This invention is not limited to the above-described optional embodiments. Anyone can derive other various forms of products under the guidance of this invention. However, regardless of any changes made in their shape or structure, any technical solution that falls within the scope of the claims of this invention shall be protected by this invention.

Claims

1. A method for extracting neurophysiological features based on a personalized audiovisual paradigm, characterized in that, Includes the following steps: S1: In the behavioral assessment phase, the subject is presented with audiovisual stimuli based on the first preset SOA set, and the subject's judgment on whether the audiovisual stimuli are synchronized under each SOA is recorded to obtain the synchronization judgment probability curve; the first preset SOA set includes multiple non-zero SOA values ​​and zero SOA values. S2: In the individualized parameter calculation stage, the synchronous judgment probability curve is fitted with a function model to calculate the individual time integration window (TIW) that characterizes the subject's multi-sensory time integration ability. S3: Individualized paradigm configuration stage, based on the TIW, a second SOA set containing the TIW value is generated for subsequent neurophysiological signal acquisition; S4: Neurophysiological feature extraction stage. Audiovisual stimuli based on the second SOA set are presented to the subject, and the subject's EEG signals are collected simultaneously. Neurophysiological features related to audiovisual integration are extracted from the EEG signals. S5: Construct a discriminative model, using the TIW calculated in step S2 and the neurophysiological features extracted in step S4 as input variables, and train a classifier model to distinguish the perceptual states of different groups. In step S2, the function model is fitted using a Gaussian function model; the standard deviation of the Gaussian function model is obtained through fitting, and this standard deviation is used as the TIW. In step S3, the second SOA set includes at least three types of SOA values: zero SOA value, critical SOA value set based on the TIW, and a far SOA value whose absolute value is much larger than the TIW. The critical SOA values ​​are +TIW and -TIW; the remote SOA values ​​are +600ms and -600ms; the second SOA set is specifically {0, +TIW, -TIW, +600ms, -600ms}.

2. The method according to claim 1, characterized in that, After step S2 and before step S3, the following is also included: The goodness of fit R of the Gaussian function model is tested. 2 If R 2 If the value is below a preset threshold, the operator is prompted to repeat steps S1 and S2 until R... 2 Not lower than the preset threshold.

3. The method according to claim 1, characterized in that, In step S4, the extraction of neurophysiological features includes: preprocessing the acquired EEG signals, and then extracting specific component amplitudes or latencies of event-related potentials, time-frequency power of EEG signals, or whole-brain functional connectivity indicators under different SOA conditions.

4. A neurophysiological feature extraction system based on a personalized audiovisual paradigm, characterized in that, include: The behavioral task module is used to present audiovisual stimuli based on a first preset SOA set to the subjects and record the subjects' synchronous judgment results to generate synchronous judgment probability curves. The individualized parameter calculation module, connected to the behavioral task module, is used to fit the synchronous judgment probability curve with a function model and calculate the subject's personal time integration window (TIW). The individualized paradigm configuration module, connected to the individualized parameter calculation module, is used to generate a second SOA set containing the TIW values ​​based on the TIW; the second SOA set includes at least three types of SOA values: zero SOA value, critical SOA value set based on the TIW, and a far SOA value whose absolute value is much larger than the TIW. The critical SOA values ​​are +TIW and -TIW; the remote SOA values ​​are +600ms and -600ms; the second SOA set is specifically {0, +TIW, -TIW, +600ms, -600ms}. The neurophysiological task module, connected to the individualized paradigm configuration module, is used to present audiovisual stimuli based on the second SOA set to the subject and synchronize with the EEG device through a trigger port to initiate EEG signal acquisition. The data processing and discrimination module is used to receive and process the collected EEG signals, extract neurophysiological features, and output discrimination results based on the TIW and the extracted neurophysiological features through a preset classifier model. The individualized parameter calculation module is specifically used to: fit the synchronization judgment probability curve with a Gaussian function model, and use the standard deviation of the fitted Gaussian function model as the TIW.

5. The system according to claim 4, characterized in that, The individualized parameter calculation module is also specifically used to test the goodness of fit R of the model. 2 And in R 2 Output a prompt message when the value is below a preset threshold.

Images