Multi-modal data acquisition and analysis system for obtaining social cognition
Through a multimodal data acquisition and analysis system, combined with social cognitive stimulation presentation, EEG and functional near-infrared spectral imaging data, combined with physiological data, the problem of early screening and evaluation of high-function autism spectrum disorders is solved, and the problem of more Accurate and efficient diagnostic support.
Patent Information
- Application Number
- CN202411242934.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Children with high-function autism spectrum disorder are often misdiagnosed or ignored, and it is difficult for the existing technology to effectively conduct early screening and evaluation.
A multimodal data acquisition and analysis system was designed, including social cognitive stimulus presentation module, electroencephalogram acquisition module, functional near-infrared spectral imaging acquisition module and data analysis module. Through correlation analysis and machine learning algorithms, ASD is constructed Early screening evaluation model, combined with physiological data acquisition module to assist in analysis.
Improve the objectivity and accuracy of screening evaluation, develop high sensitivity and high specificity ASD screening diagnostic models, improve evaluation efficiency and data quality, and discover physiological and biochemical indicators related to core symptoms of ASD, promoting early stages Discover and diagnose.
Smart Images

Figure CN118924298A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of social cognition related data collection and analysis, and specifically relates to a multimodal data collection and analysis system for acquiring social cognition. Background Art
[0002] Autism Spectrum Disorder (ASD) is a type of congenital neurodevelopmental disorder with social interaction deficits and narrow behavioral patterns as its core features. Among the patients with ASD, some have normal or superior intelligence, which is called high-functioning ASD. High-functioning ASD is the most treatable part of the ASD population.
[0003] In clinical practice, children with high-functioning autism spectrum disorder are often misdiagnosed or neglected. Therefore, screening and assessment of high-functioning autism spectrum disorder is a key issue that needs to be addressed urgently. Summary of the invention
[0004] In view of this, the present invention provides a multimodal data collection and analysis system for acquiring social cognition, comprising: A social cognition stimulation presentation module, used to present videos, images or scenes related to social cognition to users, inducing brain responses related to social cognition; An EEG acquisition module, used to collect EEG signals generated by the user under the social cognitive stimulation, the EEG acquisition module comprising an EEG signal amplification circuit, an analog-to-digital conversion circuit and a wireless transmission module; A functional near-infrared spectral imaging acquisition module, used to collect functional near-infrared spectral signals generated by the user under the social cognitive stimulation, the functional near-infrared spectral imaging acquisition module comprising a light source, a photodetector and a data acquisition circuit; The data analysis module is used to perform correlation analysis and machine learning analysis on the collected EEG signals and functional near-infrared spectroscopy imaging data, respectively, to obtain analysis results of multimodal data of social cognition.
[0005] In particular, the system also includes a physiological data acquisition module for synchronously acquiring the user's eye movements, facial expressions, and voice as behavioral data to assist in the analysis and interpretation of the multimodal data of social cognition.
[0006] In particular, the correlation analysis of the collected EEG signals and functional near-infrared spectral imaging data specifically includes: preprocessing the EEG signal data, extracting the time series of multiple collections, and convolving with the blood oxygen processing function to obtain a feature matrix X; preprocessing the functional near-infrared spectral imaging data to extract the feature matrix Y of blood oxygen changes; using correlation analysis to calculate the cross-correlation between the two feature matrices X and Y to obtain a correlation projection vector and a correlation coefficient; calculating the mean and standard deviation of the correlation coefficient, and calculating the confidence space of the correlation analysis result.
[0007] In particular, the blood oxygen processing function divides the blood oxygen processing into two models: oxygenated hemoglobin and deoxygenated hemoglobin, which are implemented by the following expression:
[0008] R i ( t ) represents the value of the blood oxygen processing function at time t; when i=1, it corresponds to the blood oxygen processing function model of oxygenated hemoglobin; when i=2, it corresponds to the blood oxygen processing function model of deoxygenated hemoglobin, t is the time independent variable; exp represents an exponential function with e as the base, and C represents a constant. In the blood oxygen processing function, the value of C is 8log2; When i=1, The time constant representing the rapid increase in oxygenated hemoglobin concentration; The time constant that represents the slow decay of oxygenated hemoglobin concentration; w 1,1 represents the width constant when the concentration of oxygenated hemoglobin increases rapidly; w 1,2 Indicates the width constant when the concentration of oxygenated hemoglobin decays slowly; when i=2, represents the time constant for the rapid decrease in deoxyhemoglobin concentration, represents the time constant for the slow recovery of deoxyhemoglobin concentration, w 2,1 represents the width constant when the concentration of deoxyhemoglobin decreases rapidly; w 2,2 represents the width constant when the concentration of deoxyhemoglobin recovers slowly; r i Represents the response intensity of oxygenated hemoglobin relative to deoxygenated hemoglobin.
[0009] In particular, the functional near-infrared spectroscopy imaging data is preprocessed to extract the characteristic matrix Y of blood oxygen changes, including: According to the near-infrared light intensity file, the modified Beer-Lambert law is used to calculate the time series data of the change of oxygenated hemoglobin concentration; the time series data of the change of oxygenated hemoglobin concentration is preprocessed by detrending and moving average filtering to obtain a smooth oxygenated hemoglobin concentration change curve; The preprocessed time series data of oxygenated hemoglobin concentration are composed into a feature matrix Y.
[0010] In particular, the machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data includes: concatenating the features of the EEG signals and functional near-infrared spectroscopy imaging data into a high-dimensional feature vector, dividing the user data into a training set and a test set, and training the SVM classifier model on the training set; using the trained SVM classifier model to perform classification prediction on the test set, obtain the probability that each sample in the test set belongs to the abnormal class, and set the output confidence probability threshold of the SVM.
[0011] In particular, the correlation analysis and machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data respectively further include: establishing a fusion confidence model of correlation analysis calculation and SVM classifier according to the confidence space of correlation analysis and the confidence space of machine learning analysis, which is realized by the following expression:
[0012] in, represents the fusion confidence model, Confidence model for correlation analysis calculation results, Confidence model of the classification results of the classifier; is the correlation coefficient of the jth sample, is the correlation coefficient The average value of is the correlation coefficient The standard deviation of is the confidence weight coefficient of the correlation analysis calculation part; is the abnormal probability output of the SVM classifier for the jth sample, is the average value of the probability output of the abnormal class, is the standard deviation of the anomaly probability output, is the weight coefficient of the SVM classifier.
[0013] In particular, the social cognitive stimulation presentation module includes a display screen and a speaker, and the social cognitive stimulation includes text, images, animation or voice.
[0014] Beneficial effects: Through the technical solution of the present invention: 1. Ability to establish an ASD early screening and evaluation model based on multimodal physiological, psychological and behavioral digital phenotypes to improve the objectivity and accuracy of screening and evaluation; 2. Use machine learning algorithms to train an ASD screening and diagnosis model with high sensitivity and specificity to provide auxiliary decision support for clinical diagnosis; 3. Develop an ASD digital assessment system that integrates electronic behavioral scales, digital neuropsychological tests, and physiological index collection to improve assessment efficiency and data quality; 4. Through brain-computer oxygen synchronization detection technology, objective physiological and biochemical indicators related to the core symptoms of ASD are discovered as a supplementary basis for ASD diagnosis; 5. Establish a sample library of ASD and healthy controls covering different age groups and intelligence levels to provide strong support for the training and verification of ASD screening models; 6. Promote the application of ASD digital screening and assessment system in clinical frontline to improve the early detection rate of ASD and reduce the burden on the families of children with the disease and society. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 Schematic diagram of the multimodal data collection and analysis system for acquiring social cognition proposed in the present invention. DETAILED DESCRIPTION
[0016] The present invention is described in detail below with reference to the accompanying drawings and embodiments.
[0017] The present invention provides a multimodal data collection and analysis system for acquiring social cognition, such as Figure 1 As shown, the system includes: The social cognition stimulus presentation module is used to present videos, images or scenes related to social cognition to users and induce brain responses related to social cognition; the main function of this module is to present visual stimulus materials related to social cognition to users, including videos, images or scenes. Videos can be short films depicting interpersonal interaction scenarios, such as two people talking or arguing. Images can be static pictures showing various facial expressions, such as happiness, sadness, anger, etc. Scenes can be simulated social environments in real life, such as offices, restaurants, parties, etc. The design of these stimulus materials needs to take into account different components of social cognition, such as emotion recognition, psychological theory reasoning, self-reference, etc. The presentation method can be full-screen playback, embedded playback, or virtual reality. This module needs to match the experimental design, be able to present stimuli in a predetermined order, and record metadata such as presentation time.
[0018] The EEG acquisition module is used to collect the EEG signals generated by the user under the social cognitive stimulation. The EEG acquisition module includes an EEG signal amplification circuit, an analog-to-digital conversion circuit and a wireless transmission module. The function of this module is to collect the EEG signals generated by the user's brain when the user watches the social cognitive stimulation. The EEG signal amplification circuit amplifies the weak EEG signal for subsequent processing. The analog-to-digital conversion circuit converts the amplified analog EEG signal into a digital signal for computer storage and analysis. The wireless transmission module transmits the digitized EEG data to the storage or analysis terminal wirelessly. This module needs to be connected to the EEG electrode array worn by the user, which usually includes 32 or 64 electrodes. EEG acquisition needs to be carried out in a shielded room or an electromagnetic shielding environment to avoid external interference; the collected EEG data needs to match the timestamp of the stimulus presentation for subsequent analysis.
[0019] The functional near-infrared spectral imaging acquisition module is used to collect the functional near-infrared spectral signals generated by the user under the social cognitive stimulation. The functional near-infrared spectral imaging acquisition module includes a light source, a photodetector and a data acquisition circuit; the function of this module is to collect the blood oxygen saturation change signal of the cerebral cortex area when the user watches the social cognitive stimulation. The light source is usually composed of an LED or a laser that emits near-infrared light, and irradiates the near-infrared light to the user's skull. The photodetector array detects the near-infrared light reflected or transmitted after passing through the skull and records the change in light intensity. The data acquisition circuit amplifies, converts and digitizes the original signal of the photodetector. This module needs to be connected to the light source-detector array on the user's head, generally including 16 to 52 measurement channels. Similar to EEG acquisition, functional near-infrared spectral imaging acquisition also needs to be carried out in an optically shielded environment to avoid interference from external light sources. The collected functional near-infrared spectral imaging data needs to be synchronized with the stimulus presentation time to prepare for subsequent analysis.
[0020] The physiological data collection module is used to synchronously collect the user's eye movements, facial expressions, and voice as behavioral data to assist in the analysis and interpretation of the multimodal data of social cognition.
[0021] The data analysis module is used to perform correlation analysis and machine learning analysis on the collected EEG signals and functional near-infrared spectroscopy imaging data, respectively, to obtain analysis results of multimodal data of social cognition.
[0022] The correlation analysis of the collected EEG signals and functional near-infrared spectral imaging data specifically includes: preprocessing the EEG signal data, extracting the time series of multiple collections, and convolving with the blood oxygen processing function to obtain a feature matrix X; preprocessing the functional near-infrared spectral imaging data to extract the feature matrix Y of blood oxygen changes; using correlation analysis to calculate the cross-correlation between the two feature matrices X and Y to obtain a correlation projection vector and a correlation coefficient; calculating the mean value and standard deviation of the correlation coefficient, and calculating the confidence space of the correlation analysis result.
[0023] Preprocessing the EEG signal specifically includes: importing the original EEG data obtained by the EEG acquisition module; Preprocess the raw data, such as filtering and removing artifacts; extract the EEG segments before and after the stimulus presentation in segments and calculate the event-related potential (ERP); select the time window and electrode position of interest from the relevant potential waveform, such as the frontocentral area Cz, FC1, and FC2; calculate the amplitude modulation (AMs) value of the relevant potential waveform in the selected time window; use the amplitude modulation value as the feature input of the EEG signal for subsequent classification, regression, and other analyses The blood oxygen processing function divides the blood oxygen processing into two models: oxygenated hemoglobin and deoxygenated hemoglobin, which are implemented by the following expression:
[0024] R i ( t ) represents the value of the blood oxygen processing function at time t; when i=1, it corresponds to the blood oxygen processing function model of oxygenated hemoglobin; when i=2, it corresponds to the blood oxygen processing function model of deoxygenated hemoglobin; t is the time independent variable; exp represents an exponential function with e as the base, and C represents a constant. In the blood oxygen processing function, the value of C is 8log2; When i=1, The time constant representing the rapid increase in oxygenated hemoglobin concentration; The time constant that represents the slow decay of oxygenated hemoglobin concentration; w 1,1 represents the width constant when the concentration of oxygenated hemoglobin increases rapidly; w 1,2 Indicates the width constant when the concentration of oxygenated hemoglobin decays slowly; when i=2, represents the time constant for the rapid decrease in deoxyhemoglobin concentration, represents the time constant for the slow recovery of deoxyhemoglobin concentration, w 2,1 represents the width constant when the concentration of deoxyhemoglobin decreases rapidly; w2,2 represents the width constant when the concentration of deoxyhemoglobin recovers slowly; r i Represents the response intensity of oxygenated hemoglobin relative to deoxygenated hemoglobin.
[0025] Preprocess the functional near-infrared spectroscopy imaging data and extract the characteristic matrix Y of blood oxygen changes, including: According to the near-infrared light intensity file, the modified Beer-Lambert law is used to calculate the time series data of the change of oxygenated hemoglobin concentration; the time series data of the change of oxygenated hemoglobin concentration is preprocessed by detrending and moving average filtering to obtain a smooth oxygenated hemoglobin concentration change curve; The preprocessed time series data of oxygenated hemoglobin concentration are composed into a feature matrix Y.
[0026] The machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data includes: concatenating the features of the EEG signals and functional near-infrared spectroscopy imaging data into a high-dimensional feature vector, dividing the user data into a training set and a test set, and training the SVM classifier model on the training set; using the trained SVM classifier model to perform classification prediction on the test set, obtain the probability that each sample in the test set belongs to the abnormal class, and set the output confidence probability threshold of the SVM.
[0027] The machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data includes: splicing the features of the EEG signals and functional near-infrared spectroscopy imaging data into a high-dimensional feature vector, and performing pre-splicing data pre-processing, pre-splicing feature extraction, pre-splicing time alignment, and finally feature-level fusion before splicing, including: The EEG signals were subjected to standard preprocessing before splicing, including filtering, artifact removal, independent component analysis (ICA), etc., to obtain clean EEG signal data.
[0028] Functional near-infrared spectral imaging data need to be preprocessed before stitching, such as motion artifact correction, baseline correction, and convolution blur kernel, to obtain a concentration change curve reflecting neural activity.
[0029] The feature extraction before splicing specifically includes: calculating the power spectral density (PSD) of different frequency bands (δ, θ, α, β, γ) of the EEG signal, or performing wavelet transform to obtain time-frequency features.
[0030] The peak value, peak delay, area under the curve and other features of each channel are extracted from the functional near-infrared spectral imaging data.
[0031] The time alignment operation before stitching requires alignment of the time series of EEG signals and functional near-infrared spectroscopy imaging data because there is a certain delay (several seconds) in the change of blood oxygen concentration relative to neural activity.
[0032] The optimal time delay can be estimated using methods such as cross-correlation or phase, and the functional near-infrared spectroscopy imaging data can be shifted accordingly.
[0033] The feature-level fusion operation includes concatenating the aligned EEG signal and functional near-infrared spectroscopy imaging data features into a high-dimensional feature vector. Principal component analysis (PCA), kernel techniques, and other methods can be used for feature extraction and dimensionality reduction.
[0034] The correlation analysis and machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data also include: establishing a fusion confidence model of correlation analysis calculation and SVM classifier according to the confidence space of correlation analysis and the confidence space of machine learning analysis, which is realized by the following expression:
[0035] in, represents the fusion confidence model, Confidence model for correlation analysis calculation results, Confidence model of the classification results of the classifier; is the correlation coefficient of the jth sample, is the correlation coefficient The average value of is the correlation coefficient The standard deviation of is the confidence weight coefficient of the correlation analysis calculation part; is the abnormal probability output of the SVM classifier for the jth sample, is the average value of the probability output of the abnormal class, is the standard deviation of the anomaly probability output, is the weight coefficient of the SVM classifier.
[0036] In this embodiment, the fusion confidence model is used to improve the accuracy of the evaluation, which can provide a quantitative evaluation result for the user's social cognitive ability and help to more accurately judge the user's social cognitive level; enhance the flexibility of the evaluation. Through the confidence information of the evaluation results, the evaluation strategy can be flexibly adjusted according to the specific situation, such as taking further inspection measures when the confidence is low, so as to improve the comprehensiveness of the evaluation; optimize the evaluation process. The evaluation model can evaluate the results of different stages in real time, identify the weak links in the evaluation process, and help optimize the evaluation process and improve work efficiency; support personalized evaluation: different users have different social cognitive characteristics. The fusion confidence model can provide differentiated evaluation suggestions based on individual characteristics to achieve personalized services. Promote the continuous optimization of the model. The confidence information of the evaluation results can be used to feedback model training, help to continuously optimize the social cognitive state evaluation model, and improve its generalization performance. Enhance the application value. The confidence of the evaluation results intuitively reflects the credibility of the evaluation results, which helps to enhance the user's trust in the system evaluation results and promote the promotion of social cognitive state evaluation technology in practical applications.
[0037] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
[0038] It is obvious to those skilled in the art that the embodiments of the present invention are not limited to the details of the above exemplary embodiments, and that the embodiments of the present invention can be implemented in other specific forms without departing from the spirit or basic features of the embodiments of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the embodiments of the present invention is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the embodiments of the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units, modules or devices stated in the system, device or terminal claims can also be implemented by the same unit, module or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.
[0039] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the embodiments of the present invention and are not intended to limit them. Although the embodiments of the present invention have been described in detail with reference to the above preferred implementation modes, those skilled in the art should understand that the technical solutions of the embodiments of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A multimodal data collection and analysis system for acquiring social cognition, characterized in that: include: A social cognition stimulation presentation module, used to present videos, images or scenes related to social cognition to users, inducing brain responses related to social cognition; An EEG acquisition module, used to collect EEG signals generated by the user under the social cognitive stimulation, the EEG acquisition module comprising an EEG signal amplification circuit, an analog-to-digital conversion circuit and a wireless transmission module; A functional near-infrared spectral imaging acquisition module, used to collect functional near-infrared spectral signals generated by the user under the social cognitive stimulation, the functional near-infrared spectral imaging acquisition module comprising a light source, a photodetector and a data acquisition circuit; The data analysis module is used to perform correlation analysis and machine learning analysis on the collected EEG signals and functional near-infrared spectroscopy imaging data, respectively, to obtain analysis results of multimodal data of social cognition.
2. The multimodal data collection and analysis system for acquiring social cognition according to claim 1, characterized in that: The system also includes a physiological data acquisition module for synchronously acquiring the user's eye movements, facial expressions and voice as behavioral data to assist in the analysis and interpretation of the multimodal data of social cognition.
3. The multimodal data collection and analysis system for acquiring social cognition according to claim 1, characterized in that: The correlation analysis of the collected EEG signals and functional near-infrared spectral imaging data specifically includes: preprocessing the EEG signal data, extracting the time series of multiple collections, and convolving with the blood oxygen processing function to obtain a feature matrix X; preprocessing the functional near-infrared spectral imaging data to extract the feature matrix Y of blood oxygen changes; using correlation analysis to calculate the cross-correlation between the two feature matrices X and Y to obtain a correlation projection vector and a correlation coefficient; calculating the mean value and standard deviation of the correlation coefficient, and calculating the confidence space of the correlation analysis result.
4. The multimodal data collection and analysis system for acquiring social cognition according to claim 3, characterized in that: The blood oxygen processing function divides the blood oxygen processing into two models: oxygenated hemoglobin and deoxygenated hemoglobin, which are implemented by the following expression: R i ( t ) represents the value of the blood oxygen processing function at time t; When i=1, it corresponds to the blood oxygen processing function model of oxygenated hemoglobin; when i=2, it corresponds to the blood oxygen processing function model of deoxyhemoglobin. t is the time independent variable; exp represents an exponential function with e as the base, and C represents a constant. In the blood oxygen processing function, the value of C is 8log2; When i=1, The time constant representing the rapid increase in oxygenated hemoglobin concentration; The time constant that represents the slow decay of oxygenated hemoglobin concentration; w 1,1 represents the width constant when the concentration of oxygenated hemoglobin increases rapidly; w 1,2 Indicates the width constant when the concentration of oxygenated hemoglobin decays slowly; when i=2, represents the time constant for the rapid decrease in deoxyhemoglobin concentration, represents the time constant for the slow recovery of deoxyhemoglobin concentration, w 2,1 represents the width constant when the concentration of deoxyhemoglobin decreases rapidly; w 2,2 represents the width constant when the concentration of deoxyhemoglobin recovers slowly; r i Represents the response intensity of oxygenated hemoglobin relative to deoxygenated hemoglobin.
5. The multimodal data collection and analysis system for acquiring social cognition according to claim 4, characterized in that: Preprocess the functional near-infrared spectroscopy imaging data and extract the characteristic matrix Y of blood oxygen changes, including: According to the near-infrared light intensity file, the modified Beer-Lambert law is used to calculate the time series data of the change of oxygenated hemoglobin concentration; the time series data of the change of oxygenated hemoglobin concentration is preprocessed by detrending and moving average filtering to obtain a smooth oxygenated hemoglobin concentration change curve; The preprocessed time series data of oxygenated hemoglobin concentration are composed into a feature matrix Y.
6. The multimodal data collection and analysis system for acquiring social cognition according to claim 1, characterized in that: The machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data includes: concatenating the features of the EEG signals and functional near-infrared spectroscopy imaging data into a high-dimensional feature vector, dividing the user data into a training set and a test set, and training the SVM classifier model on the training set; using the trained SVM classifier model to perform classification prediction on the test set, obtain the probability that each sample in the test set belongs to the abnormal class, and set the output confidence probability threshold of the SVM.
7. The multimodal data collection and analysis system for acquiring social cognition according to claim 3 or 6, characterized in that: The correlation analysis and machine learning analysis of the collected EEG signals and functional near-infrared spectroscopy imaging data also include: establishing a fusion confidence model of correlation analysis calculation and SVM classifier according to the confidence space of correlation analysis and the confidence space of machine learning analysis, which is realized by the following expression: in, represents the fusion confidence model, Confidence model for correlation analysis calculation results, Confidence model of the classification results of the classifier; is the correlation coefficient of the jth sample, is the correlation coefficient The average value of is the correlation coefficient The standard deviation of is the confidence weight coefficient of the correlation analysis calculation part; is the abnormal probability output of the SVM classifier for the jth sample, is the average value of the probability output of the abnormal class, is the standard deviation of the anomaly probability output, is the weight coefficient of the SVM classifier.
8. The multimodal data collection and analysis system for acquiring social cognition according to claim 7, characterized in that: The social cognitive stimulation presentation module includes a display screen and a speaker, and the social cognitive stimulation includes text, images, animation or voice.
Citation Information
Patent Citations
Cognitive enhancement training system and method
CN116530938A
Brain cognitive function rehabilitation state analysis method based on multi-modal data
CN116616704A
System for assisting in screening mild cognitive impairment
CN116999027A
Multi-modal emotion recognition method and system based on confidence fusion
CN117591967A
Multi-parameter social cognition assessment method, system and product based on virtual reality
CN118568576A
