System platform for psychological assessment and emotion feedback

Abstract

The invention provides a system platform for psychological assessment and emotion feedback. The system platform comprises a multi-source data acquisition module, an emotion data analysis module, a psychological state assessment module, an intervention scheme generation module, an emotion change feedback module, a scheme optimization module and a data security protection module. The emotion recognition accuracy is improved by synchronously acquiring physiological-behavior-environment data and combining with a dynamic feature fusion method, a personalized intervention strategy is generated based on a reinforcement learning algorithm, real-time feedback is realized, and a data security protection system is constructed by adopting federal learning and homomorphic encryption technologies, so that the data security is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of emotion recognition, and in particular to a system platform for psychological assessment and emotion feedback. Background Art

[0002] At present, mental health issues have become a major challenge in the field of global public health. However, traditional mental health services have problems such as uneven resource distribution and low diagnostic efficiency. With the development of wearable devices and artificial intelligence technology, psychological assessment and intervention systems based on physiological signals have gradually become a research hotspot.

[0003] Existing technologies mainly rely on scale questionnaires and unimodal physiological signal analysis for psychological assessment. They lack real-time physiological data calibration and are difficult to dynamically adjust the intensity of intervention. In addition, existing systems mostly use static assessment models that cannot adapt to individual differences, resulting in large deviations in assessment results. They are also unable to make dynamic adjustments based on the user's real-time status, and the data security protection mechanism is weak, posing a risk of privacy leakage. Therefore, it is very necessary to design a system platform for psychological assessment and emotional feedback. Summary of the invention

[0004] The purpose of the present invention is to provide a system platform for psychological assessment and emotional feedback, which provides real-time status feedback of users through physiological-behavioral-environmental data combined with dynamic feature fusion and reinforcement learning algorithm to improve the accuracy of emotion recognition and data security.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A system platform for psychological assessment and emotional feedback, including: a multi-source data acquisition module, an emotional data analysis module, a psychological state assessment module, an intervention program generation module, an emotional change feedback module, a program optimization module and a data security protection module;

[0007] The multi-source data acquisition module obtains the original data set by performing signal denoising and format standardization on the collected physiological signals, environmental parameters and behavioral characteristics;

[0008] The sentiment data analysis module obtains a fused feature vector by performing key feature separation, data alignment, and abnormal data removal on the original data set;

[0009] The psychological state assessment module predicts psychological activities based on the fusion feature vector through the intelligent network model to obtain comprehensive assessment results;

[0010] The intervention plan generation module generates an initial plan based on the comprehensive evaluation results through a reinforcement learning algorithm and conducts simulation rehearsals to obtain an intervention strategy package;

[0011] The emotional change feedback module obtains the emotion regulation efficacy index based on the intervention strategy package through multimodal execution and real-time monitoring;

[0012] The scheme optimization module conducts effect evaluation and parameter iteration according to the emotion regulation efficacy index to obtain the final scheme;

[0013] The data security protection module encrypts the data of the final scheme through federated learning and homomorphic encryption technology.

[0014] Optionally, the multi-source data acquisition module includes: a biosensing unit, an environmental perception unit, an audiovisual acquisition unit, and a data preprocessing unit;

[0015] The biosensing unit conducts data analysis on the collected data through a dual-threshold QRS wave detection algorithm and Kalman filtering operation to obtain physiological signals;

[0016] The environmental perception unit conducts data analysis on the collected data through a thermal comfort model and spectrum analysis to obtain environmental parameters;

[0017] The audiovisual acquisition unit collects data through a binocular camera and a microphone array to obtain behavioral characteristics;

[0018] The data preprocessing unit conducts wavelet denoising and standardization alignment operations on the physiological signals, environmental parameters, and behavioral characteristics respectively to obtain the original data set.

[0019] Optionally, the emotion data analysis module includes: a feature extraction unit, a dynamic fusion unit, and a quality calibration unit;

[0020] The feature extraction unit conducts feature extraction on the original data set through a Poincaré scatter plot and baseline calibration technology to obtain emotion features;

[0021] The dynamic fusion unit conducts importance evaluation on the emotion features through a dynamic time warping algorithm and dynamic weight allocation technology to obtain normalized features;

[0022] The quality calibration unit conducts quality detection on the normalized features through an adaptive filtering technology to obtain a fused feature vector.

[0023] Optionally, the mental state assessment module includes: an emotion recognition unit, a comprehensive evaluation unit, and a baseline calibration unit;

[0024] The emotion recognition unit conducts classification output on the fused feature vector through time series modeling and cross-entropy loss function to obtain an emotion score;

[0025] The comprehensive evaluation unit calculates the mental health risk index based on the emotion score through the Markov chain Monte Carlo method;

[0026] The baseline calibration unit corrects the error of the mental health risk index through a Gaussian mixture model to obtain a comprehensive evaluation result.

[0027] Optionally, the intervention plan generation module includes: a strategy calculation unit, a digital twin unit, and a risk assessment unit;

[0028] The strategy calculation unit performs multi-objective optimization on the comprehensive evaluation result through a Pareto front objective function to obtain an initial plan;

[0029] The digital twin unit obtains a simulated rehearsal effect by performing finite element analysis on the initial plan and solving Maxwell's equations;

[0030] The risk assessment unit screens strategies for the simulated rehearsal effect through a logistic regression model and an adverse event database to obtain an intervention strategy package.

[0031] Optionally, the emotional change feedback module includes: a multi-modal execution unit and a real-time monitoring unit;

[0032] The multi-modal execution unit performs emotional regulation through neuromodulation tDCS, light therapy, and acoustic wave intervention according to the intervention strategy package to obtain an emotional regulation result;

[0033] The real-time monitoring unit calculates the change rate of the emotional regulation result through a linear mixed effects model to obtain an emotional regulation efficacy index.

[0034] Optionally, the plan optimization module includes: an effect evaluation unit and a parameter iteration unit;

[0035] The effect evaluation unit performs a statistical test on the emotional regulation efficacy index through HRV analysis and the paired-sample t-test method to obtain an effectiveness index;

[0036] The parameter iteration unit performs evolutionary operations and strategy updates through a genetic algorithm based on the effectiveness index to obtain a final plan.

[0037] Optionally, the data security protection module includes: a privacy protection unit and an access control unit;

[0038] The privacy protection unit encrypts the data of the final plan through a distributed deep learning training library and an encryption algorithm;

[0039] The access control unit performs anomaly detection on system access personnel through a zero-trust policy and a blockchain evidence storage method.

[0040] According to the specific embodiments provided by the present invention, the following technical effects are disclosed: The system platform for psychological assessment and emotional feedback provided by the present invention includes: a multi-source data acquisition module, an emotional data analysis module, a psychological state assessment module, an intervention plan generation module, an emotional change feedback module, a plan optimization module, and a data security protection module. By synchronously acquiring physiological-behavioral-environmental data and combining the method of dynamic feature fusion, the accuracy of emotion recognition is improved. Based on the reinforcement learning algorithm, personalized intervention strategies are generated, real-time feedback is achieved, and a data security protection system is constructed using federated learning and homomorphic encryption technology, thereby ensuring data security. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0042] Figure 1 It is the structural diagram of the system platform of the present invention;

[0043] Figure 2 It is the internal structural diagram of the system platform of the present invention;

[0044] Figure 3 It is the work flow chart of the multi-source data acquisition module in the embodiment of the present invention;

[0045] Figure 4 It is the work flow chart of the emotional data analysis module in the embodiment of the present invention;

[0046] Figure 5 It is the work flow chart of the psychological state assessment module in the embodiment of the present invention;

[0047] Figure 6 It is the work flow chart of the intervention plan generation module in the embodiment of the present invention;

[0048] Figure 7 It is the work flow chart of the emotional change feedback module in the embodiment of the present invention;

[0049] Figure 8 It is the work flow chart of the plan optimization module in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a system platform for psychological assessment and emotional feedback, including: a multi-source data acquisition module, an emotional data analysis module, a psychological state assessment module, an intervention plan generation module, an emotional change feedback module, a plan optimization module and a data security protection module;

[0053] The multi-source data acquisition module obtains the original data set by performing signal denoising and format standardization on the collected physiological signals, environmental parameters and behavioral characteristics;

[0054] The sentiment data analysis module obtains a fused feature vector by performing key feature separation, data alignment, and abnormal data removal on the original data set;

[0055] The psychological state assessment module predicts psychological activities based on the fusion feature vector through the intelligent network model to obtain comprehensive assessment results;

[0056] The intervention plan generation module generates an initial plan based on the comprehensive evaluation results through a reinforcement learning algorithm and conducts simulation rehearsals to obtain an intervention strategy package;

[0057] The emotion change feedback module is based on the intervention strategy package and obtains the emotion regulation efficacy index through multimodal execution and real-time monitoring;

[0058] The solution optimization module evaluates the effect and iterates the parameters according to the emotion regulation effectiveness index to obtain the final solution;

[0059] The data security protection module encrypts the data of the final solution through federated learning and homomorphic encryption technology.

[0060] Specifically, the multi-source data acquisition module includes: a biosensor unit, an environmental perception unit, an audio-visual acquisition unit, and a data preprocessing unit; the workflow of this module is as follows: Figure 3 shown.

[0061] The biosensor unit performs data analysis on the collected data through a dual-threshold QRS wave detection algorithm and Kalman filter operation to obtain physiological signals;

[0062] In this embodiment, the ADS1294 analog front-end chip is used to collect three-lead electrocardiogram signals at a sampling rate of 250 Hz. Then, the Pan-Tompkins algorithm is adopted for high-pass filtering, differential operation, and moving window integration operations to eliminate baseline drift, enhance QRS wave characteristics, and locate the R-wave peak, making the RR interval sequence error ≤ 3 ms. Next, time-domain analysis of heart rate variability (HRV) is performed according to the standard deviation (SDNN) and root mean square successive difference (RMSSD), and the LF and HF powers are calculated through the Lomb-Scargle periodogram to achieve the frequency-domain analysis of HRV. Finally, the constant false alarm rate (CFAR) detection algorithm and Kalman filter are used for noise removal and signal smoothing processing, so as to capture the user's physiological state in real time and accurately and reflect it as a digital signal.

[0063] The environmental perception unit performs data analysis on the collected data through a thermal comfort model and spectral analysis to obtain environmental parameters; the specific implementation process includes: collecting environmental light intensity and color temperature through a multispectral sensor, calculating the correlated color temperature according to the CIE 1931 standard, and evaluating the light quality in combination with the melatonin suppression effect (MEL). Measuring the environmental temperature and humidity through a temperature and humidity sensor, and generating an environmental comfort score based on the thermal comfort model. The evaluation of light quality helps to identify the potential role of the light environment in mood regulation to improve the accuracy of mood assessment.

[0064] The audio-visual acquisition unit collects data through a binocular camera and a microphone array to obtain behavior characteristics; the specific implementation process includes: using the binocular camera to capture the user's facial images at a frame rate of 60 fps, then extracting 68 facial landmark points and calculating the intensity values of 20 facial action units (AUs) through the OpenFace framework, and at the same time generating an eye movement fixation point coordinate sequence using the pupil tracking algorithm. Suppressing background noise through the microphone array using beamforming technology, extracting the fundamental frequency of speech through the YIN algorithm and identifying emotion keywords in combination with the BERT model. By capturing the user's behavior details, it provides a rich data source for emotion recognition, thus improving the multi-dimensionality and accuracy of emotion recognition.

[0065] The data preprocessing unit performs wavelet denoising and standardization alignment operations on physiological signals, environmental parameters, and behavior characteristics respectively to obtain the original data set. The specific implementation process includes: performing denoising processing on the signal using wavelet transform and eliminating high-frequency noise through the soft threshold rule. Using the Z-score standardization method to eliminate the dimensional difference between different signals, and dynamically updating the mean through the exponentially weighted moving average. Achieving time synchronization of multi-source data based on the PTP protocol, and using the cubic spline interpolation method to align the timestamps. Wavelet denoising and time synchronization technologies ensure the time consistency of multi-modal data.

[0066] Specifically, the emotional data analysis module includes: a feature extraction unit, a dynamic fusion unit, and a quality calibration unit; the working process of this module is as Figure 4 shown.

[0067] The feature extraction unit extracts features from the original data set through the Poincaré scatter plot and baseline calibration technology to obtain emotional features; the specific implementation process includes: using wavelet transform technology to extract the low-frequency (LF) and high-frequency (HF) power ratios of the heart rate variability (HRV) signal, and analyzing its non-linear features through the Poincaré scatter plot, which reflects the balance state of the autonomic nervous system. Using the OpenFace framework to extract the intensity values of 20 facial action units (AUs) in the facial image data, and calculating the relative intensity change through the method of dynamic calibration to capture subtle facial expression changes. Then calculate the fundamental frequency jitter and spectral tilt of the speech signal through the fundamental frequency extraction algorithm. By separating multi-modal features of different signals, the extraction efficiency of emotion-related information is significantly improved, and thus the feature extraction is more accurate and robust.

[0068] The dynamic fusion unit evaluates the importance of emotional features through the dynamic time warping algorithm and dynamic weight allocation technology to obtain normalized features; the specific implementation process includes: synchronizing the data in time through the Transformer architecture, and at the same time capturing the time dependence relationship between different modal signals through the self-attention mechanism, calculating the contribution degree of each modal feature based on the dynamic weight allocation algorithm and gradient backpropagation significance analysis as the feature importance, then using the Softmax function for normalization processing, and splicing the weighted feature vectors. Through time alignment and weight allocation, the problems of time series inconsistency and uneven feature contribution degree in multi-modal data fusion are solved, and the accuracy and stability of the fusion result are improved.

[0069] The quality calibration unit detects the quality of the normalized features through adaptive filtering technology to obtain a fused feature vector. The specific implementation process includes: evaluating the consistency of multi-modal data by calculating the phase locking value of respiration-heart rate (PLV), marking the data as abnormal when the PLV is lower than 0.35, using the NLMS algorithm to perform adaptive filtering and denoising processing on the signal, the step size parameter μ of the NLMS algorithm is 0.01, and then dynamically updating the data quality score based on the sliding time window) to achieve the effect of removing low-quality data segments, thus significantly improving the reliability of the data.

[0070] Specifically, the mental state evaluation module includes: an emotion recognition unit, a comprehensive evaluation unit, and a baseline calibration unit; the working process of this module is as Figure 5 shown.

[0071] The emotion recognition unit classifies and outputs the fused feature vector through time series modeling and cross-entropy loss function to obtain the emotion score; the specific implementation process includes: inputting the fused feature vector into the ResNet-50 network, extracting spatial features through residual blocks, and compressing the feature dimensions using global average pooling, and then inputting the feature sequence into the Transformer encoder, capturing the temporal dependency through the multi-head self-attention mechanism, and finally passing through the fully connected layer to output the six-dimensional emotion score, and using the cross-entropy loss function with category weights to optimize the network.

[0072] The comprehensive assessment unit calculates the mental health risk index based on the emotion score through the Markov chain Monte Carlo method; the specific implementation process includes: constructing a Bayesian network topology structure containing observation nodes and hidden nodes, initializing the conditional probability table through expert knowledge; then using Gibbs sampling in the Markov chain Monte Carlo method for posterior probability inference, setting 5000 sampling iterations and removing the first 1000 burning period data to eliminate the impact of the initial value, and calculating the mental health risk index at the same time. It avoids the direct calculation of high-dimensional integrals and reduces the risk assessment error.

[0073] The baseline calibration unit uses a Gaussian mixture model to perform error correction on the mental health risk index to obtain a comprehensive evaluation result. The specific implementation process includes: collecting the user's resting state data for 7 consecutive days (3 times a day, 5 minutes each time), using a 3-component GMM Gaussian mixture model to fit the baseline distribution of HRV, and then calculating the Mahalanobis distance between the real-time data and the baseline through a sliding time window (the window size of this embodiment is 24 hours, and the step size is 1 hour). When the Mahalanobis distance exceeds the distribution threshold, the baseline update is triggered, and the linear regression calibration formula is applied to correct the output result of the GMM model to obtain a comprehensive evaluation result. The Mahalanobis distance can identify abnormal states caused by environmental mutations, and the linear regression calibration eliminates the system errors caused by equipment drift, which greatly improves the system stability.

[0074] It should be noted that this module achieves dual optimization of emotion recognition and risk assessment through the collaboration of hybrid models, probabilistic graphical models and dynamic calibration technology. At the same time, each unit forms a complementary technology, jointly ensuring that the assessment results have both high accuracy and individual characteristics.

[0075] Specifically, the intervention plan generation module includes: a strategy calculation unit, a digital twin unit, and a risk assessment unit; the workflow of this module is as follows: Figure 6 shown.

[0076] The strategy calculation unit performs multi-objective optimization on the comprehensive evaluation results through the Pareto front objective function to obtain an initial plan. The specific implementation process includes: encoding the comprehensive evaluation results into a 12-dimensional vector (including emotion scores, HRV parameters, environmental light values, etc.), and at the same time constructing an action space including tDCS intensity, light therapy color temperature, and sound wave frequency. The Q-learning algorithm is used to update the state-action value matrix of the space, where the discount factor is 0.9 and the learning rate is 0.1. Then, the NSGA-II algorithm is used to solve the Pareto optimization objective function, and the strategy set after non-dominated sorting is output, so as to obtain the initial plan. The dynamic learning ability of the Q-learning algorithm enables the system to adapt to individual differences. At the same time, Pareto optimization balances multi-objective conflicts, and the elitist retention strategy of the NSGA-II algorithm avoids local optima, ensuring that the initial plan achieves the best trade-off between emotional stability and physiological safety.

[0077] The digital twin unit obtains the simulated preview effect by performing finite element analysis on the initial plan and solving Maxwell's equations. The specific implementation process includes: dividing bones, muscles, and nerve tissues into a finite number of grids through a standard human bioelectric model, applying current density boundary conditions to the grids, simulating the electric field distribution of the prefrontal cortex by solving Maxwell's equations, and at the same time generating a light therapy retinal response curve to analyze the effect of different color temperatures on melatonin secretion. The simulation step size is set to 1 ms, and a preview report including expected effects and risk indicators is output.

[0078] The risk assessment unit screens the strategies for the simulated preview effect through a logistic regression model and an adverse event database to obtain an intervention strategy package. The adverse event database stores historical risk cases related to the intervention plan. Using this database to construct a training set, the logistic regression coefficients are solved by maximum likelihood estimation. The results with a risk probability greater than 5% in the output of the logistic regression model are excluded, and the remaining results are sorted in descending order of the effect-risk ratio to generate the final intervention strategy package, enabling the system to achieve quantitative prediction and active avoidance of risks.

[0079] Specifically, the emotion change feedback module includes: a multi-modal execution unit and a real-time monitoring unit; the working process of this module is as Figure 7 shown.

[0080] The multimodal execution unit performs emotion regulation through neuromodulation tDCS, light therapy, and acoustic wave intervention according to the intervention strategy package, and obtains the emotion regulation result. The specific implementation process includes: adjusting the stimulation parameters through a constant current source according to the intervention strategy to achieve tDCS regulation, and real-time monitoring of the electrode contact impedance to ensure stimulation safety; the multi-spectral LED array dynamically adjusts the light intensity and color temperature through PWM dimming drive and spectral peak matching curve to achieve light therapy control; using a class D amplifier to drive the oscillator, generating a low-frequency acoustic wave signal based on sinusoidal carrier modulation, and controlling the sound pressure level within 85 dB, reducing the system output error.

[0081] The real-time monitoring unit calculates the change rate of the emotion regulation result through a linear mixed effects model to obtain the emotion regulation efficacy index. The specific implementation process includes: collecting key indicators such as HRV, GSR, and respiratory rate at a frequency of 10 Hz, and using the first-order difference method to calculate the change rate of each indicator; then constructing a linear mixed effects model to output the emotion regulation efficacy index, and identifying abnormal responses through dynamic threshold detection. This unit realizes the precise capture of user responses through dynamic modeling and anomaly detection, and reduces system latency.

[0082] Specifically, the scheme optimization module includes: an effect evaluation unit and a parameter iteration unit; the workflow of this module is as Figure 8 shown.

[0083] The effect evaluation unit conducts a statistical test on the emotion regulation efficacy index through HRV analysis and paired sample t-test to obtain the effectiveness index. The specific implementation process includes: extracting the heart rate intervals through a dual-threshold QRS wave detection algorithm, generating heart rate variability indicators (HRV) through time-domain analysis and frequency-domain analysis, judging whether the emotion regulation efficacy index has statistical significance through paired sample t-test, and adding the statistically significant emotion regulation efficacy index and HRV in a weighted manner to obtain the effectiveness index. This unit intuitively reflects the impact of the intervention strategy and quantifies the significance of the intervention effect.

[0084] The parameter iteration unit performs evolutionary operations and strategy updates through a genetic algorithm based on the effectiveness index to obtain the final scheme. The specific implementation process includes: using the effectiveness index as the fitness function, retaining the parameter combinations with higher fitness through selection operations, exchanging some genes of two excellent parameter combinations through crossover operations to generate new parameter combinations, then randomly adjusting some parameters through mutation operations to introduce diversity and avoid falling into local optimal solutions, and finally generating the final scheme with the highest fitness after multiple rounds of iteration. This unit can handle complex multi-parameter optimization problems, continuously improve the intervention strategy by simulating the biological evolution process, ensure the scientificity and efficiency of the scheme, and significantly improve the accuracy and adaptability of the intervention scheme.

[0085] Specifically, the data security protection module includes: a privacy protection unit and an access control unit;

[0086] The privacy protection unit encrypts the data of the final solution through a distributed deep learning training library and an encryption algorithm. The specific implementation process includes: dividing the data of the final solution into multiple parts and storing them on different distributed nodes respectively; using federated learning technology, training the deep learning model on each node without sharing the original data to generate encryption keys; using a symmetric encryption algorithm to perform preliminary encryption on the data; using an asymmetric encryption algorithm to perform secondary encryption on the symmetric key to ensure the security of the key; and storing the encrypted data and key in different security areas respectively.

[0087] The access control unit detects anomalies of system access personnel through zero-trust strategy and blockchain evidence storage method. The specific implementation process includes: multi-factor authentication of system access personnel, including password, fingerprint, facial recognition, etc.; dynamically adjust access rights according to the role and authority of the access personnel to ensure the principle of least privilege; monitor access behavior in real time, detect and block abnormal operations. At the same time, the system access log and operation record are generated into a hash value and stored on the blockchain. The authenticity and integrity of the log are ensured through the tamper-proof characteristics of the blockchain. When a security incident occurs, the access record is traced through the blockchain to locate the abnormal behavior.

[0088] It should be noted that this module achieves efficient model training while protecting data privacy, and ensures the security of data during storage and transmission, while enhancing the transparency and traceability of the system.

[0089] The beneficial effects of the present invention are as follows:

[0090] 1) The dynamic weight allocation algorithm of the Transformer architecture integrates physiological signals, behavioral characteristics and environmental parameters, solving the misjudgment problem caused by single data in traditional methods, improving the accuracy of emotion recognition and reducing the false alarm rate;

[0091] 2) The Q-learning-based reinforcement learning algorithm is used to perform digital twin rehearsal verification, which avoids the lag of traditional static intervention and improves the system response speed;

[0092] 3) Combining the federated learning framework and homomorphic encryption, it reduces the risk of data leakage and improves the accuracy of log tampering detection;

[0093] 4) Iteratively update the intervention plan through genetic algorithms, making the plan more adaptable to individual physiological changes and environmental fluctuations, ensuring the continued effectiveness of the intervention strategy;

[0094] 5) Combining tDCS neuromodulation and Bayesian risk assessment realizes the synergy of interdisciplinary technologies and ensures the scientificity and practicality of data.

[0095] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. Specific examples are used in the present invention to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention. At the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A system platform for psychological assessment and emotional feedback, characterized in that: include: Multi-source data acquisition module, emotional data analysis module, psychological state assessment module, intervention plan generation module, emotional change feedback module, plan optimization module and data security protection module; The multi-source data acquisition module obtains the original data set by performing signal denoising and format standardization on the collected physiological signals, environmental parameters and behavioral characteristics; The emotion data analysis module obtains a fused feature vector by performing key feature separation, data alignment and abnormal data removal on the original data set; The mental state assessment module predicts mental activities based on the fused feature vector through an intelligent network model to obtain a comprehensive assessment result; The intervention plan generation module generates an initial plan based on the comprehensive evaluation results through a reinforcement learning algorithm and performs a simulation preview to obtain an intervention strategy package; The emotion change feedback module obtains the emotion regulation effectiveness index through multimodal execution and real-time monitoring based on the intervention strategy package; The solution optimization module performs effect evaluation and parameter iteration according to the emotion regulation efficacy index to obtain a final solution; The data security protection module encrypts the data of the final solution through federated learning and homomorphic encryption technology.

2. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The multi-source data acquisition module includes: a biological sensing unit, an environmental sensing unit, an audio-visual acquisition unit and a data pre-processing unit; The biosensor unit performs data analysis on the collected data through a dual-threshold QRS wave detection algorithm and a Kalman filter operation to obtain the physiological signal; The environmental sensing unit performs data analysis on the collected data through a thermal comfort model and spectrum analysis to obtain the environmental parameters; The audio-visual collection unit collects data through a binocular camera and a microphone array to obtain the behavior characteristics; The data preprocessing unit performs wavelet denoising and standardized alignment operations on the physiological signal, the environmental parameter and the behavioral feature respectively to obtain the original data set.

3. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The emotion data analysis module includes: a feature extraction unit, a dynamic fusion unit and a quality calibration unit; The feature extraction unit extracts features from the original data set by using a Poincare scatter plot and a baseline calibration technique to obtain emotional features; The dynamic fusion unit evaluates the importance of the emotional features through a dynamic time warping algorithm and a dynamic weight allocation technology to obtain a normalized feature; The quality calibration unit performs quality detection on the normalized feature through an adaptive filtering technology to obtain the fused feature vector.

4. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The mental state assessment module includes: an emotion recognition unit, a comprehensive assessment unit and a baseline calibration unit; The emotion recognition unit classifies and outputs the fused feature vector through time series modeling and a cross entropy loss function to obtain an emotion score; The comprehensive evaluation unit calculates the mental health risk index based on the emotion score by using the Markov Chain Monte Carlo method; The baseline calibration unit performs error correction on the mental health risk index through a Gaussian mixture model to obtain the comprehensive assessment result.

5. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The intervention plan generation module includes: a strategy calculation unit, a digital twin unit and a risk assessment unit; The strategy calculation unit performs multi-objective optimization on the comprehensive evaluation result through a Pareto frontier objective function to obtain the initial solution; The digital twin unit obtains a simulation preview effect by performing finite element analysis on the initial scheme and solving Maxwell equations; The risk assessment unit performs strategy screening on the simulation rehearsal effect through a logistic regression model and an adverse event database to obtain the intervention strategy package.

6. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The emotion change feedback module includes: a multimodal execution unit and a real-time monitoring unit; The multimodal execution unit performs emotion regulation through neural tDCS regulation, light therapy and sound wave intervention according to the intervention strategy package to obtain an emotion regulation result; The real-time monitoring unit calculates the change rate of the emotion regulation result through a linear mixed effect model to obtain the emotion regulation efficacy index.

7. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The scheme optimization module includes: an effect evaluation unit and a parameter iteration unit; The effect evaluation unit performs a statistical test on the emotion regulation efficacy index through HRV analysis and paired sample t-test method to obtain an effectiveness index; The parameter iteration unit performs evolutionary operations and strategy updates through a genetic algorithm based on the effectiveness index to obtain the final solution.

8. The system platform for psychological assessment and emotional feedback according to claim 1, characterized in that: The data security protection module includes: a privacy protection unit and an access control unit; The privacy protection unit encrypts data of the final solution through a distributed deep learning training library and an encryption algorithm; The access control unit performs anomaly detection on system access personnel through zero-trust strategy and blockchain evidence storage method.

Images