Personalized psychological portrait generation evaluation method

Abstract

The invention relates to the technical field of psychology and artificial intelligence, in particular to a personalized psychological portrait generation evaluation method, which comprises the steps of constructing a multi-source heterogeneous data acquisition system, updating a time sequence behavior pattern, performing multi-dimensional association analysis, dynamically adjusting an adaptive optimization algorithm and the like. According to the method, real-time capturing and modeling of the psychological state of the user can be achieved by fusing multi-source data, the accuracy and adaptability of the psychological portrait are improved by combining the graph neural network and the genetic algorithm, and the defects in the aspects of dynamic psychological state capturing, individual difference modeling and multi-modal data processing in the prior art are overcome; and technical support is provided for psychological health intervention, behavior prediction and psychological assessment in a complex environment.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of cross-application of information technology and psychology, specifically to a personalized psychological portrait generation and evaluation method.

[0002] The present application relates to the technical field of psychology and artificial intelligence, specifically to a personalized psychological portrait generation and evaluation method.

[0003] With the deep integration of psychology and artificial intelligence technology, the personalized psychological portrait generation and evaluation method has shown wide application prospects in the fields of psychological health intervention, user behavior prediction, and personalized service recommendation. However, the existing psychological portrait generation and evaluation technology still has certain deficiencies in data processing depth, model adaptability, and individualization degree, which limits its efficient application in actual scenarios.

[0004] After searching, a psychological portrait recognition model training method and a collection action recommendation system with publication number CN117272159B were disclosed on January 7, 2025. This patent obtains the score value of the sample object on multiple psychological evaluation labels, maps the psychological characteristics of different dimensions to the Euclidean space, and realizes the automatic allocation of psychological portrait categories by combining clustering analysis. However, this technical solution mainly relies on the pre-set psychological evaluation label system, and has certain limitations in real-time capture of dynamically changing psychological state, which may lead to the lag and insufficient accuracy of the psychological portrait generation result. In addition, this solution has limited modeling ability for individual differences, and it is difficult to meet the needs of highly personalized application scenarios.

[0005] After searching, a psychological counseling strategy generation method, device, equipment, and medium based on an intelligent agent with publication number CN118964598B were disclosed on January 21, 2025. This patent collects audio and video signals of the target user, extracts multi-modal signals using computer vision and speech recognition technology, and generates psychological counseling strategies by combining decision trees and multi-modal large models. At the same time, a user portrait model is constructed based on the user's historical counseling records to further adjust the counseling strategy for personalized recommendation. However, this technical solution is relatively static in describing the user's psychological state during the psychological portrait generation process, and fails to fully explore the deep relationship between the user's potential psychological characteristics and behavior patterns. In addition, in complex scenarios, the collection and processing of multi-modal signals may be affected by environmental noise and device performance, thereby bringing certain challenges to the accuracy and stability of psychological portrait generation.

[0006] The above problems show that the existing psychological portrait generation evaluation technology still has certain improvement space in dynamic psychological state capture, individual difference modeling and multi-modal data processing capability, etc. Therefore, the present application provides a personalized psychological portrait generation evaluation method, aiming to improve the real-time performance, accuracy and adaptability of psychological portrait generation by fusing multi-source heterogeneous data, introducing a dynamic updating mechanism and optimizing a personalized modeling algorithm, so as to meet the needs of the modern psychology and artificial intelligence field for efficient and intelligent psychological portrait generation evaluation methods. SUMMARY

[0007] To solve the above problems, the present application aims to provide a personalized psychological portrait generation evaluation method, which fuses the dynamic acquisition and processing mechanism of multi-source heterogeneous data, introduces a behavior pattern updating module based on time series, establishes a real-time capture and modeling system of user psychological state, and through deep mining and multi-dimensional correlation analysis of user's potential psychological characteristics, combines an adaptive optimization algorithm to dynamically adjust the psychological portrait model, thereby solving the deficiencies of the prior art in dynamic psychological state capture, individual difference modeling and multi-modal data processing capability, and providing technical support for efficient and intelligent psychological portrait generation evaluation in the fields of psychology and artificial intelligence.

[0008] To achieve the above purpose, the technical scheme of the present application is as follows:

[0009] A personalized psychological portrait generation evaluation method, comprising the following steps:

[0010] A multi-source heterogeneous data acquisition system is constructed, a plurality of sensor nodes are set to obtain the physiological signals, behavior trajectories and environmental information of the user, the sensor nodes are connected to a central data processing unit through a wireless communication protocol, and a data synchronization module is configured to realize timestamp alignment of multi-source data; an adaptive filtering algorithm is used to suppress noise and extract features from the collected multi-source data, forming a basic behavior feature set of the user;

[0011] A time series analysis module is introduced in the basic behavior feature set, the behavior patterns of the user are divided into stages through a sliding window mechanism; the similarity between the current behavior and the historical behavior of the user is calculated based on the historical behavior record of the user, and a behavior pattern updating matrix is generated; the psychological state label distribution of the user is dynamically adjusted according to the behavior pattern updating matrix, forming a real-time psychological state capture model;

[0012] A multi-dimensional correlation analysis module is configured to correlate and mine the psychological state label distribution and potential psychological characteristics of the user; the psychological state label of the user is mapped to its potential psychological characteristics by constructing a user psychological feature map, forming a psychological feature correlation network; the psychological feature correlation network is trained using a graph neural network, generating an initial model of the user's psychological portrait;

[0013] An adaptive optimization algorithm is introduced into the initial model of the user psychological portrait to dynamically adjust the model parameters; the deviation between the initial model of the user psychological portrait and the actual behavior data is compared to calculate the adaptability score of the model; the model parameters are iteratively optimized according to the adaptability score to generate the final personalized psychological portrait model;

[0014] Further, a plurality of sensor nodes are provided, including but not limited to wearable devices, cameras, microphones and environmental sensors, wherein each sensor node is connected to a central data processing unit through a wireless communication protocol; the central data processing unit is configured with a data storage module and a data synchronization module for receiving and storing data from each sensor node, and realizing time synchronization of multi-source data through a timestamp alignment algorithm;

[0015] Further, the adaptive filtering algorithm adopts a combination of Kalman filtering and wavelet transform to denoise the collected multi-source data; the physiological signal features, behavior trajectory features and environmental features of the user are extracted from the denoised data to form the user's basic behavior feature set;

[0016] Further, the time series analysis module divides the user's behavior pattern in stages through a sliding window mechanism, and the length of the sliding window is set to between 10 seconds and 60 seconds; combined with the user's historical behavior records, the cosine similarity between the user's current behavior and the historical behavior is calculated to generate a behavior pattern update matrix;

[0017] Further, the multi-dimensional association analysis module uses the Apriori algorithm to associate and mine the psychological state label distribution and potential psychological features of the user; in the psychological feature association network, the user's psychological state label is used as a node, and the potential psychological feature is used as an edge, and the weight of the edge is determined by the association strength of the two;

[0018] Further, the graph neural network uses the GraphSAGE algorithm to train the psychological feature association network, and uses the cross-entropy loss function to optimize the model during training; after training, an initial model of the user psychological portrait is generated, which includes the user's psychological state label distribution and its corresponding potential psychological features;

[0019] Further, the adaptive optimization algorithm uses a genetic algorithm to dynamically adjust the model parameters, and the population size of the genetic algorithm is set to 50 to 100, and the number of iterations is set to 200; the mean square error between the initial model of the user psychological portrait and the actual behavior data is calculated to evaluate the adaptability score of the model;

[0020] Further, in the step, the calculation formula of the adaptive score is: [S = \frac{1}{N} \sum_{i=1}^{N} (P_i - A_i)^2] Wherein, SS represents the adaptive score, PiPi represents the psychological state value predicted by the user psychological portrait model, AiAi represents the psychological state value in the actual behavior data, NN represents the sample quantity;

[0021] Further, the expression of the final personalized psychological portrait model is: [M = \alpha \cdot F + \beta \cdot G + \gamma \cdot H] Wherein, MM represents the final personalized psychological portrait model, FF represents the basic behavior feature set of the user, GG represents the psychological feature association network, HH represents the output result of the adaptive optimization algorithm, and αα, ββ, γγ respectively represent the weight coefficients of each part;

[0022] In order to achieve the above purpose, the application further provides an application of a personalized psychological portrait generation and evaluation method, which can be used in a psychological health intervention scene, or used as a personalized service recommendation system for user behavior prediction, or used as a psychological evaluation tool driven by multi-modal data in a complex environment.

[0023] Beneficial effects: the application realizes the stage division and update of user behavior patterns by fusing the dynamic acquisition and processing mechanism of multi-source heterogeneous data and combining a time series analysis module, solves the insufficient ability of the prior art in dynamic psychological state capture, deeply mines user psychological state labels and potential psychological features through a multi-dimensional correlation analysis module, improves the accuracy of psychological portrait generation, dynamically adjusts model parameters through an adaptive optimization algorithm, and enhances the adaptability and stability of the model. In addition, the application can be used as an intelligent tool for psychological health intervention, and can be used as a multi-modal data processing and psychological evaluation solution in a complex scene. DESCRIPTION OF DRAWINGS

[0024] Figure 1 It is a whole flowchart of the personalized psychological portrait generation and evaluation method in the embodiment of the application, which shows the complete steps from multi-source data acquisition to the generation of the final psychological portrait model.

[0025] Figure 2 It is a working principle diagram of the time series analysis module in the embodiment of the application, which details the process of the stage division of user behavior patterns and the generation of the behavior pattern update matrix by the sliding window mechanism.

[0026] Figure 3The embodiment of the present application is a schematic diagram of the construction and training process of the psychological feature association network, which shows the process of training the psychological feature association network through the graph neural network and generating the initial psychological portrait model.

[0027] The reference signs are as follows: 1, multi-source data acquisition system; 2, time series analysis module; 3, psychological feature association network; 4, graph neural network; 5, adaptive optimization algorithm; 6, sliding window mechanism; 7, behavior pattern update matrix; 8, initial psychological portrait model; 9, final psychological portrait model. DETAILED DESCRIPTION

[0028] The present application provides a personalized psychological portrait generation and evaluation method, which is described in detail in the specific implementation combined with the Figure 1 , Figure 2 and Figure 3 . Figure 1 The overall process of the personalized psychological portrait generation and evaluation method is shown, including the complete steps from multi-source data acquisition to final psychological portrait model generation; Figure 2 The working principle of the time series analysis module is described in detail, including the process of phase division of user behavior patterns by the sliding window mechanism and the generation of the behavior pattern update matrix; Figure 3 The construction and training process of the psychological feature association network, as well as the process of generating the initial psychological portrait model through the graph neural network, are shown. The specific embodiments of the present application will be described in detail in combination with these drawings and reference signs.

[0029] First, the multi-source data acquisition system 1 is the basic component of the entire method, and its core function is to obtain multi-source heterogeneous data of users, including physiological signals, behavior trajectories and environmental information. The multi-source data acquisition system 1 is composed of multiple sensor nodes, which can be wearable devices, cameras, microphones or environmental sensors, etc. Each sensor node is connected to the central data processing unit through a wireless communication protocol, and the central data processing unit is configured with a data storage module and a data synchronization module. The data storage module is used to receive and store data from each sensor node, while the data synchronization module realizes the time synchronization of multi-source data through a timestamp alignment algorithm. The core of the timestamp alignment algorithm is to add accurate time labels to each piece of data, and to align data from different sources according to the time labels, so as to ensure the consistency of the data in the subsequent analysis process. In practical applications, for example, in the context of mental health intervention, wearable devices can monitor the user's heart rate, skin electrical response and other physiological signals in real time, cameras can capture the user's facial expressions and body movements, and environmental sensors can record the intensity of light, temperature and other environmental information. These data together constitute the user's basic behavior feature set.

[0030] Next, an adaptive filtering algorithm is applied to the collected multi-source data for noise suppression and feature extraction. The adaptive filtering algorithm combines Kalman filtering and wavelet transform, where Kalman filtering removes random noise by recursively estimating the state space model of the data, and wavelet transform extracts key features through multi-scale decomposition of the signal. In the denoised data, further extract the user's physiological signal features, behavior trajectory features and environmental features to form the user's basic behavior feature set F. The basic behavior feature set F is an important input for subsequent analysis, and its content includes but is not limited to heart rate variation curve, gait features, speech tone features, and environmental light intensity trend, etc. The extraction process of these features needs to be combined with specific sensor types and application scenarios, for example, in the personalized service recommendation system of user behavior prediction, speech tone features can be obtained through spectral analysis of audio data collected by the microphone.

[0031] On the basis of the basic behavior feature set F, a time series analysis module 2 is introduced to divide the user's behavior pattern into stages. The time series analysis module 2 achieves this goal through a sliding window mechanism 6, the length of which is set to 10-60 seconds, and the specific length can be adjusted according to the needs of the application scenario. The core of the sliding window mechanism 6 is to divide the continuous time series data into several time periods, and the data in each time period is regarded as an independent sample. On this basis, combined with the user's historical behavior records, the cosine similarity between the user's current behavior and the historical behavior is calculated to generate a behavior pattern update matrix 7. The generation process of the behavior pattern update matrix 7 is as follows: first, represent the user's current behavior feature vector and historical behavior feature vector as points in a high-dimensional space, then calculate the cosine angle between the two to measure their similarity. The value range of the cosine similarity is -1 to 1, and the closer the value is to 1, the more similar the two are. According to the behavior pattern update matrix 7, dynamically adjust the user's psychological state label distribution to form a real-time psychological state capture model. The real-time psychological state capture model can reflect the user's psychological state changes in different time periods, for example, in a multi-modal data-driven psychological assessment tool in a complex environment, the user's emotional fluctuations can be judged by capturing their psychological state changes in real time.

[0032] Subsequently, a multi-dimensional association analysis module is configured to associate the user's psychological state label distribution with potential psychological characteristics. The multi-dimensional association analysis module uses the Apriori algorithm to achieve this goal. The Apriori algorithm calculates the support and confidence of frequent item sets to mine the association rules between the user's psychological state label and potential psychological characteristics. In the psychological characteristic association network 3, the user's psychological state label is used as a node, and the potential psychological characteristics are used as edges. The weight of the edge is determined by the association strength of the two. For example, in the psychological health intervention scenario, the user's low mood label may have a high association strength with potential psychological characteristics such as poor sleep quality and reduced social activity. By constructing the user's psychological characteristic map, the user's psychological state label is mapped to its potential psychological characteristics to form the psychological characteristic association network 3. The construction process of the psychological characteristic association network 3 needs to be combined with specific application scenarios. For example, in the personalized service recommendation system of user behavior prediction, the user's potential psychological characteristics can be mined by analyzing the user's historical purchase behavior and interest preferences.

[0033] The training process of the psychological characteristic association network 3 is implemented by a graph neural network 4. The graph neural network 4 uses the GraphSAGE algorithm to train the psychological characteristic association network 3. The GraphSAGE algorithm samples and aggregates the neighbor information of the nodes to generate the embedding representation of each node. During the training process, the cross-entropy loss function is used to optimize the model. The cross-entropy loss function compares the probability distribution predicted by the model with the probability distribution of the true label, calculates the error of the model, and performs backpropagation. After training, the initial model 8 of the user's psychological portrait is generated. This model contains the user's psychological state label distribution and its corresponding potential psychological characteristics. The generation process of the user's psychological portrait initial model 8 needs to be combined with specific application scenarios. For example, in the psychological assessment tool driven by multi-modal data in complex environments, the initial model obtained by training can be used to preliminarily evaluate the user's psychological state.

[0034] An adaptive optimization algorithm 5 is introduced in the user psychological portrait initial model 8 to dynamically adjust the model parameters. The adaptive optimization algorithm 5 uses a genetic algorithm to achieve this goal. The genetic algorithm gradually optimizes the model parameters by selecting, crossing, and mutating individuals in the population. The population size of the genetic algorithm is set to 50 to 100, and the number of iterations is set to 200. The specific parameters can be adjusted according to the needs of the application scenario. The adaptability score S of the model is evaluated by calculating the mean square error between the user psychological portrait initial model 8 and the actual behavior data. The calculation formula of the adaptability score S is S equals 1 divided by N multiplied by the sum of Pi minus Ai squared from i equals 1 to N, where S represents the adaptability score, Pi represents the psychological state value predicted by the user psychological portrait model, Ai represents the psychological state value in the actual behavior data, and N represents the number of samples. According to the adaptability score S, the model parameters are iteratively optimized to generate the final personalized psychological portrait model 9. The expression of the final personalized psychological portrait model 9 is M equals alpha multiplied by F plus beta multiplied by G plus gamma multiplied by H, where M represents the final personalized psychological portrait model, F represents the user's basic behavior feature set, G represents the psychological feature association network, H represents the output result of the adaptive optimization algorithm, and alpha, beta, and gamma represent the weight coefficients of each part, respectively. The determination of the weight coefficients needs to be combined with the specific application scenario, for example, in the psychological health intervention scenario, different weights can be assigned according to the importance of the user's psychological state label distribution.

[0035] The present application also provides application examples of the personalized psychological portrait generation and evaluation method. In the psychological health intervention scenario, the personalized psychological portrait generation and evaluation method can capture the user's psychological state changes in real time and provide personalized psychological health intervention solutions for the user. For example, when detecting that the user's mood is low, the system can automatically recommend relaxation training or psychological counseling and other intervention measures. In the personalized service recommendation system of user behavior prediction, the personalized psychological portrait generation and evaluation method can analyze the user's psychological state label and potential psychological features to provide personalized service recommendations for the user. For example, when detecting that the user's interest preferences have changed, the system can automatically adjust the recommended content to meet the user's new needs. In the psychological evaluation tool driven by multi-modal data in complex environments, the personalized psychological portrait generation and evaluation method can fuse multi-source heterogeneous data to provide comprehensive psychological evaluation reports for users. For example, in the disaster rescue scenario, the system can analyze the psychological state changes of rescue personnel to provide timely psychological support.

[0036] The above embodiments describe the specific steps and operating principles of the personalized psychological portrait generation evaluation method in detail, covering the complete process from multi-source data acquisition to the final psychological portrait model generation. By combining the dynamic acquisition and processing mechanism of multi-source heterogeneous data, the time series analysis module, the multi-dimensional correlation analysis module, and the adaptive optimization algorithm, the present application realizes real-time capture and modeling of user psychological state, solving the shortcomings of existing technologies in dynamic psychological state capture, individual difference modeling, and multi-modal data processing capability. In order to better enable relevant personnel in this technical field to fully understand and implement the present application, the specific implementation principles of the present application are further supplemented below in conjunction with a specific application scenario.

[0037] In the psychological health intervention scenario, the personalized psychological portrait generation evaluation method acquires the user's psychological state change information in real time through the multi-source data acquisition system 1. First, the wearable device as one of the sensor nodes is used to monitor the user's physiological signals such as heart rate and skin electrical response. These data are transmitted to the central data processing unit through a wireless communication protocol, and are time-stamped by the data synchronization module to ensure time consistency with other source data. At the same time, the camera captures the user's facial expressions and body movements, and the environmental sensor records environmental information such as light intensity and temperature. These data together constitute the user's basic behavior feature set F. The extraction process of the basic behavior feature set F relies on an adaptive filtering algorithm that combines Kalman filtering and wavelet transform to remove random noise and extract key features. For example, in the psychological health intervention scenario, by analyzing the heart rate variation curve and the voice tone features, the user's emotional fluctuation trend can be preliminarily judged.

[0038] Subsequently, the time series analysis module 2 divides the user's behavior patterns into stages through the sliding window mechanism 6. The length of the sliding window is set to 30 seconds, which can be adjusted according to actual needs. On this basis, the cosine similarity between the user's current behavior feature vector and the historical behavior feature vector is calculated, and a behavior pattern update matrix 7 is generated. For example, when the user shows a trend of accelerated heart rate and low-pitched voice tone in consecutive time periods, the system will dynamically adjust the user's psychological state label distribution according to the behavior pattern update matrix 7 to form a real-time psychological state capture model. This dynamic adjustment mechanism can effectively cope with the rapid changes in the user's psychological state, thereby improving the real-time performance of the psychological portrait generation.

[0039] In the construction process of the psychological feature association network 3, the multi-dimensional association analysis module mines the association rules between the user psychological state labels and the potential psychological features through the Apriori algorithm. For example, in the psychological health intervention scenario, the user's low mood label may have a high association strength with potential psychological features such as poor sleep quality and reduced social activities. These association rules are mapped into the psychological feature association network 3, where the user's psychological state labels are nodes, the potential psychological features are edges, and the weight of the edge is determined by the association strength of the two. The psychological feature association network 3 is trained through the graph neural network 4, and the GraphSAGE algorithm is used to generate the embedding representation of each node. After training, the initial model 8 of the user psychological portrait is generated, which can reflect the distribution of the user's psychological state labels and their corresponding potential psychological features.

[0040] To further optimize the model performance, the adaptive optimization algorithm 5 uses a genetic algorithm to dynamically adjust the parameters of the user psychological portrait initial model 8. The population size of the genetic algorithm is set to 80, and the number of iterations is 200 times. In the optimization process, the mean square error between the user psychological portrait initial model 8 and the actual behavior data is calculated to evaluate the adaptability score S of the model. The calculation formula of the adaptability score S is: [S = \frac{1}{N} \sum_{i=1}^{N} (P_i - A_i)^2] where PiPi represents the user psychological portrait model predicted psychological state value, AiAi represents the psychological state value in the actual behavior data, NN represents the sample size. According to the adaptability score S, the model parameters are iteratively optimized, and finally the personalized psychological portrait model 9 is generated. The expression of the model is: [M = \alpha \cdot F + \beta \cdot G + \gamma \cdot H] where MM represents the final personalized psychological portrait model, FF represents the user's basic behavior feature set, GG represents the psychological feature association network, HH represents the output result of the adaptive optimization algorithm, αα, ββ, γγ respectively represent the weight coefficients of each part. In the psychological health intervention scenario, different weights can be assigned according to the importance of the user's psychological state label distribution, for example, the weight of the low mood label is set to a higher value to highlight its impact on the generation of the psychological portrait.

[0041] In practical applications, when the personalized psychological portrait generation and evaluation method detects that the user's mood is low, the system will automatically recommend interventions such as relaxation training or psychological counseling. For example, when the user's heart rate continues to rise and the voice tone is low, the system will judge that the user may have emotional stress according to the real-time psychological state capture model, and analyze the potential psychological characteristics such as poor sleep quality or reduced social activities through the psychological characteristic association network 3. Based on these analysis results, the system generates a personalized psychological health intervention plan and feeds back the suggestions to the user through the user interface.

[0042] The above embodiments describe in detail the specific operation principle of the personalized psychological portrait generation and evaluation method in the psychological health intervention scene, covering the complete process from multi-source data acquisition to the final psychological portrait model generation. By combining the dynamic acquisition and processing mechanism of multi-source heterogeneous data, the time series analysis module, the psychological characteristic association network and the adaptive optimization algorithm, the present application realizes the real-time capture and modeling of the user's psychological state, and solves the shortcomings of the prior art in dynamic psychological state capture, individual difference modeling and multi-modal data processing capability.

Claims

1. A method for generating and assessing personalized psychological profiles, characterized in that, Includes the following steps: S1: Construct a multi-source heterogeneous data acquisition system, set up multiple sensor nodes to acquire users' physiological signals, behavioral trajectories and environmental information. Each sensor node is connected to the central data processing unit through a wireless communication protocol, and a data synchronization module is configured to achieve timestamp alignment of multi-source data. S2: Use an adaptive filtering algorithm to suppress noise and extract features from the collected multi-source data to form a basic set of user behavior features; S3: Introduce a time series analysis module into the basic behavioral feature set, divide the user's behavior pattern into stages through a sliding window mechanism, calculate the cosine similarity between the current behavior and the historical behavior records of the user, and generate a behavior pattern update matrix. S4: Dynamically adjust the distribution of users' psychological state labels based on the behavior pattern update matrix to form a real-time psychological state capture model; S5: Configure a multi-dimensional correlation analysis module to mine the correlation between the distribution of users' psychological state labels and potential psychological features, and map psychological state labels and potential psychological features by constructing a user psychological feature map to form a psychological feature correlation network. S6: Use graph neural networks to train the psychological feature association network to generate an initial model of the user psychological profile; S7: Introduce an adaptive optimization algorithm into the initial model of the user psychological profile to dynamically adjust the model parameters. Evaluate the model's adaptability score by calculating the mean square error between the initial model of the user psychological profile and the actual behavioral data. Iterate and optimize the model parameters based on the adaptability score to generate the final personalized psychological profile model.

2. The personalized psychological profile generation and assessment method according to claim 1, characterized in that, The adaptive filtering algorithm uses a combination of Kalman filtering and wavelet transform to denoise the collected multi-source data, and extracts the user's physiological signal features, behavioral trajectory features and environmental features from the denoised data to form the user's basic behavioral feature set.

3. The personalized psychological profile generation and assessment method according to claim 1, characterized in that, The length of the sliding window mechanism is set between 10 and 60 seconds, the value of the cosine similarity is between -1 and 1, the adaptive optimization algorithm adopts a genetic algorithm, the population size is set between 50 and 100, and the number of iterations is set to 200.

Images