CN117877682A

CN117877682A - VR psychological test table compiling method, system, equipment and storage medium

Info

Publication number: CN117877682A
Application number: CN202311871945.6A
Authority: CN
Inventors: 段新; 金海燕; 卢嬿
Original assignee: Foshan Chuangshijia Technology Co ltd
Current assignee: Foshan Chuangshijia Technology Co ltd
Priority date: 2023-12-29
Filing date: 2023-12-29
Publication date: 2024-04-12

Abstract

The invention provides a VR psychological test table compiling method, a system, equipment and a storage medium, comprising the following steps: determining a psychological trait; constructing original VR scene item content based on psychology-related theory and clinical priori knowledge of the psychological trait; in the original VR scene item content, a sample (calibration group) and a normal sample (normal group) with the psychological trait are tested respectively to obtain calibration group data and normal group data, and VR scene item content which is the most characterized by the psychological trait is selected and optimized from the original VR scene item content based on a computer information theory, so that VR psychological test list item content of the psychological trait is determined. The method provided by the invention takes the principles of psychological theory, clinical priori knowledge and computer information theory as guidance, and can improve the pertinence of the contents of the compiled VR psychological test list item, and improve the compiling efficiency and the test accuracy of the VR psychological test list.

Description

VR psychological test table compiling method, system, equipment and storage medium

Technical Field

The invention relates to the fields of psychological tests, virtual reality and artificial intelligence, in particular to a VR psychological test meter compiling method, a VR psychological test meter compiling system, VR psychological test meter compiling equipment and a VR psychological test meter compiling storage medium.

Background

Psychological tests are a scientific means for deducing and quantitatively analyzing psychological characteristics of all behavioral activities of a person by observing a few representative behaviors of the person, and are used for testing, analyzing and evaluating specific diathesis or psychological traits of the person through scientific, objective and standard testing means. Quality refers to a personal characteristic of perception, memory, ability, gas quality, personality, interest, or motivation. While psychological traits are an inherent tendency of individuals to respond to stimuli under genetic and environmental influences, an internal behavior.

Since the psychological traits (internal display behaviors) cannot be directly obtained through testing, only the external display behaviors of the person can be tested through the psychological test items, and further the psychological traits of the person can be deduced through the reactions obtained through the psychological test items. Therefore, how to obtain effective psychological test items and to evaluate objective criteria becomes the key to whether psychological tests are accurate.

Objectivity of the measurement requires standardization of tests, such as test projects or job execution instructions; the speech attitude of the applicator; the physical environment, the score, the scoring principle and procedure, the conversion and interpretation of the score and the like are standardized. Only then is each test score tested objective. The test results between different tests can be compared with each other. While virtual reality technology naturally meets these characteristics. The virtual reality technology, namely VR technology, is a computer simulation system capable of creating and experiencing a virtual world, and utilizes a computer to generate a simulation environment, so that a user is immersed in the environment, and can feel surrounding changes personally, and further, corresponding actions and reactions are performed. In the VR environment, the content is comprehensive and deep, the tested reaction is free, the tested reaction can be fully expressed, no intention is kept or anxiety is caused, and the reflection is more true; the testing process of the instruction, the environment and the like is standardized, the scoring and scoring is standardized, and the score interpretation is standardized. The measuring system can be ensured to be stable and objective, and random errors and systematic errors are reduced.

The psychological test scale is a tool for psychological tests, and different psychological characteristics need to be tested corresponding to different test scales. The compilation of the current psychological test scale is mainly carried out manually, is very complicated, and is time-consuming and labor-consuming. The psychological test scale often needs a larger number of test items to obtain better effectiveness, and the content needs to be rich, so that the test content is not biased, and the representativeness of the behavior sample can be improved. Meanwhile, a plurality of test questions with compensatory property are inevitably repeated in the test table, so that the test efficiency is low, and tedious and lengthy test items easily cause contradiction emotion of the testee, thereby influencing the accuracy of the final test.

Disclosure of Invention

The invention aims to provide a VR psychological test table compiling method, system, equipment and storage medium, which are used for solving the technical problems, and in the VR psychological test table compiling process, psychological related theory, clinical priori knowledge and computer information theory are used as guidance to realize the optimal selection of psychological characteristics, so that the pertinence of the content of the VR psychological test table is improved, the compiling efficiency is improved, and the test accuracy of the VR psychological test table is improved.

In order to solve the technical problems, a VR psychological test scale programming method determines a psychological trait; acquiring preliminary psychological characteristics based on psychology-related theory and clinical priori knowledge of the psychological traits;

Constructing original VR scene item content based on the preliminary psychological characteristics; in the original VR scene item content, testing a calibration group and a normal group with the psychological characteristics respectively to obtain calibration group data and normal group data; and selecting and optimizing VR scene item contents which most represent the psychological characteristics from the original VR scene item contents based on the calibration group data and the normal group data and determining the VR psychological test list item contents which are the psychological characteristics.

The proposal takes the principle of psychological correlation theory and clinical priori knowledge as the guide, namely the dimension and the framework of the characteristics of the psychological trait to be tested (the characteristics are defined as primary psychological characteristics), further utilizes artificial intelligence to optimally select the primary psychological characteristics based on computer information theory, and can effectively improve the pertinence of the content of the compiled VR psychological test table, the compiling efficiency of the VR psychological test table and the test accuracy thereof.

Further, psychological traits include personality, intelligence, neuropsychology, behavioral or mental health of the child, and the like.

Further, in the obtaining of the preliminary psychological characteristics based on the psychology-related theory and the clinical prior knowledge of the psychological characteristics, the preliminary psychological characteristics include effective information characteristics and duplicate information characteristics, and the obtaining process specifically includes the following steps: based on a psychological correlation theory, obtaining effective information characteristics capable of representing the psychological characteristics; duplicate information features representing the psychological trait are obtained from a generally accepted psychological test form based on clinical prior knowledge.

Further, in constructing the original VR scene item content based on the preliminary psychological characteristics, constructing the original VR scene item content includes a scene design and a content design, specifically: the scene design is used for enabling the VR scene item content to reflect the text, sound, picture scene, smell or touch of the detected psychological characteristics; the content design is used for enabling the VR scene item content to reflect daily life and social scenes, special scenes, operation test scenes or projection test scenes.

In the above scheme, the construction of the VR scene item content is based on the preliminary psychological characteristics, after the preliminary psychological characteristics are obtained by taking the psychology-related theory and the clinical priori knowledge as the guide, the preliminary psychological characteristics are compiled, the corresponding VR scenes and the display content of the scenes are designed, so that the VR scene item content can show a certain psychological trait of a tested sample, and the response tendency of the obtained tested sample, namely the sample with the psychological trait is representative and can be measured by objective indexes.

Further, in the VR scene item content most characterized by the psychological characteristics is selected and optimized from the original VR scene item content based on the calibration group data and the normal group data, and is determined as the VR psychological test table item content of the psychological characteristics, the calibration group data and the normal group data are specifically implemented by adopting an artificial intelligent model based on the computer information theory, which specifically includes: training the artificial intelligent model based on the calibration set data and the normal set data to adjust parameters of the artificial intelligent model, sequencing the primary psychological characteristics and screening candidate psychological characteristics corresponding to the high weight; and determining the VR scene item content of the psychological trait from the candidate psychological characteristics with larger weights, and determining the VR psychological test scale item content of the psychological trait.

It should be noted that there are also prior art techniques that use learning models as inputs for features, but that either use hundreds of features or do an exhaustive search of a pre-selected subset of smaller features (with insufficient information content), which results in the prior art being either prone to overfitting or too limited in search space, the interpretation of the resulting learning model may be difficult to verify by existing neuropsychological theory. The above protocol, based on psycho-related theory and clinical prior knowledge and computer information theory, can reduce statistical power (the statistical power) with a relatively small sample size. From the extracted number of features, meaningful discriminating variables are determined. Therefore, the problems existing in the prior art are effectively avoided, the compiling efficiency of the VR psychological test scale is improved, and the accuracy of the scale is not lost.

Further, training the artificial intelligent model based on the calibration set data and the normal set data to adjust parameters of the artificial intelligent model, and screening out candidate psychological characteristics with larger weight, wherein the method specifically comprises the following steps: training and parameter adjustment are carried out on the artificial intelligent model based on the calibration set data and the normal set data, mutual information scores among all features in the primary psychological features are calculated, and the primary psychological features are ordered to screen out candidate psychological features with larger weights.

According to the scheme, the candidate psychological characteristics with larger weight can be efficiently screened out, and then a plurality of scenes with distinguishing characteristics corresponding to the candidate psychological characteristics can be obtained to serve as VR scenes of a certain psychological characteristic, so that the certain psychological characteristic of the subject can be tested. Compared with the existing paper pen test which is time-consuming and tedious, and has the inconvenience that a trained main test is necessary to observe a subject, the VR scene determined by the scheme is objective and convenient to operate; meanwhile, the same psychological trait can be tested by designing different types of VR scenes, and consistent test results can be obtained through a plurality of VR scenes. The analysis of the content of the item made only according to the result of one prediction is insufficient, and in order to check whether the performance of the selected content of the item really meets the requirement, another dial sample from the same population is sometimes selected for one time to check whether the results of the two analyses are consistent, and the process is called rechecking. The above-described procedure is much simpler than the current paper pen test scales.

Further, the ranking of the preliminary psychological characteristics is specifically: and sorting the preliminary psychological characteristics based on the mutual information score through a minimum redundancy maximum correlation algorithm or a minimum absolute value contraction and selector method so as to ensure the optimal selection of the characteristics.

Further, the determination of the artificial intelligence model specifically includes the following steps: selecting a plurality of machine learning models, and constructing a plurality of preliminary models based on the plurality of machine learning models; taking calibration group data and normal group data with the psychological characteristics as the input of a plurality of preliminary models, carrying out parameter optimization on the plurality of preliminary models, carrying out performance evaluation on the plurality of preliminary models by adopting a cross-validation algorithm, and obtaining statistical data of the plurality of preliminary models; and selecting an optimal preliminary model based on the statistical data, and determining the optimal preliminary model as an artificial intelligent model, wherein the VR scene which is screened by the model and corresponds to the candidate psychological characteristics with earlier obtained sequences is the scene of the finally determined scale.

Further, the selecting and determining the optimal preliminary model based on the statistical data is performed as an artificial intelligence model, specifically: drawing a plurality of ROC curves corresponding to the preliminary models based on the statistical data; obtaining the average area AUC values under the plurality of the initial model ROC curves; the optimal preliminary model is determined by the AUC values and is determined as an artificial intelligence model.

The VR psychological test scale programming method provided by the above scheme is performed according to the principle of combining psychological related theory, clinical priori knowledge and computer information theory, and comprises the following steps: determining a psychological trait to be measured; based on the psychological theory and clinical knowledge of the psychological trait, acquiring the dimension and the frame of the psychological trait; constructing a large amount of original VR scene item contents based on the psychological trait dimension and the frame; based on different machine learning models, respectively training and verifying the sample with the psychological trait and the normal sample to obtain the distinguishing characteristics capable of representing the sample data of the psychological trait: the training set carries out machine learning model parameter optimization, completes feature sequencing and optimization selection, obtains ROC curves through cross verification, and compares the performances of different machine learning models; selecting an optimal machine learning model based on prediction Accuracy (AUC) of different artificial intelligent models, and obtaining VR scene item contents which are preferably selected from the machine learning model and have the psychological characteristics; and finally, determining the VR scene item content of the VR psychological test scale and the corresponding artificial intelligent model, and judging the measurement result through the artificial intelligent model algorithm.

The method takes psychological related knowledge, clinical priori knowledge and computer information theory as guidance, optimizes and selects effective information from a large number of original VR scene items, obtains VR scene items which can most effectively represent corresponding psychological characteristics, and completes the establishment of various VR psychological test tables.

The invention also provides a VR psychological test meter compiling system, which is used for realizing the VR psychological test meter compiling method, and comprises the following steps: the psychological trait determining module is used for determining and inputting a psychological trait; the primary psychological characteristic acquisition module is used for acquiring primary psychological characteristics based on psychology-related theory and clinical priori knowledge of the psychological characteristics; the VR scene construction module is used for constructing original VR scene item content based on the primary psychological characteristics; the test and input module is used for respectively testing the calibration group and the normal group with the psychological characteristics in the original VR scene item content to obtain calibration group data and normal group data; and the VR scene screening module is used for selecting and optimizing VR scene item contents which most represent the psychological characteristics from the original VR scene item contents based on the calibration group data and the normal group data and computer information theory, and determining the VR psychological test list item contents which are the psychological characteristics.

The system provided by the scheme is simple in construction and convenient to realize, and the psychological characteristics are optimally selected by taking psychological related theory, clinical priori knowledge and computer information theory as guidance, so that the pertinence of the content of the compiled VR psychological test table can be effectively improved, and the compiling efficiency of the VR psychological test table and the test accuracy of the VR psychological test table are improved.

Further, the preliminary psychological characteristic acquisition module includes an effective information unit and a duplicate information unit, wherein: the effective information unit is used for acquiring effective information characteristics capable of representing psychological characteristics based on a psychological related theory; the duplicate information unit is used for acquiring duplicate information features which can represent psychological characteristics from the recognized psychological test table based on clinical priori knowledge; the effective information features and the duplicate information features constitute preliminary psychological features.

Further, the VR scene construction module constructs original VR scene item content based on the preliminary psychological characteristics, where the original VR scene item content includes a scene design and a content design, specifically: the scene is designed to enable the VR scene item content to reflect the text, sound, pictorial scene, smell or touch of the measured psychological characteristics; the content is designed to enable VR scene item content to reflect daily life, social scenes, special scenes, operational quiz scenes, or projective quiz scenes.

Further, the VR scene screening module includes an artificial intelligence module and a scene determination module, wherein: the artificial intelligent module is used for training the artificial intelligent model based on the calibration set data and the normal set data so as to adjust parameters of the artificial intelligent model and screen out candidate psychological characteristics corresponding to larger weight; the scene determining module is used for determining the VR scene item content of the psychological trait based on the candidate psychological characteristics corresponding to the larger satisfying weight from the original VR scene item content, and determining the VR psychological test list item content of the psychological trait.

Further, the artificial intelligence module comprises a mutual information calculation unit and a preliminary psychological characteristic sorting unit; wherein: the mutual information calculation unit is used for calculating mutual information scores among all the features in the preliminary psychological features; the preliminary psychological characteristic sorting unit is used for sorting the preliminary psychological characteristics based on the mutual information score so as to ensure the optimal selection of the characteristics.

Further, the artificial intelligence module comprises a mathematical statistical model construction unit, a model training unit and a cross verification unit; wherein: the mathematical statistical model construction unit is used for selecting a plurality of machine learning models and constructing a plurality of preliminary models based on the plurality of machine learning models; the model training unit is used for taking the sample data with the calibration set and the normal set data as the input of a plurality of preliminary models so as to perform parameter optimization on the plurality of preliminary models; the cross verification unit is used for performing performance evaluation on the plurality of preliminary models by adopting a cross verification algorithm, acquiring statistical data of the plurality of preliminary models, selecting an optimal preliminary model based on the statistical data, and determining the optimal preliminary model as an artificial intelligent model.

Further, the cross-validation unit comprises an ROC curve drawing subunit and an AUC value calculating subunit; wherein: the ROC curve drawing subunit is used for drawing ROC curves corresponding to the plurality of preliminary models based on the statistical data; the AUC value calculation subunit is used for obtaining the average area AUC values under the plurality of initial model ROC curves.

Further, the scene determining module comprises a model comparing unit and a candidate feature screening unit, wherein the candidate feature screening unit is used for carrying out training and parameter adjustment on the artificial intelligent model based on calibration group data, normal group data and the sorting result of the preliminary psychological features in the training process of the artificial intelligent model so as to screen out candidate psychological features corresponding to larger weights; the model comparison unit is used for determining an optimal preliminary model through AUC values and selecting the optimal preliminary model as an artificial intelligent model.

The present invention also provides a data processing apparatus comprising: the device comprises a processor and a memory coupled with the processor, wherein the memory stores a program, and the program is executed by the processor to cause the data processing device to execute a VR psychological test meter programming method.

The invention also provides a computer storage medium which stores computer instructions for executing the VR psychological test meter programming method.

Drawings

FIG. 1 is a flowchart of a method for creating a VR psychological test meter according to an embodiment of the present invention;

FIG. 2 is a flowchart illustrating steps for determining an optimal artificial intelligence model according to an embodiment of the present invention;

FIG. 3 is a flowchart illustrating a step of selecting an optimal preliminary model as an artificial intelligence model based on statistical data according to an embodiment of the present invention;

fig. 4 is a schematic diagram of a VR psychological test meter compiling system according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Psychological tests are a type of relative behavior that merely compares an individual with the behavior of the majority of the individuals in the community, or some "artificially determined criteria", and finally categorizes based on the comparison to determine psychological test results based on the categorization results.

Referring to fig. 1, the present embodiment provides a VR psychological test meter compiling method, which includes the following steps:

step S1: determining psychological traits. Step S1 is to select and determine a psychological characteristic to be tested to compile a corresponding VR psychological test chart.

Step S2: based on the psychology-related theory and clinical priori knowledge of the psychological traits, preliminary psychological characteristics are obtained.

Step S3: constructing original VR scene item content based on the preliminary psychological characteristics; the number of the content of the original VR scene item is far greater than the number of the VR scenes finally determined, and the variety of the test contents is also realized due to the richness of the scene layers, so that the bias of test results is avoided on one hand, and the representativeness of samples can be improved on the other hand.

Step S4: in the original VR scene item content, a sample (calibration set) and a normal sample (normal set) with the psychological characteristics are tested respectively to obtain calibration set data and normal set data. In step S4, the sample with psychological trait refers to a crowd who has confirmed to have a certain psychological trait, and responds to the scene content in the original VR scene item content, so as to complete the test and obtain psychological trait sample data. The normal sample refers to a common crowd without a certain psychological characteristic, and normal sample data can be acquired similarly.

It should be noted that, the obtained sample data is an evaluation index for a certain psychological trait, which can show psychological tendencies or expectations of the subject to respond to the VR scene, and can be used for qualitative and quantitative determination of the current perception and the nature and direction of the response. Wherein the evaluation index may be acquired by a biometric sensor or a physiological characteristic sensor, and the corresponding sensor may measure or detect heart rate, pulse, respiration rate, temperature, pupil dilation, eye movement, pressure level, brain electrical and functional magnetic resonance, etc. by being attached to or in proximity to the body of the subject.

Specifically, the evaluation index may feedback the tendency, accuracy, or reaction speed of the subject to answer the selection question, etc.: counting accumulated scores, accuracy and the like of the scene task stimulus response or the selected tendency, and acquiring behavior indexes such as response speed, correct number and the like by using dynamic capture equipment; the evaluation index can be obtained through an eye movement tracking device, such as obtaining the interest content and the region of the subject, recording the response time, the attention span and the like; the evaluation index can be obtained by a skin resistance device, and the neurophysiologic index reflecting emotion, sympathology and parasympathetic characteristics and the like is obtained as the evaluation index by a sensor for measuring Galvanic Skin Response (GSR) or galvanic skin activity; the evaluation index can be obtained by monitoring heart rate and heart rate variability equipment, and can comprise a photoelectric volume pulse wave (PPG) sensor which measures blood volume change in microvascular tissues so as to reflect pulse rate, and the electric activity of the heart can be sensed by the sensor in the heart rate monitor so as to reflect heart rate or reflect physiological indexes such as emotion pressure; the evaluation index can be obtained through an electroencephalogram device, such as obtaining electroencephalogram data reflecting attention, emotion, memory and the like so as to reflect the corresponding neurophysiologic index; the evaluation index can also be obtained by brain function imaging equipment, such as brain function magnetic resonance and other metabolite imaging reflecting nerve physiological indexes such as sensory perception, thinking, memory and the like. The evaluation index is only an illustration of this embodiment, and is not unique, so the description is omitted here.

Step S5: and selecting and optimizing VR scene item contents which most represent the psychological characteristics from the original VR scene item contents based on the calibration group data and the normal group data and determining the VR psychological test list item contents which are the psychological characteristics.

In this embodiment, the principles of psychological correlation theory and clinical priori knowledge are used as guidance, so that the dimension and the framework of the psychological characteristics can be determined, and then the psychological characteristics are optimally selected, so that the pertinence of the content of the compiled VR psychological test table can be effectively improved, and the compiling efficiency and the testing accuracy of the VR psychological test table can be improved.

It should be noted that, in this embodiment, people who have been recognized as different types of psychological traits are selected and used as experience targets. Then, people with certain psychological characteristics and normal people enter a VR environment to complete tasks or games, namely VR scene items are applied to a calibration group and a normal group, and people with predicted psychological characteristics form the calibration group and the normal group formed by common people. And comparing the responses of the two groups of tested items to each VR scene, such as the response tendency or task index of the subject, and the like, distinguishing the effective standard group from the normal group by utilizing an artificial intelligent model according to the data acquired by the evaluation index, and reserving the VR scene with the distinction degree to form the VR scene of a certain psychological quality test.

It should be noted that, the artificial intelligent model is essentially a classifier, which can determine the type according to the specific classification quantity, such as classification problem, SVM classifier can be selected, or neural network adopts sigmoid and other activation functions; for multi-classification problems, a softmax classifier may be employed.

Further, in acquiring the preliminary psychological characteristics based on psychology-related theory and clinical prior knowledge of psychological characteristics, the preliminary psychological characteristics include effective information characteristics and duplicate information characteristics, and the acquisition process specifically includes the following steps: based on the theory of psychological correlation, obtaining effective information characteristics capable of representing psychological characteristics; duplicate information features representing psychological traits are obtained from a recognized psychological test meter based on clinical prior knowledge.

VR scenes can be designed based on the valid information features and the duplicate information features and a large number of candidate scenes can be generated for screening. The scene content can well test certain psychological characteristics of the subjects, and the subjects can react based on the content in the scene, so that effective clustering is carried out through reaction data, and the method has good representativeness.

It should be noted that, the preliminary psychological characteristics obtained in this embodiment are obtained specifically according to the purpose, property and function of the psychological characteristics of the test, and the VR test scene and content design are performed according to the frame of certain dimensions (i.e. emphasis or category) of the psychological characteristics, so that a lot of characteristic variables can be generated, and at this time, the experiment of the psychological characteristics of multiple scenes can be performed through limited samples, so that the artificial intelligence model can learn, thereby effectively improving the efficiency of compiling the VR psychological test table.

Specifically, the effective information features are information that, according to psychology-related theory, extracts preliminary psychological features concerning a certain psychological trait and finds out therefrom representative features with distinction, which can widely characterize the psychological trait. The duplicate information features are to refer to the content of the existing recognized psychological test meter item and directly convert the content into a scene for VR content design, and can be used for VR scene conversion directly from the content and form by the traditional meter or can be used for replacing conversion but contains the same content of the traditional meter.

Further, in constructing the original VR scene item content based on the preliminary psychological characteristics, constructing the original VR scene item content includes a scene design and a content design, specifically: the scene design is used for enabling the content of VR scene items to reflect the text, sound and picture scenes (including various contents such as life, emotion, pressure and the like) of the detected psychological characteristics, and can also be matched with smell, touch and the like; the content design is used for enabling the VR scene item content to reflect daily life and social scenes, special scenes, operation test scenes or projection test scenes. The content design has the characteristic of presenting more psychological characteristics, thereby improving the representativeness of psychological experiments.

In this embodiment, the construction of the VR scene item content is based on a preliminary psychological characteristic, and after the preliminary psychological characteristic is obtained by taking psychology-related theory and clinical priori knowledge as guidance, the preliminary psychological characteristic is compiled and the corresponding VR scenes and the display content of the scenes are designed, so that the VR scene item content can present a certain psychological characteristic of the test, and the reactivity of the obtained test sample, that is, the sample with the psychological characteristic is representative and can be measured by an objective index.

Further, in the calibration group data and the normal group data, VR scene item contents which are most characterized by the psychological characteristics are selected and optimized from the original VR scene item contents based on a computer information theory, and VR psychological test table item contents which are the psychological characteristics are determined, wherein the calibration group data and the normal group data are specifically realized by adopting an artificial intelligent model based on the computer information theory, and specifically comprises the following steps:

training and parameter adjustment are carried out on the artificial intelligent model based on the calibration set data and the normal set data, the preliminary psychological characteristics are ordered, and candidate psychological characteristics corresponding to large weights are screened out;

and determining the VR scene item content of the psychological trait from the candidate psychological characteristics with larger weights, and determining the VR psychological test scale item content of the psychological trait.

Further, training and parameter adjustment are carried out on the artificial intelligent model based on the calibration set data and the normal set data, and candidate psychological characteristics corresponding to larger weight are screened out, specifically:

training the artificial intelligent model based on the calibration set data and the normal set data, adjusting parameters by the artificial intelligent model, calculating mutual information scores among all features in the preliminary psychological features, and sorting the preliminary psychological features to screen out candidate psychological features with larger weights.

According to the method and the device, the candidate psychological characteristics with larger weight can be efficiently screened out, and then a plurality of scenes with distinguishing characteristics corresponding to the candidate psychological characteristics can be obtained to serve as VR scenes of a certain psychological characteristic, so that the certain psychological characteristic of a subject can be tested. Compared with the existing paper pen test which is time-consuming and tedious, and has the inconvenience that a trained main test is necessary to observe a subject, the VR scene determined by the embodiment is objective and convenient to operate; meanwhile, the same psychological trait can be tested by designing different types of VR scenes, and consistent test results can be obtained through a plurality of VR scenes. It is not enough to make a test according to the content analysis of the item only based on the result of one prediction, and in order to check whether the performance of the selected item content really meets the requirement, another dial sample from the same population is sometimes selected and tested again to see whether the two analysis results are consistent. This process is called rechecking. The above embodiment is much simpler than the current pen test meter.

In the embodiment, the preliminary psychological characteristics are ranked through mutual information scores, and the optimal psychological characteristics are selected to be used as subsequent input, so that the statistical efficacy can be reduced, and the most significant characteristics are obtained; the psychological characteristics are ranked through mutual information scores, so that the characteristics with the highest information quantity can be identified, and the effective characteristics are reserved. And for the selection of psychological characteristics, the method can search in all psychological characteristic spaces and keep sparsity, can improve the accuracy of the finally constructed artificial intelligent model, and ensures the prediction accuracy of the VR scene constructed by the artificial intelligent model. In the above embodiment, the minimum redundancy maximum correlation algorithm or the minimum absolute value contraction and selector method can be utilized to screen out the highly relevant psychological characteristics with low redundancy. By analyzing all psychological characteristics in advance and sorting the psychological characteristics of the meaningful information according to the mutual information score, the optimal psychological characteristics can be found, and the optimal psychological characteristic selection of multiple parameters in the VR scene is realized so as to improve the effectiveness of the test scale.

Further, referring to fig. 2, the determining of the artificial intelligence model specifically includes the following steps:

selecting a plurality of machine learning models, and constructing a plurality of preliminary models based on the plurality of machine learning models;

Taking calibration group data and normal group data with the psychological characteristics as the input of a plurality of preliminary models, carrying out parameter optimization on the plurality of preliminary models, carrying out performance evaluation on the plurality of preliminary models by adopting a cross-validation algorithm, and obtaining statistical data of the plurality of preliminary models;

an optimal preliminary model is selected based on the statistical data and determined as an artificial intelligence model.

In this embodiment, a cross-validation algorithm may be used to verify accuracy, sensitivity, specificity, false positives, false negatives to evaluate the performance of the model.

It should be noted that, before determining the artificial intelligence model, normalization processing may be performed on the preliminary psychological characteristics to ensure consistency of the characteristic forms. The cross-validation algorithm may be a 5-fold (or 10-fold) cross-validation algorithm, which is one method for evaluating the performance of a machine learning model by testing the trained machine learning model on new different data sets (i.e., different inputs) to validate the model. The performance evaluation can also be realized by comparing the data sets under different evaluation algorithms, such as application of methods such as Neural Networks (NN) and Support Vector Machines (SVM) to different psychological trait classifications, and comparison verification among a plurality of data sets.

Further, referring to fig. 3, the optimal preliminary model is selected based on the statistical data and determined as an artificial intelligence model, specifically:

drawing a plurality of ROC curves corresponding to the preliminary models based on the statistical data;

obtaining the average area AUC values under the plurality of the initial model ROC curves;

the optimal preliminary model is determined by the AUC values and is determined as an artificial intelligence model.

In this embodiment, performance comparison and verification of different machine learning models is achieved by obtaining ROC curves of the different machine learning models. Wherein the AUC values can visually represent the accuracy of the machine learning model, thereby enabling comparison of different machine learning models.

In order to further describe the above technical process, in this embodiment, the selected plurality of machine learning models may include machine learning models such as GLM, RF, SVM or Lasso, by performing data training by using a plurality of machine learning models and performing evaluation comparison of model performance by using a cross-validation algorithm, the best choice of the finally determined artificial intelligence model may be ensured, and further, a VR scene having the best distinction degree for representing a certain psychological trait may be corresponding to the candidate psychological trait in the optimal artificial intelligence model, and the artificial intelligence model may be determined as a VR psychological test table classifier, so that people with the specific psychological trait can be accurately distinguished by using fewer features, thereby facilitating the performance of the VR psychological test, greatly reducing the work difficulty of the VR psychological test, and reducing the psychological test cost.

The VR psychological test scale compiling method provided in the above embodiment is performed according to the principles of the clinical priori knowledge of the psychology-related theory and the computer information theory, and includes the following steps: determining a psychological trait to be measured; based on the psychological theory and clinical knowledge of the psychological trait, acquiring the dimension and the frame of the psychological trait; constructing a large amount of original VR scene item contents based on the psychological trait dimension and the frame; based on different machine learning models, respectively training and verifying the sample with the psychological trait and the normal sample to obtain the distinguishing characteristics capable of representing the sample data of the psychological trait: the training set carries out machine learning model parameter optimization, completes feature sequencing and optimization selection, obtains ROC curves through cross verification, and compares the performances of different machine learning models; selecting an optimal machine learning model based on prediction Accuracy (AUC) of different artificial intelligent models, wherein VR scenes corresponding to candidate psychological characteristics which are screened out from the machine learning model and are obtained in a front sequence are the scenes of a finally determined scale; and finally, determining the VR scene item content of the VR psychological test table and the corresponding artificial intelligent model.

The above embodiments can efficiently create test tables of various psychological traits using VR technology and artificial intelligence technology. According to the method, the VR scene project kernel can be determined by only a small sample, and the measurement result is judged through an artificial intelligent model algorithm; and the scene content is subjected to feature sequencing through extraction of psychological features and mutual information score, so that screening and revision of the VR scene are realized, and finally, the optimal VR scene item content and an optimal artificial intelligent model algorithm are determined.

Based on the method provided by the embodiment, if enough samples with certain psychological traits exist, the samples can be reliably classified, and then a screening or evaluating tool for the psychological traits can be obtained.

Further, in order to make the test meter reasonably practical and applicable, a test manual may be compiled, for example, the field, purpose, function and other specifications to be measured by the test meter; and whether the measurement results of each test meter can distinguish between different parties defined by the calibration measurements (tag data). The test may be considered valid if the measurement results better distinguish between different calibration communities.

Referring to fig. 4, the present embodiment also provides a VR psychological test meter compiling system, configured to implement the above VR psychological test meter compiling method, which includes: the psychological trait determining module is used for determining and inputting a psychological trait; the primary psychological characteristic acquisition module is used for acquiring primary psychological characteristics based on psychology-related theory and clinical priori knowledge of the psychological characteristics; the VR scene construction module is used for constructing original VR scene item content based on the primary psychological characteristics; the test and input module is used for respectively testing the calibration group and the normal group with the psychological characteristics in the original VR scene item content to obtain calibration group data and normal group data; and the VR scene screening module is used for selecting and optimizing VR scene item contents which most represent the psychological characteristics from the original VR scene item contents based on the calibration group data and the normal group data and computer information theory, and determining the VR psychological test list item contents which are the psychological characteristics.

The system provided by the embodiment is simple in construction and convenient to realize, and the psychological characteristics are optimally selected by taking psychological related theory, clinical priori knowledge and computer information theory as guidance, so that the pertinence of the content of the compiled VR psychological test table can be effectively improved, and the compiling efficiency of the VR psychological test table and the testing accuracy of the VR psychological test table are improved.

Further, the preliminary psychological characteristic acquisition module includes an effective information unit and a duplicate information unit, wherein: the effective information unit is used for acquiring effective information characteristics capable of representing the psychological characteristics based on the theory of the related psychology; the duplicate information unit is used for acquiring duplicate information characteristics capable of representing the psychological characteristics from a recognized psychological test table based on clinical priori knowledge; the effective information features and the duplicate information features constitute preliminary psychological features.

Further, the VR scene construction module constructs original VR scene item content based on the preliminary psychological characteristics, where the original VR scene item content includes a scene design and a content design, specifically:

the scene is designed to enable the VR scene item content to reflect the text, sound, pictorial scene, smell or touch of the measured psychological characteristics;

The content is designed to enable VR scene item content to reflect daily life, social scenes, special scenes, operational quiz scenes, or projective quiz scenes.

Further, the VR scene screening module includes an artificial intelligence module and a scene determination module, wherein: the artificial intelligent module is used for training the artificial intelligent model based on the calibration set data and the normal set data so as to adjust parameters of the artificial intelligent model and screen out candidate psychological characteristics with larger weight; the scene determination module is used for determining the VR scene item content of the psychological trait from the candidate psychological characteristics with larger weights and determining the VR psychological test list item content of the psychological trait _。

Further, the artificial intelligence module comprises a mathematical statistical model construction unit, a model training unit and a cross verification unit; wherein: the mathematical statistical model construction unit is used for selecting a plurality of machine learning models and constructing a plurality of preliminary models based on the plurality of machine learning models; the model training unit is used for taking the sample data with the calibration set and the normal set data as the input of a plurality of preliminary models so as to perform parameter optimization on the plurality of preliminary models; the cross verification unit is used for performing performance evaluation on the plurality of preliminary models by adopting a cross verification algorithm, acquiring statistical data of the plurality of preliminary models and transmitting the statistical data to the scene determination module.

Further, the scene determining module comprises a model comparing unit and a candidate feature screening unit, wherein the candidate feature screening unit is used for adjusting parameters of the artificial intelligent model based on calibration group data, normal group data and the sorting result of the preliminary psychological features in the training process of the artificial intelligent model so as to screen candidate psychological features corresponding to larger weights; the model comparison unit is used for determining an optimal preliminary model through AUC values and selecting the optimal preliminary model as an artificial intelligent model.

The present embodiment also provides a data processing apparatus including: the device comprises a processor and a memory coupled with the processor, wherein the memory stores a program, and the program is executed by the processor to cause the data processing device to execute a VR psychological test meter programming method.

The embodiment also provides a computer storage medium, and the computer storage medium stores computer instructions for executing the VR psychological test meter compiling method.

The VR psychological test meter programming method, system, device and storage medium provided in the above embodiments do not need project difficulty and distinction analysis in the traditional paper pen meter programming process; the obtained VR psychological test meter is combined with the VR scene in the actual application process, oral feedback is not needed, and the measurement result can be intelligently obtained, so that the test efficiency is greatly improved, and the test cost and difficulty are obviously reduced; the VR psychological test scale is simple in programming process, time-saving and low in cost.

It should be noted that, the tested behavior is observed in the VR scene, which is more realistic and natural and is not easy to be imitated. In the VR environment, the content is comprehensive and deep, the tested reaction is free, the tested reaction can be fully expressed, no intention is kept or anxiety is caused, and the reflection is more true; the testing process of instruction, environment and the like is standardized, the score of the measurement result is standardized, and the score interpretation is standardized, so that the stability and objectivity of the measurement system can be ensured, and the random error and the system error are reduced, thereby having better credibility.

The above embodiment has a good degree of effectiveness, and if a test has a high correlation with a calibration (tag data), the nature and kind of calibration predicted by the test can be used as the conception index measured by the test. The conventional evaluation of the content effectiveness of the paper pen test is an expert judgment method or an empirical method, often a qualitative analysis method, and the above embodiment has reliable quantitative indexes compared with the expert empirical method, and can be quantitatively analyzed.

In order to further illustrate the technical features of the present invention, and to highlight the technical advantages thereof, the present embodiment is exemplified by a mental scale programming method for personality testing. Before the description, the definition and determination of psychological traits should be described:

and (3) a step of: psychological traits refer in particular to regularity in the behavior and consistency in a broad range of behaviors. In view of this feature, psychological traits may represent basic categories of functional individual differences. Thus, when a person is described as a straight person, it is in describing a common characteristic of the person and distinguishing the person from those with shy characteristics. And II: psychological traits can be used as the basic unit of personality. Based on this, the task of the psychologist of the personality is to find out the basic characteristics of the personality and to design individual differences that occur in a specific scenario, giving a satisfactory interpretation. While psychological traits cannot be directly observed, observing a person's repetitive behavior can indirectly infer its presence.

Through the patterns or the environment of the "energized" of artists and psychologists, the characteristics of the life experience, emotion, personality tendency and the like of the subjects can be induced, and the subjects can present the respective psychological characteristics in VR scenes. There is some relationship between the person's perception and personality, and the person's perception of the stimulus throws out of his personality, similar to a projection test. Projection refers to a psychological effect in which an individual does not voluntarily reflect his or her mind, attitude, wish, emotion, or character, etc., on an outside or other person. The projective test is to provide the subject with several unorganized stimulation scenarios that allow the subject to freely express the subject's response in an unrestricted scenario.

The basic assumptions of the projection test are: (1) the human response to external stimuli is causal and predictable, rather than occasional; (2) the individual's response is naturally dependent on the current stimulus and situation, but the individual's current mental state, existing experience, future expectations, etc., will have a great effect on the current perception and the nature and direction of the response; (3) most of the personality structure is in subconscious, and the individual cannot explain himself by consciousness, but when the individual faces an unknown stimulation situation, the desire, demand, motivation, conflict and the like hidden in subconscious can be revealed.

Thus, people who have been recognized as different types of psychological traits can be selected and used as experience targets. Then, the psychological trait to be detected and the ordinary person enter the VR scene to complete tasks or games, and the responses of the two groups of subjects to the VR scene content, such as the response tendency of the subjects, task indexes and the like, are compared, so that the VR scene task of the effective standard group and the VR scene task of the normal group can be distinguished, and the VR scene item content of the psychological test scale of the psychological trait to be detected is formed.

The embodiment uses the psychological scale compiling method of personality test as an example, according to the personality theory of psychologists, the conscious functions of the person are divided into 8 parts and 8 models, namely Rong Geba dimension theory, the content structure of which is 6 aspects, including 2 kinds of potential and 4 kinds of thought functions. Wherein, the fixed potential refers to the characteristic direction of the multi-movement, including the inward inclination and the outward inclination, namely the inward and outward directions frequently mentioned in the daily life of people. The psychological castration of the person who leans inwards is directed to the individual's internal world, while the psychological castration of the person who leans outwards is directed outwards. While thought functions refer to typical ways of individuals understanding internal and external stimuli, two main categories of functions are divided: rational and irrational. There are two functions of irrational nature, sensory (sense) -intuitiveness (point); there are two functions involved in rationality, thought (thinking) -emotion (felt). The method specifically comprises the following four steps: (1) feel (sense): indicating where the thing is present; (2) intuition (point): rough inferences about past and future events when insufficient information is utilized; (3) thinking (thinking): explaining what the perceived object is, and naming it; (4) emotion (emotion): reflecting whether the thing is acceptable to the individual or not, determines what value to the individual.

In fact, everyone often exhibits a function and a predominance of potentials, all else being unconscious. Each person's type is composed of these six permutations of functions, which creates 8 psychological types: camber feeling type, and toe-in feeling type; an external-tilt intuition type, an internal-tilt intuition type; camber thinking type, camber thinking type; camber emotion, and toe-in emotion. Wherein:

irrational function:

camber feeling: sensitive to the external environment, enjoys contact with the external environment and self-experience; feel of inward inclination: can continuously accumulate experience in growth, follow tradition and have short predictions in the future; camber intuition: the relationship among things can be found, the thinking is jumped and diverged, and the method has strong predictability in the future; intuition of internal inclination: having a convergent thinking is good at refining the nature and core of things, enabling a realistic conclusion to be formed.

Rational function:

camber thinking: the problem can be effectively solved, and the organization structure and the logic of things are emphasized, and the method is characterized in that the things are acted according to fixed rules, and the problems, the armed, the emotion and the depression are objectively and calmly considered actively; internal tilting thinking: is enthusiastically in abstract thinking and creating new theory, has good sense of number and space ability, and pays attention to the principle of things; the method has the tendency of neglecting daily life practice, is rarely influenced by external things to make decisions, and is firm, intractable and indifferent; camber emotion: emotion depends on the outside, focuses on the harmony of social requirements and morals, and wants to care for the feeling of each person; introverted emotion: emotion is independent, attention is paid to the value of the heart and the bottom line of the heart, and the heart sounds are listened to.

In this embodiment, six dimensions (2 kinds of setups, 4 kinds of thought functions) are used as preliminary psychological characteristics to perform VR scene design, and specific reference may be made to table 1:

table 1: candidate psychology characteristic table

And then in a plurality of machine learning models, based on the screened candidate psychological characteristics, comparing the AUC of each machine learning model, and further determining an optimal artificial intelligent model, wherein the VR scene which is screened by the model and corresponds to the candidate psychological characteristics with the earlier obtained rank is the scene of the finally determined scale. Specifically:

regarding the construction of VR scenes: VR scene item content is derived from daily life and social scenes, special scenes or objective materials (e.g., operation tests, projection tests, etc.). The questions are answered in 6 dimensions classified by the Rongge personality, each dimension, or the tasks are operated on, to form data (clustered questions) of 6 answer reaction sets, such as 80 entries as scene content.

Acquisition of evaluation index: indexes such as tendency, accuracy and response speed of response (behavior) to 6-dimensional content are selected, and the current perception, the nature and direction of response and the like of the subject are qualitatively and quantitatively determined.

Determination of mental feature ordering, selection, and artificial intelligence model: sorting the psychological characteristics through the mutual information scores of the preliminary psychological characteristics, and preferentially selecting the candidate psychological characteristics with the front sorting as the item content basis of the test scale; and simultaneously, determining an optimal artificial intelligent model from a plurality of machine learning models by utilizing the ROC curve, wherein the VR scene which is screened by the model and corresponds to the candidate psychological characteristics with earlier obtained sequences is the scene content of the finally determined scale.

After the VR scene item content and the artificial intelligence model are determined, the preparation of VR psychological test table is completed, and the measurement result is judged through the artificial intelligence model algorithm.

The present embodiment chooses a psychology-related theory as a framework for the design and feature selection of the guided project. A certain dimension frame is selected to generate a plurality of psychological characteristic variables, and the characteristics are ordered and selected, wherein the theoretical basis is the correlation index of 6 dimension characteristics in Rong Geba dimensions. By means of computer information theory, through feature sorting and feature selection of 6 feature relevant indexes and training and learning by selecting several machine learning models, one optimized artificial intelligent model may be constructed to distinguish various personality characteristic persons from normal persons.

Such psychologically related theory-oriented feature selection structures can incorporate the most useful discriminative predictors into the final artificial intelligence model, using computer information theory to remove less informative or redundant variables, creating an algorithmic model with accuracy (sparse predictive model). In addition, with mutual information, one can flexibly integrate clinical prior knowledge into a feature selection model.

Specifically, the present embodiment takes 100 samples as input and trains and validates 6 machine learning models using a 10-fold cross validation algorithm. Suppose that 60 of the 100 samples are people of a certain psychological trait (e.g., a certain class of personality test) and 40 are ordinary people. They complete VR scene tests, such as 80 item contents (i.e., preliminary test meter item contents) designed using the eight-dimensional principle of rong, and acquire corresponding evaluation index data. Training and verification are carried out through 6 machine learning models, parameter adjustment is carried out through evaluation index data training, mutual information ordering is carried out, and optimized selection is carried out on the corresponding 80 item contents. Finally, assuming that under the SVM machine training model, there are 30 items in the 80 VR items, which are sufficient to distinguish the person with the psychological trait from the ordinary person with high accuracy, so as to obtain the feature items and the number with the distinction, then the 30 items are the items in the VR psychological test scale finally, and the SVM machine training model is the artificial intelligence model of the psychological trait test. According to the method, VR scenes corresponding to other 50 items which have no distinguishing significance or are not strong in distinguishing degree can be deleted, and the time efficiency of evaluation can be improved.

For each machine learning model, 10 fold cross-validation was performed, i.e., 10 sets of experiments were performed for each machine learning model, where classifier training could be performed using 90% of the samples (i.e., 9 sets), feature ordering of mutual information was performed with training on 9 sub-sets of samples, and the remaining 10% of the samples (i.e., 1 set) were used for validation. For each machine learning model, the model is subjected to parameter optimization adjustment, namely parameter adjustment training, by using default parameters and ordered features, the performance of each machine learning model is recorded, namely the construction of a classifier is performed, and the sensitivity, the specificity and the accuracy of the classifier are measured. In this process, 10-fold cross-validation experiments were repeated for each machine-learning model to obtain the average area under ROC curve (AUC) thereof. All machine learning analyses were performed in Python using the Scikit-learn software package. Therefore, ROC curves of each machine learning model can be drawn, the accuracy of each machine learning model can be obtained through AUC, and further performances of different machine learning models can be evaluated.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.

Claims

1. A VR psychological test meter programming method, comprising the steps of:

determining a psychological trait;

acquiring preliminary psychological characteristics based on psychology-related theory and clinical priori knowledge of the psychological traits;

constructing original VR scene item content based on the preliminary psychological characteristics;

in the original VR scene item content, testing a calibration group and a normal group with the psychological characteristics respectively to obtain calibration group data and normal group data;

and selecting and optimizing VR scene item contents which most represent the psychological characteristics from the original VR scene item contents based on the calibration group data and the normal group data and determining the VR psychological test list item contents which are the psychological characteristics.

2. The VR psychological test scale programming method of claim 1, wherein the psychological trait comprises personality, intelligence, neuropsychology, child behavior, or mental health.

3. The VR psychological test scale programming method according to claim 1, wherein in said obtaining preliminary psychological characteristics based on psychology-related theory and clinical prior knowledge of said psychological traits, said preliminary psychological characteristics include valid information characteristics and duplicate information characteristics, and the obtaining process specifically includes the steps of:

Based on a psychological correlation theory, obtaining effective information characteristics capable of representing the psychological characteristics;

duplicate information features representing the psychological trait are obtained from a generally accepted psychological test form based on clinical prior knowledge.

4. The VR psychological test meter programming method according to claim 1, wherein in said constructing the original VR scene item content based on the preliminary psychological characteristics, said constructing the original VR scene item content includes a scene design and a content design, specifically:

5. The VR psychological test meter compiling method according to claim 1, wherein, in the VR psychological test meter item content that is most characterized by the psychological characteristics and is determined to be the psychological characteristics, which is selected and optimized from the original VR scene item content based on the calibration group data and the normal group data, based on a computer information theory, the computer information theory is implemented specifically by using an artificial intelligence model, specifically:

Training the artificial intelligent model based on the calibration set data and the normal set data to adjust parameters of the artificial intelligent model, sequencing the primary psychological characteristics and screening candidate psychological characteristics corresponding to the high weight;

6. The VR psychological test scale programming method according to claim 5, wherein the training of the artificial intelligent model based on the calibration set data and the normal set data to tune the artificial intelligent model, and the screening of the candidate psychological characteristics with larger weight specifically comprises:

7. The VR psychological test meter programming method of claim 6, wherein said ranking said preliminary psychological characteristics comprises:

And sorting the preliminary psychological characteristics based on the mutual information score through a minimum redundancy maximum correlation algorithm or a minimum absolute value contraction and selector method so as to ensure the optimal selection of the characteristics.

8. The VR psychological test meter formulation method according to any one of claims 5 to 7, wherein said artificial intelligence model determination comprises the steps of:

9. The VR psychological test scale programming method according to claim 8, wherein said selecting and determining an optimal preliminary model based on said statistical data as an artificial intelligence model is specifically:

10. A VR psychological test meter formulation system for implementing a VR psychological test meter formulation method as claimed in any one of claims 1 to 9 comprising:

the psychological trait determining module is used for determining and inputting a psychological trait;

the primary psychological characteristic acquisition module is used for acquiring primary psychological characteristics based on psychology-related theory and clinical priori knowledge of the psychological characteristics;

the VR scene construction module is used for constructing original VR scene item content based on the primary psychological characteristics;

the test and input module is used for respectively testing the calibration group and the normal group with the psychological characteristics in the original VR scene item content to obtain calibration group data and normal group data;

and the VR scene screening module is used for selecting and optimizing VR scene item contents which most represent the psychological characteristics from the original VR scene item contents based on the calibration group data and the normal group data and computer information theory, and determining the VR psychological test list item contents which are the psychological characteristics.

11. The VR psychological test meter formulation system according to claim 10, wherein said preliminary psychological characteristic acquisition module comprises a valid information unit and a duplicate information unit, wherein:

the effective information unit is used for acquiring effective information characteristics capable of representing the psychological characteristics based on the theory of the related psychology;

the duplicate information unit is used for acquiring duplicate information characteristics capable of representing the psychological characteristics from a recognized psychological test table based on clinical priori knowledge;

the effective information features and the duplicate information features constitute preliminary psychological features.

12. The VR psychological test meter formulation system according to claim 10, wherein said VR scene construction module constructs original VR scene item content based on preliminary psychological characteristics, said original VR scene item content comprising scene design and content design, in particular:

13. The VR psychological test meter programming system of claims 10-12, wherein the VR scene screening module comprises an artificial intelligence module and a scene determination module, wherein:

the artificial intelligent module is used for training the artificial intelligent model based on the calibration set data and the normal set data so as to adjust parameters of the artificial intelligent model and screen out candidate psychological characteristics with larger weight;

the scene determination module is used for determining the VR scene item content of the psychological trait from the candidate psychological characteristics with larger weights and determining the VR psychological test list item content of the psychological trait _。

14. The VR psychological test meter formulation system according to claim 13, wherein said artificial intelligence module comprises a mutual information calculation unit and a preliminary psychological characteristic ordering unit; wherein:

the mutual information calculation unit is used for calculating mutual information scores among all the features in the preliminary psychological features;

the preliminary psychological characteristic sorting unit is used for sorting the preliminary psychological characteristics based on the mutual information score so as to ensure the optimal selection of the characteristics.

15. The VR psychological test meter formulation system according to claim 14, wherein said artificial intelligence module comprises a mathematical statistical model construction unit, a model training unit, and a cross-validation unit; wherein:

The mathematical statistical model construction unit is used for selecting a plurality of machine learning models and constructing a plurality of preliminary models based on the plurality of machine learning models;

the model training unit is used for taking the sample data with the calibration set and the normal set data as the input of a plurality of preliminary models so as to perform parameter optimization on the plurality of preliminary models;

the cross verification unit is used for performing performance evaluation on the plurality of preliminary models by adopting a cross verification algorithm, acquiring statistical data of the plurality of preliminary models and transmitting the statistical data to the scene determination module.

16. The VR psychological test meter formulation system according to claim 15, wherein said cross-validation unit comprises an ROC curve drawing subunit and an AUC value calculating subunit; wherein:

the ROC curve drawing subunit is used for drawing ROC curves corresponding to the plurality of preliminary models based on the statistical data;

the AUC value calculation subunit is used for obtaining the average area AUC values under the plurality of initial model ROC curves.

17. The VR psychological test scale programming system of claim 16, wherein the scene determining module comprises a model comparing unit and a candidate feature screening unit, wherein the candidate feature screening unit is configured to train and tune the artificial intelligent model based on calibration group data, normal group data and the ranking result of the preliminary psychological features during the training of the artificial intelligent model, so as to screen candidate psychological features corresponding to a larger weight; the model comparison unit is used for determining an optimal preliminary model through AUC values and selecting the optimal preliminary model as an artificial intelligent model.

18. A data processing apparatus, comprising:

a processor and a memory coupled to the processor, the memory storing a program for execution by the processor to cause the data processing apparatus to perform a VR psychological test meter formulation method as set forth in any one of claims 1-9.

19. A computer storage medium having stored thereon computer instructions for performing a VR psychological test meter formulation method according to any one of the preceding claims 1 to 9.