Psychological health state projection assessment method and system based on AI identification

Abstract

The invention discloses a psychological health state projection assessment method and system based on AI recognition, and relates to the technical field of psychological health assesses.The method comprises the steps of analysis model construction, user psychological analysis and group psychological analys.The method comprises the steps that firstly, house tree person data of historical users without external intervention are collected through analysis model construction; building a fusion data AI analysis model through distributed GPU cluster training, then collecting house tree person drawing and dynamic behavior data of a target user and model output system state data through user psychological analysis, analyzing dynamic behaviors to obtain a psychological positive feedback coefficient, and outputting a psychological level after verification; and finally, through group psychological analysis, analyzing the state change of a target user system to identify an abnormal psychological state, summarizing user group data, and performing early warning according to the focused attention state when the user group is abnormal, objective and efficient evaluation is realized through AI, the precision is improved through two-dimensional verification, and the method is used for supporting individual and group psychological health management and control.

Description

Technical Field

[0001] This invention relates to the field of mental health assessment technology, specifically to a method and system for assessing mental health status projection based on AI recognition. Background Technology

[0002] The core technology system in the field of mental health assessment mainly includes three categories: questionnaire tests, projective tests, and situational tests. Among them, questionnaire tests and projective tests are the two most widely used technologies in clinical diagnosis, educational screening, and corporate management. As society's demand for mental health services increases, there is a need for a mental health status projective assessment method and system based on AI recognition.

[0003] Existing technologies, such as questionnaires, collect subjective feedback from test takers through standardized written questions to achieve quantitative assessment of psychological characteristics. The core defects of these technologies are: 1. Difficulty in ensuring the authenticity of responses: Test takers are easily influenced by social desirability bias and may deliberately conceal negative psychological characteristics such as anxiety, depression, and aggression, resulting in a significant deviation between the assessment results and the actual psychological state, which cannot meet the needs of accurate diagnosis.

[0004] 2. The applicable population is limited, relying on the test subject's language comprehension and written expression ability. It cannot cover young children, people with low education levels, people with language impairments, and cross-cultural groups, and the applicable scenarios have obvious boundaries.

[0005] 3. The assessment has a single dimension and is prone to fatigue. It focuses on a single explicit psychological trait and cannot simultaneously analyze deeper psychological characteristics such as personality structure and subconscious conflicts. At the same time, repeated administration can easily lead to test subject fatigue, and the reliability of the results decreases with the frequency of administration, making it difficult to meet the needs of high-frequency dynamic monitoring.

[0006] Existing technologies, such as projective tests, are based on Freud's projective theory. They avoid the subjective faking problem of questionnaire tests by analyzing the unconscious drawing responses of test subjects to vague stimuli such as houses, trees and people. The core defects of their technical solutions are: 1. The analysis process is highly dependent on human experience and has a low degree of standardization: The assessment results depend entirely on the professional ability and experience of the psychological counselor. There is no unified technical standard for interpreting the characteristics of the drawings. The conclusions drawn by different assessors have a high rate of difference and cannot guarantee the consistency of the assessment results. 2. Low processing efficiency and inability to adapt to large-scale scenarios: Manual analysis of a single house-tree-person drawing requires the completion of the entire process of recording image features, dynamic diagnosis, psychological feature association, and report writing, which takes a long time on average and is difficult to meet the large-scale needs of scenarios such as school student screening, corporate stress assessment, and batch psychological monitoring of judicial detainees. 3. Lack of dynamic behavioral data and incomplete assessment dimensions: Only static drawing results are collected, ignoring key dynamic behavioral data during the drawing process. Such data is a core indicator reflecting the psychological oscillation state of the test subject, resulting in incomplete assessment dimensions and difficulty in accurately capturing deep psychological states. Summary of the Invention

[0007] In view of the above-mentioned technical shortcomings, the purpose of this invention is to provide a method and system for assessing mental health status projection based on AI recognition.

[0008] To solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a mental health status projection assessment method based on AI recognition, including the following steps: Step 1, analysis model construction: collect house-tree-person drawing data of historical users without external intervention, analyze the house-tree-person drawing data of historical users without external intervention, and obtain a fusion data AI analysis model.

[0009] Step 2: User Psychological Analysis: Collect visual feature data and dynamic behavior data of the target user's paintings. Based on the fusion data AI analysis model, input the visual feature data of the target user's paintings to obtain the target user's system state data. Analyze the target user's dynamic behavior data to obtain the target user's psychological positive feedback coefficient, thereby verifying the target user's system state data and outputting the target user's psychological level.

[0010] Step 3: Group Psychological Analysis: Collect system state change data of target users, analyze the system state change data of target users to obtain abnormal psychological states of target users, thereby obtaining psychological state data of user group, and analyze the psychological state data of target user group to issue early warning prompts.

[0011] Preferably, the verification of the target user's system status data is specifically analyzed as follows: The target user's system status data consists of various system status scores. After normalization, the various system status scores of the target user are weighted and calculated to obtain the target user's system status psychological index. The system status psychological index interval corresponding to each system status psychological level is obtained from the database. If the target user's system status psychological index belongs to the system status psychological index interval corresponding to a certain system status psychological level, it indicates that the target user is at that system status psychological level.

[0012] The range of positive psychological feedback coefficients corresponding to each level of psychological fluctuation is obtained from the database. If the positive psychological feedback coefficient of the target user belongs to the range of positive psychological feedback coefficients corresponding to a certain level of psychological fluctuation, it indicates that the target user is at that level of psychological fluctuation.

[0013] If the target user's system state psychological level and psychological fluctuation level are not the same, obtain the various system state correction rates corresponding to each psychological fluctuation level from the database to obtain the various system state correction rates of the target user. Multiply the various system state scores of the target user by the correction rates of various system state data to obtain the various system state correction scores of the target user, and use this to verify.

[0014] On the other hand, the present invention provides a mental health status projection assessment system based on AI recognition, including the following modules: an analysis model construction module, used to collect house-tree-person drawing data of historical users without external intervention, analyze the house-tree-person drawing data of historical users without external intervention, and obtain a fusion data AI analysis model.

[0015] The user psychological analysis module is used to collect the visual feature data and dynamic behavior data of the target user's paintings. Based on the fusion data AI analysis model, the target user's visual feature data is input to obtain the target user's system status data. The target user's dynamic behavior data is analyzed to obtain the target user's psychological positive feedback coefficient, thereby verifying the target user's system status data and outputting the target user's psychological level.

[0016] The group psychological analysis module is used to collect system state change data of target users, analyze the system state change data of target users to obtain abnormal psychological states of target users, thereby obtaining psychological state data of user groups, analyzing the psychological state data of target user groups, and issuing early warning prompts.

[0017] The beneficial effects of this invention are as follows: 1. This method first collects historical House-Tree-Person (HTP) data of users without external intervention by constructing an analysis model, and then trains it through a distributed GPU cluster to build a fusion data AI analysis model. Secondly, it collects target users' HTP drawing and dynamic behavior data through user psychological analysis. The model outputs system state data, analyzes dynamic behavior to obtain psychological positive feedback coefficients, and outputs psychological levels after verification. Finally, through group psychological analysis, it analyzes changes in the target users' system state to identify abnormal psychological states, summarizes user group data, and issues warnings based on key attention states when abnormalities occur. This invention achieves objectivity and efficiency in assessment through AI, and improves accuracy through dual-dimensional verification, which is used to support the management of individual and group mental health.

[0018] 2. This invention combines static house-tree-person drawing data with dynamic behavioral data for dual-dimensional assessment. On one hand, it analyzes the visual features of the drawing using a data-integrated AI analysis model, outputting systemic state data such as personality traits, personality structure, and psychological state. On the other hand, it calculates the psychological positive feedback coefficient using dynamic behavioral data such as the interval between strokes, the duration of revisions, and the order in which elements are drawn, thus validating the systemic state data. Compared to traditional scales that rely solely on subjective responses and traditional projective tests that focus only on static results, this invention, through cross-validation of static features and dynamic processes, can effectively identify potential psychological conflicts in test subjects and reduce the false negative rate of high-risk cases.

[0019] 3. This invention uses house-tree-person drawing as the core assessment vehicle, combined with dynamic behavioral data, without the need for written answers. It completely eliminates the reliance of traditional scales on language comprehension and written expression ability, and can effectively cover young children, people with low education levels, people with language disorders and cross-cultural groups. In the mental health screening of special populations, it reduces the invalid data caused by the difficulty of understanding the text in traditional scales.

[0020] 4. On the one hand, by analyzing the improvement rate of the system state scores of target users in each period, this invention can identify self-excitation phenomena and locate abnormal types. Compared with traditional assessment of single cross-sectional detection, this invention can capture the dynamic evolution of psychological state, provide early warning of potential psychological crises, and provide a window period for early intervention. On the other hand, this invention summarizes user group data, analyzes and warns, and determines key attention states, which can improve the speed of group organization in locating high-risk groups and improve the accuracy of allocating intervention resources. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention.

[0023] Figure 2 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] according to Figure 1 As shown, the present invention provides a mental health status projection assessment method based on AI recognition, including the following steps: Step 1, analysis model construction: collect house-tree-person drawing data of historical users without external intervention, analyze the house-tree-person drawing data of historical users without external intervention, and obtain a fusion data AI analysis model.

[0026] In one specific embodiment, the collection of house-tree-person drawing data from historical users without external intervention is specifically carried out as follows: the house-tree-person drawing data from historical users without external intervention includes the visual feature data of each historical test image and the corresponding psychological feature data.

[0027] The system collects historical test images using cameras and uses machine vision to collect visual feature data of the paintings in each historical test image. The visual feature data of the paintings includes, but is not limited to, the length of the roof of the house, the area of ​​the head of the figure, and the number of clothing and decorations of the figure.

[0028] The psychological characteristic data was set by double-blind labeling by psychology experts. The specific scores were set by the psychology experts. The corresponding psychological characteristics included, but were not limited to, scores of various personality traits, scores of various levels of personality structure, and scores of various psychological states.

[0029] It should be noted that the personality traits include, but are not limited to, introversion / extroversion, emotional / rational, and passive / positive; the personality structures include, but are not limited to, the id, the ego, and the superego; and the psychological states include, but are not limited to, psychological security, self-identity, and anxiety tendency.

[0030] In one specific embodiment, the analysis of historical user house-tree-person drawing data without external intervention is carried out as follows: the visual feature data of each historical test drawing, the corresponding psychological features and risk level labels are used to train a model based on a distributed GPU cluster to obtain a fusion data AI analysis model.

[0031] By training the model using a distributed GPU cluster, a knowledge graph of the association rules between image features and psychological indicators is first established to obtain the quantitative mapping from each historical test image to various psychological indicators. This yields the quantitative mapping data of various psychological indicators corresponding to each historical test image. Then, a random forest classification model is trained to obtain the scores of various personality traits, personality structure, and psychological state indicators corresponding to each historical test image. This is used to construct a fusion data AI analysis model.

[0032] The fusion data AI analysis model: input the visual feature data of the target user's painting, and obtain various system state scores of the target user.

[0033] Step 2: User Psychological Analysis: Collect visual feature data and dynamic behavior data of the target user's paintings. Based on the fusion data AI analysis model, input the visual feature data of the target user's paintings to obtain the target user's system state data. Analyze the target user's dynamic behavior data to obtain the target user's psychological positive feedback coefficient, thereby verifying the target user's system state data and outputting the target user's psychological level.

[0034] In one specific embodiment, the process of collecting the target user's drawing visual feature data and dynamic behavior data is as follows: based on the historical house-tree-person drawing data of the user without external intervention, the process of collecting the target user's drawing visual feature data is as follows.

[0035] The target user's dynamic behavior data includes the target user's pen stroke interval data, modification duration data, and element drawing order data.

[0036] The target user's pen stroke interval data is the duration of each pen stroke by the target user; the target user's modification duration data is the duration of each modification by the target user; and the target user's element drawing order data is the element and timestamp of each pen stroke by the target user.

[0037] The special drawing paper has a built-in pressure sensor and positioning module. The built-in pressure sensor collects the time interval between each stroke and the time of each modification. The built-in pressure sensor and positioning module determine the stroke timestamp of each stroke position. Machine vision identifies the element type of each stroke position, thereby obtaining the element and timestamp of each stroke.

[0038] In one specific embodiment, the process of obtaining the target user's system status data is as follows: input the target user's drawing visual feature data into the fusion data AI analysis model, and output the target user's various system status scores.

[0039] In one specific embodiment, the analysis of the target user's dynamic behavior data is carried out as follows: the target user's pen stroke interval data is normalized to obtain the target user's normalized pen stroke interval value; the target user's modification time data is normalized to obtain the target user's normalized modification time value; the target user's element drawing order data is normalized to obtain the target user's element drawing normalized value; after normalization, the target user's oscillation state index is calculated by weighting; the target user's oscillation state index is subtracted from the oscillation state index threshold, and then divided by the oscillation state index threshold to obtain the target user's psychological positive feedback coefficient.

[0040] It should be noted that the weighted calculation process is as follows: retrieve the pen stroke interval weight factor, modification duration weight factor, and element drawing weight factor from the database, and multiply the normalized values ​​of the target user's pen stroke interval, modification duration, and element drawing by the corresponding weight factors to obtain the target user's oscillation state index.

[0041] In one specific embodiment, the normalization process for the pen stroke interval data of the target user is as follows: The pen stroke interval data of the target user is the duration of each pen stroke interval of the target user. The maximum pen stroke interval duration is subtracted from the minimum pen stroke interval duration, and then divided by the average pen stroke interval duration to obtain the pen stroke interval fluctuation range. The pen stroke interval fluctuation range of the target user is divided by the pen stroke interval fluctuation range threshold to obtain the normalized value of the pen stroke interval of the target user.

[0042] It should be noted that the threshold for the fluctuation range of the pen stroke interval represents the upper limit of the fluctuation range of the pen stroke interval for psychologically stable individuals, and the specific value is set by the staff.

[0043] In one specific embodiment, the normalization process for the modification duration data of the target user is as follows: The modification duration data of the target user is the modification duration of each modification by the target user. The maximum modification duration is subtracted from the minimum modification duration, and then divided by the average modification duration to obtain the modification duration fluctuation range. The modification duration fluctuation range of the target user is divided by the modification duration fluctuation range threshold to obtain the normalized value of the modification duration of the target user.

[0044] It should be noted that the threshold for the fluctuation range of modification time is the upper limit of the fluctuation range of modification time during the drawing process for a historical user group whose psychological state is determined to be basically stable under the condition of no external intervention.

[0045] In one specific embodiment, the normalization process for the element drawing order data of the target user is as follows: The element drawing order data of the target user consists of the element and timestamp of each stroke made by the target user, thereby obtaining the timestamps of various elements made by the target user. The minimum timestamp of a certain type of element stroke is recorded as the lower limit of the timestamp interval of the element stroke for that type, and the maximum timestamp is recorded as the upper limit of the timestamp interval of the element stroke for that type, thereby obtaining the timestamp interval of the element stroke for that type. Each stroke that is not a stroke of the element stroke within the timestamp interval of the element stroke for that type is recorded as an abnormal stroke, thereby obtaining the abnormal strokes of the element stroke for that type. The abnormal values ​​of various abnormal strokes of various types of element strokes are obtained from the database, thereby obtaining the abnormal value of the abnormal stroke of the element stroke for that type. The total abnormal value of the abnormal strokes of the element stroke for that type is obtained by summing them up, thereby obtaining the total abnormal value of the element drawing order, and thus obtaining the total abnormal value of the element drawing order of the target user.

[0046] Divide the total outlier value of the element drawing order of the target user by the threshold of the total outlier value of the element drawing order to obtain the normalized value of the element drawing of the target user.

[0047] It should be noted that the threshold for total outliers in the element drawing order is the upper limit of the total outliers in the element drawing order during the drawing process for a historical user group whose psychological state is determined to be basically stable under the condition of no external intervention.

[0048] In one specific embodiment, the abnormal values ​​of various abnormal strokes for each type of element are set as follows: The timestamps of each type of element stroke in each test are obtained from the database. The minimum value of the timestamps of each type of element stroke in each test is recorded as the starting timestamp of that type of element, thus obtaining the starting timestamps of each type of element stroke in each test. The elements are then sorted in ascending order of the starting timestamps to obtain the order of each type of element stroke in each test. The order of each type of element stroke is then summarized to obtain the number of tests for each order. The order type with the largest number of tests is recorded as the standard order type, thus obtaining the standard order of each type of element.

[0049] According to the standard order of various elements, staff assign values ​​to each element, with the values ​​set from small to large according to the standard order of the elements. The numerical intervals between adjacent elements in the standard order are the same, thus obtaining the order characteristic value of each element. The order characteristic value of a certain type of abnormal stroke is subtracted from the characteristic value of a certain type of stroke, and then the absolute value is obtained to obtain the abnormal value of that type of abnormal stroke for that type of element. In this way, the abnormal values ​​of various types of abnormal strokes for each type of element are obtained.

[0050] It should be noted that values ​​are assigned to various elements, such as trees, people, and houses, in the order of tree-person-house. The order feature value of the tree is 1, the order feature value of the person is 3, and the order feature value of the house is 5.

[0051] In one specific embodiment, the verification of the target user's system status data is specifically analyzed as follows: The target user's system status data consists of various system status scores. After normalization, the various system status scores of the target user are weighted and calculated to obtain the target user's system status psychological index. The system status psychological index interval corresponding to each system status psychological level is obtained from the database. If the target user's system status psychological index belongs to the system status psychological index interval corresponding to a certain system status psychological level, it indicates that the target user is at that system status psychological level.

[0052] It should be noted that the weighted calculation process for the normalized system status scores of the target user is as follows: Subtract the corresponding standard score from the various system status scores of the target user, convert the absolute value, and divide by the corresponding standard score to obtain the normalized value of each system status score. Obtain the weight factor of each system status score from the database, multiply the normalized value of each system status score by the corresponding weight factor, and then add them together to obtain the target user's system status psychological index. The standard scores and weight factors for each system status are set by the staff.

[0053] The range of positive psychological feedback coefficients corresponding to each level of psychological fluctuation is obtained from the database. If the positive psychological feedback coefficient of the target user belongs to the range of positive psychological feedback coefficients corresponding to a certain level of psychological fluctuation, it indicates that the target user is at that level of psychological fluctuation.

[0054] If the target user's system state psychological level and psychological fluctuation level are not the same, obtain the various system state correction rates corresponding to each psychological fluctuation level from the database to obtain the various system state correction rates of the target user. Multiply the various system state scores of the target user by the correction rates of various system state data to obtain the various system state correction scores of the target user, and use this to verify.

[0055] In one specific embodiment, the specific analysis process for outputting the target user's psychological level is as follows: if the target user's system state psychological level is the same as the psychological fluctuation level, the target user's system state psychological level is output as the psychological level.

[0056] If the target user's system state psychological level and psychological fluctuation level are different, the target user's various system state correction scores are normalized and weighted to obtain the target user's system state correction psychological index. If the target user's system state correction psychological index belongs to the system state psychological index interval corresponding to a certain system state psychological level, it indicates that the target user's corrected system state psychological level is the system state psychological level, and the target user's corrected system state psychological level is output as the psychological level.

[0057] It should be noted that the weighted calculation process for the various system state correction scores of the target user after normalization is as follows: the various system state correction scores of the target user are subtracted from the corresponding standard scores, the absolute values ​​are then divided by the corresponding standard scores to obtain the normalized values ​​of the various system state correction scores of the target user. The normalized values ​​of the various system state correction scores are multiplied by the corresponding weight factors, and then added together to obtain the target user's system state psychological index. The weight factors are set by the staff.

[0058] Step 3: Group Psychological Analysis: Collect system state change data of target users, analyze the system state change data of target users to obtain abnormal psychological states of target users, thereby obtaining psychological state data of user group, and analyze the psychological state data of target user group to issue early warning prompts.

[0059] In one specific embodiment, the system state change data of the target user is collected in the following process: The system state change data of the target user includes the improvement rate of various system state scores in each period of the target user's tests. The various system state scores of the target user in each period of the tests are obtained through historical records. The corresponding standard value is subtracted from each system state score, and after absolute value conversion, it is divided by the corresponding standard value to obtain the performance score of each system state score. The performance score of each system state in the last historical test is subtracted from the current performance score of each system state score, and then divided by the performance score of each system state in the last historical test to obtain the improvement rate of the current system state score. Thus, the improvement rate of various system state scores of the target user in each period of the tests is obtained.

[0060] In one specific embodiment, the analysis of the target user's system state change data is carried out as follows: system state types with an improvement rate greater than the corresponding preset improvement rate are recorded as abnormal system state types, thereby obtaining various abnormal system state data of the target user in each period of testing. If the target user's system state data of a certain type are abnormal system state data for two consecutive periods, or if the number of abnormal system state types of the target user in a certain period of testing is greater than the preset number of types, it indicates that the target user has a self-excitation phenomenon. The type corresponding to the psychological state index of the maximum score improvement rate is recorded as the type of abnormal psychological state, thereby obtaining the abnormal psychological state of the target user.

[0061] It should be noted that the number of preset types is obtained by staff through experience or internet research, and the specific values ​​are set by staff.

[0062] In one specific embodiment, the process of obtaining the psychological state data of the user group is as follows: The psychological state data of the target user group includes the number of users in various abnormal system states of self-excitation and the improvement rate of the psychological positive feedback coefficient of each user. Based on the method of analyzing the system state change data of the target users, the number of users in various abnormal system states of self-excitation is obtained by statistically analyzing the self-excitation state of each user in the target user group.

[0063] Based on the process of analyzing the dynamic behavioral data of the target users, the psychological positive feedback coefficient of each user in each period of testing is obtained. The current psychological positive feedback coefficient is subtracted from the historical last psychological positive feedback coefficient, and then divided by the historical last psychological positive feedback coefficient to obtain the current psychological positive feedback coefficient improvement rate. This gives the improvement rate of the psychological positive feedback coefficient of each user.

[0064] In one specific embodiment, the analysis of the psychological state data of the target user group is carried out as follows: the average improvement rate of the psychological positive feedback coefficient of each user is calculated to obtain the improvement rate of the psychological positive feedback coefficient of the target user group; if the improvement rate of the psychological positive feedback coefficient of the target user group is less than zero, an early warning is issued.

[0065] The early warning process is as follows: users are sorted in descending order of number to obtain the abnormal system status order. The first preset number of various abnormal system status types in the abnormal system status order are recorded as various key attention statuses, and the supervisors are notified.

[0066] It should be noted that the preset quantity is obtained by staff through experience or internet research, and the specific value is set by staff.

[0067] according to Figure 2 As shown, the present invention provides a mental health status projection assessment system based on AI recognition, including the following modules: an analysis model construction module, a user psychological analysis module, a group psychological analysis module, and a database.

[0068] The user psychological analysis module is connected to the analysis model construction module and the group psychological analysis module, respectively. The analysis model construction module, the user psychological analysis module, and the group psychological analysis module are all connected to the database.

[0069] The analysis model building module is used to collect house-tree-person drawing data from historical users without external intervention, analyze the historical house-tree-person drawing data without external intervention, and obtain a fused data AI analysis model.

[0070] The user psychological analysis module is used to collect the target user's house-tree-person drawing data and dynamic behavior data. Based on the fusion data AI analysis model, the target user's house-tree-person drawing data is input to obtain the target user's system status data. The target user's dynamic behavior data is analyzed to obtain the target user's psychological positive feedback coefficient, thereby verifying the target user's system status data and outputting the target user's psychological level.

[0071] The group psychological analysis module is used to collect system state change data of target users, analyze the system state change data of target users to obtain abnormal psychological states of target users, thereby obtaining psychological state data of user groups, analyzing the psychological state data of target user groups, and issuing early warning prompts.

[0072] The database stores the weighting factors for pen stroke interval, modification duration, and element drawing, as well as the abnormal values ​​of various abnormal pen strokes for each element, the timestamps of various element pen strokes in each test, the system state psychological index range corresponding to each system state psychological level, the weighting factors for various system state scores, the psychological positive feedback coefficient range corresponding to each psychological fluctuation level, and the system state correction rate corresponding to each psychological fluctuation level.

[0073] The psychological expert double-blind annotation, distributed GPU cluster model training, and random forest classification model training described in this invention are all existing technologies that can be found on the Internet, so they will not be described in detail here.

[0074] The examples described in this invention are not limited to the specific embodiments listed above. The examples are merely illustrative to facilitate understanding of the invention and do not constitute a limitation on the scope of protection of this invention. Any modifications, equivalent substitutions, etc., made within the spirit and principles of this invention should be included within the scope of protection.

[0075] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. An AI recognition-based mental health state projection evaluation method, characterized in that, Comprising the following steps: Step one, analysis model construction: collect historical user's house tree man drawing data without external intervention, analyze historical user's house tree man drawing data without external intervention, obtain fusion data AI analysis model; Step two, user psychological analysis: collect target user's drawing visual feature data and dynamic behavior data, input target user's drawing visual feature data based on fusion data AI analysis model, obtain target user's system state data, analyze target user's dynamic behavior data, obtain target user's psychological positive feedback coefficient, thereby verify target user's system state data, output target user's psychological level; Step three, group psychological analysis: collect target user's system state change data, analyze target user's system state change data, obtain target user's abnormal psychological state, thereby obtain user group's psychological state data, analyze target user group's psychological state data, and give early warning prompt.

2. The psychological health state projection evaluation method based on AI recognition according to claim 1, characterized in that, The analysis of historical user's house tree man drawing data without external intervention is as follows: The historical user's house tree man drawing data without external intervention includes drawing visual feature data and corresponding psychological feature data of each historical test picture; The drawing visual feature data, corresponding psychological feature and risk level label of each historical test picture are trained based on distributed GPU cluster to obtain fusion data AI analysis model; Through distributed GPU cluster model training, first, establish a knowledge graph of picture feature-psychological index association rule, obtain the quantitative mapping of each historical test picture to each type of psychological index, thereby obtain the quantitative mapping data of each type of psychological index corresponding to each historical test picture, and then perform random forest classification model training to obtain the state score, structure score and psychological state index score of each type of personality trait corresponding to each historical test picture, thereby constructing fusion data AI analysis model; The fusion data AI analysis model: input target user's drawing visual feature data to obtain target user's each type of system state score.

3. The AI recognition-based mental health state projection evaluation method according to claim 1, characterized in that, The analysis of target user's dynamic behavior data is as follows: The dynamic behavior data of target user includes target user's pen drop interval data, modification time data and element drawing order data; The pen drop interval data of target user is normalized to obtain the pen drop interval normalized value of target user, the modification time data of target user is normalized to obtain the modification time normalized value of target user, the element drawing order data of target user is normalized to obtain the element drawing normalized value of target user, and the normalized weighted calculation obtains the oscillation state index of target user, the oscillation state index threshold is subtracted from the oscillation state index of target user, and then divided by the oscillation state index threshold to obtain the psychological positive feedback coefficient of target user.

4. The psychological health state projection evaluation method based on AI recognition according to claim 3, characterized in that, The normalization process of target user's element drawing order data is as follows: The element drawing order data of target user is the element and timestamp of each pen drop of target user, thereby obtaining the timestamp of each element pen drop of target user; The minimum timestamp of the element pen-down is recorded as the lower limit of the timestamp interval of the element pen-down, and the maximum timestamp is recorded as the upper limit of the timestamp interval of the element pen-down, so as to obtain the timestamp interval of the element pen-down, and the non-element pen-down in the timestamp interval of the element pen-down is recorded as each abnormal pen-down of the element pen-down, so as to obtain each abnormal pen-down of the element pen-down, the abnormal value of each abnormal pen-down of each element pen-down is obtained from the database, and then the abnormal value of each abnormal pen-down of the element pen-down is obtained, and the total abnormal value of the element abnormal pen-down is obtained, so as to obtain the total abnormal value of each element pen-down, and the total abnormal value of the element drawing order is obtained, so as to obtain the total abnormal value of the element drawing order of the target user; The total abnormal value of the element drawing order of the target user is divided by the element drawing order total abnormal value threshold to obtain the element drawing normalization value of the target user.

5. The AI recognition-based mental health state projection evaluation method according to claim 4, characterized in that, The abnormal value of each abnormal pen-down of each element pen-down is specifically set as follows: The timestamps of each element pen-down of each test are obtained from the database, the minimum value of the timestamps of each element pen-down of each test is recorded as the starting timestamp of the element, so as to obtain the starting timestamp of each element pen-down of each test, and the starting timestamps are sorted in ascending order to obtain the order of each element pen-down of each test, and the test quantity of each order is obtained by summarizing according to the order of each element pen-down, and the order type with the maximum test quantity is recorded as the standard order type, so as to obtain the standard order of each element; According to the standard order of each element, the staff assigns values to each element, and the values are set in ascending order according to the standard order of each element, and the value interval of adjacent elements in the standard order is the same, so as to obtain the order characteristic value of each element, the order characteristic value of the abnormal pen-down of the element is subtracted from the characteristic value of the element pen-down, and then the absolute value is obtained to obtain the abnormal value of the abnormal pen-down of the element, so as to obtain the abnormal value of each abnormal pen-down of each element pen-down.

6. The psychological health state projection evaluation method based on AI recognition according to claim 3, characterized in that, The specific analysis process of verifying the system state data of the target user is as follows: The system state data of the target user is the system state score of each system state of the target user, and the system state psychological index of the target user is obtained by normalizing and weighting the system state scores of each system state of the target user, the system state psychological index interval corresponding to each system state psychological level is obtained from the database, and if the system state psychological index of the target user belongs to the system state psychological index interval corresponding to a certain system state psychological level, it indicates that the target user is of the system state psychological level; The psychological positive feedback coefficient interval corresponding to each psychological fluctuation level is obtained from the database, and if the psychological positive feedback coefficient of the target user belongs to the psychological positive feedback coefficient interval corresponding to a certain psychological fluctuation level, it indicates that the target user is of the psychological fluctuation level. If the system state psychological level of the target user is not the same as the psychological fluctuation level, the system state correction rates corresponding to each psychological fluctuation level are obtained from the database, the system state correction rates of the target user are obtained, the system state scores of the target user are multiplied by the correction rates of the system state data, and the system state correction scores of the target user are obtained. Verification is carried out in this way.

7. The AI recognition-based mental health state projection evaluation method according to claim 6, characterized in that, The output target user's psychological level is analyzed in detail as follows: If the system state psychological level of the target user is the same as the psychological fluctuation level, the system state psychological level of the target user is output as the psychological level; If the system state psychological level of the target user is not the same as the psychological fluctuation level, the system state correction scores of the target user are normalized and weighted to obtain the system state correction psychological index of the target user. If the system state correction psychological index of the target user belongs to the system state psychological index interval corresponding to a certain system state psychological level, it indicates that the target user is corrected to this system state psychological level. The system state psychological level of the target user after correction is output as the psychological level.

8. The AI recognition-based mental health state projection evaluation method according to claim 1, characterized in that, The system state change data of the target user is analyzed as follows: The system state change data of the target user includes the improvement rate of each type of system state score of the target user in each test. The system state type with an improvement rate greater than the corresponding preset improvement rate is recorded as an abnormal system state type. In this way, the abnormal system state data of each type of the target user in each test is obtained. If the system state data of a certain type of the target user is abnormal for two consecutive periods, or the number of abnormal system state types of the target user in a certain test is greater than the preset type number, it indicates that the target user has a self-activation phenomenon. The type corresponding to the maximum score improvement rate psychological state index is recorded as the type of psychological state abnormality. In this way, the abnormal psychological state of the target user is obtained.

9. The AI recognition-based mental health state projection evaluation method according to claim 8, characterized in that, The psychological state data of the target user group is analyzed as follows: The psychological state data of the target user group includes the number of abnormal system state users in each type of self-activation state and the improvement rate of the psychological positive feedback coefficient of each user. The improvement rate of the psychological positive feedback coefficient of each user is calculated by mean value to obtain the improvement rate of the psychological positive feedback coefficient of the target user group. If the improvement rate of the psychological positive feedback coefficient of the target user group is less than zero, a warning is given. The warning prompt process is as follows: based on the order from large to small, the abnormal system state order is obtained, and the first preset number of each type of abnormal system state type in the abnormal system state order is recorded as each type of focus state. The supervisor is prompted.

10. A projection assessment system using the AI recognition-based mental health state projection assessment method according to any one of claims 1-9, characterized in that, The following modules are included: The analysis model construction module is used to collect historical user house tree painting data without external intervention, analyze the historical user house tree painting data without external intervention, and obtain a fusion data AI analysis model. The user psychological analysis module is configured to collect drawing visual feature data and dynamic behavior data of a target user, input the drawing visual feature data of the target user based on a fusion data AI analysis model, obtain system state data of the target user, analyze dynamic behavior data of the target user, obtain a psychological positive feedback coefficient of the target user, thereby verifying the system state data of the target user, and output a psychological level of the target user. The group psychological analysis module is configured to collect system state change data of a target user, analyze the system state change data of the target user, obtain an abnormal psychological state of the target user, thereby obtaining psychological state data of a user group, analyze the psychological state data of the target user group, and provide a warning prompt.

Images