CN-121981859-A - Intelligent sentry personnel education evaluation and transformation integrated system

CN121981859ACN 121981859 ACN121981859 ACN 121981859ACN-121981859-A

Abstract

The invention discloses an intelligent prisoner education evaluation and transformation integrated system, and relates to the technical field of intelligent correction of prisoners. The system comprises a data acquisition module for acquiring multi-modal data in real time, a map construction module for extracting psychological factor characteristics, emotion coding characteristics, behavior strategy characteristics and physiological abnormality index characteristics after aligning the multi-modal data according to a unified time axis, and constructing a state map with psychological factors as nodes and the association relationship among factors as edges, a risk assessment module for inputting the state map into a graph neural network to output psychological risk levels, a reconversion trend path and key nodes, a task generation module for generating VR training instructions or dialogue scene instructions according to the key nodes, and a feedback update module for receiving the instructions and then acquiring new behavior and physiological data in the training process, and dynamically updating the state map. And by a multi-mode data fusion and dynamic map closed-loop updating mechanism, the risk accurate assessment of the prisoner and the self-adaptive optimization of the personalized correction scheme are realized.

Inventors

ZHONG WEIFANG
MO LEI
GUO YONGXING
GAO RUIXIANG
CHEN RUNLONG

Assignees

广东司法警官职业学院

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. An integrated system for education assessment and transformation of intelligent prisoners is characterized by comprising the following components: the data acquisition module is used for acquiring multi-modal data of the target object in real time, wherein the multi-modal data comprises a psychological scale vector, a behavior sequence label, a dialogue text, an emotion label and a physiological time sequence value; The map construction module is used for aligning the multi-mode data according to a unified time axis, extracting psychological factor characteristics, emotion coding characteristics, behavior strategy characteristics and physiological abnormality index characteristics, and constructing a state map taking psychological factors as nodes and the association relationship among the factors as edges; the risk assessment module is used for inputting the state map into a graph neural network to obtain a current psychological risk level, a current reimbursement tendency path and a corresponding key node; The task generation module is used for generating a corresponding VR training instruction or dialogue scene instruction according to the key node; And the feedback updating module is used for receiving the VR training instruction or the dialogue scene instruction, collecting a new behavior sequence label and a new generation time sequence value of the target object in the training process, and dynamically updating the state map.
2. The integrated system for intelligent prisoner education assessment and transformation according to claim 1, wherein the extraction of psychological factor characteristics from the atlas construction module specifically comprises: And screening the psychological scale vector of the target object through a predefined activation threshold, and only reserving psychological factors exceeding the activation threshold as nodes of the state map, wherein the nodes distribute weights according to the proportion of the scoring values of the psychological factors.
3. The intelligent prisoner educational evaluation and transformation integrated system of claim 1, wherein the extraction of emotion encoding features in the atlas construction module specifically comprises calculating the duty ratio of the active emotion words of the target object in dialogue text, and generating special nodes associated with transformation priority lifting marks when the duty ratio is lower than a preset duty ratio threshold.
4. The integrated system for intelligent prisoner education assessment and transformation according to claim 1, wherein the extraction of the behavior policy features in the map construction module specifically comprises: And inputting the behavior sequence label into a preset behavior pattern template library for matching, and outputting a behavior code representing a decision pattern, wherein the behavior code is used for triggering the generation condition judgment of the associated edge in the state map.
5. The integrated system for intelligent prisoner education assessment and transformation according to claim 1, wherein the construction of a state map with psychological factors as nodes and the association between factors as edges, the generation of which is that the edges follow a predefined psychological rule, comprises: Generating an edge representing that the aggression is associated with the aggression when the aggressive behavior code of the target object is detected to coexist with the high anger emotion; when a high cognitive deviation node value of the target object is detected to accompany physiological abnormal fluctuation, generating an edge which is used for representing that the cognitive deviation is associated with physiological disorder.
6. The integrated system for intelligent prisoner educational assessment and transformation according to claim 1, wherein the specific steps of inputting the state atlas into a graphic neural network, and outputting the current psychological risk level and the path of the offensiveness comprise: The graph neural network calculates a current node risk value based on the risk value of the adjacent node and the edge weight through an edge weight propagation mechanism; wherein the contribution of the edge weight to the risk value is controlled by adopting a preset attenuation coefficient.
7. The intelligent prisoner educational evaluation and transformation integrated system according to claim 1, wherein said step of dynamically updating said state atlas by said feedback updating module comprises: when the state map executes node value updating, synchronously acquiring the new generation time sequence value, and when the descending rate of the new generation data index of the new generation time sequence value does not reach a preset target value, carrying out weight strengthening operation on the key nodes; The weight strengthening operation is according to a formula The execution is performed, wherein, For the original value of the node weight, And alpha is a strengthening coefficient, which is the value after updating the node weight.
8. The integrated intelligent prisoner educational assessment and transformation system of claim 1, wherein the specific step of generating corresponding VR training instructions or dialogue scene instructions from the key nodes comprises: Based on the topological structure of the re-strike trend path, matching a path node set with similarity exceeding a preset similarity threshold value in a history correction case library, and preferentially calling a verification valid instruction type corresponding to the path node set; meanwhile, the conflict instruction types in the failed cases are marked to generate an avoidance list.
9. The integrated intelligent prisoner educational assessment and transformation system of claim 7, further comprising a path optimization module for invoking a historical correction case library and switching a transformation task type and resetting an operation count when the node weight reinforcement operation reaches a preset number of times.
10. The integrated intelligent prisoner educational assessment and transformation system of claim 1, further comprising a data security module that performs an automatic triggering of a proprietary protocol channel that individually encrypts the sensitive index using a preset encryption algorithm to generate encrypted data when the physiological time series value contains a predefined sensitive index; the encrypted data transmits only a decryption key fragment to the map construction module.

Description

Intelligent sentry personnel education evaluation and transformation integrated system Technical Field The invention relates to the technical field of intelligent correction of prisoners, in particular to an integrated system for education evaluation and transformation of intelligent prisoners. Background In the field of intelligent correction systems, the prior art has been able to achieve static risk assessment of the prisoner through multimodal data collection (such as mental scale, behavioral records and physiological indicators) and trigger standardized modification schemes based on fixed thresholds. The system has a certain effect in basic risk screening, but exposes a fundamental defect that the multi-source heterogeneous data is subjected to fracture analysis in the time dimension and the causal dimension, so that the misjudgment rate of the dynamic psychological state and the behavioral risk of the prisoner is high. Specifically, the hysteresis of the psychological scale, the discreteness of the behavior sequence and the instantaneous fluctuation of the physiological index form a dynamic network which is mutually influenced in the actual scene (for example, the increase of the emotion fluctuation is often accompanied by the abnormal skin-to-electricity reaction and the increase of the attack behavior), but the traditional system independently processes various data, namely the psychological assessment depends on a periodic questionnaire, the behavior analysis is based on discrete event records, and the physiological monitoring adopts independent threshold value warning. Particularly for high risk prisoners, their psychological, behavioral and physiological states often exhibit non-linear coupling characteristics (e.g. positive feedback loops of anger emotion and violent behaviour). The split analysis model of the existing system is difficult to quantify the complex association, so that risk assessment deviates from a real state, and further, the correction scheme is misplaced with the actual demand, potential risks are accumulated, or excessive intervention is implemented to cause contradiction psychology. Disclosure of Invention Aiming at the defects existing in the prior art, the invention provides an intelligent sentry personnel education evaluation and transformation integrated system. In order to achieve the above object, the technical scheme of the present invention is as follows: The invention discloses an intelligent prisoner education evaluation and transformation integrated system, which comprises: The system comprises a data acquisition module, a target object acquisition module and a data processing module, wherein the data acquisition module is used for acquiring multi-modal data of the target object in real time, the multi-modal data comprises a psychological scale vector, a behavior sequence label, a dialogue text, an emotion label and a physiological time sequence value; The map construction module is used for aligning the multi-mode data according to a unified time axis, extracting psychological factor characteristics, emotion coding characteristics, behavior strategy characteristics and physiological abnormality index characteristics, and constructing a state map taking psychological factors as nodes and the association relationship among the factors as edges; the risk assessment module is used for inputting the state map into a graph neural network to obtain a current psychological risk level, a current reimbursement tendency path and a corresponding key node; The task generation module is used for generating a corresponding VR training instruction or dialogue scene instruction according to the key node; And the feedback updating module is used for receiving the VR training instruction or the dialogue scene instruction, collecting a new behavior sequence label and a new generation time sequence value of the target object in the training process, and dynamically updating the state map. Compared with the prior art, the invention has the beneficial effects that: 1. According to the invention, through constructing a multi-mode data alignment mechanism with a unified time axis, heterogeneous data such as psychological states, behavioral decisions, language emotion, physiological reactions and the like are precisely coupled in a time domain, so that the problem of cognitive deviation caused by data splitting in traditional evaluation is solved. Based on dynamic graph modeling, the collaborative expression of psychological factor nodes and cross-modal associated edges is realized, so that originally isolated risk indexes (such as impulsive psychology characteristics) can form causal chain mapping with corresponding behavioral patterns (aggressive action sequences) and physiological abnormal signals (skin electric peak fluctuation). The risk evolution path hidden in multidimensional data interaction (such as critical state of anger emotion accumulation triggering violent behavior) can be