Intelligent police call receiving and automatic report filling device based on semantic analysis

Abstract

The application relates to the technical field of police receiving equipment, and discloses an intelligent police receiving and automatic reporting equipment based on semantic analysis, which comprises an equipment main body, a communication module and a processing module installed in the equipment main body, wherein the communication module and the processing module are connected, the communication module is used for receiving alarm information, the alarm information comprises a voice alarm signal, and the processing module comprises a context cognition module, a dynamic reporting module and a verification calculation module which are connected in sequence. The context cognition module is used for performing deep semantic understanding on the alarm voice, actively identifying information gaps and dynamically adjusting a follow-up strategy in combination with the state of an alarm person, so that the problem of missing key information caused by the fuzzy and fragmented description of the alarm person in the traditional passive recording mode is solved, and the completeness of police information collection is improved.

Description

Technical Field

[0001] This invention relates to the field of alarm receiving equipment technology, and in particular to an intelligent alarm receiving and automatic reporting device based on semantic analysis. Background Technology

[0002] In public safety and emergency response, receiving and handling alarms is the primary step in responding to emergencies, and its efficiency and accuracy directly affect the protection of people's lives and property. Traditional alarm receiving and reporting systems are mainly based on a model of manual answering of calls, manual recording of information, and filling out standardized work forms, followed by manual dispatch. With the development of information technology, some auxiliary systems with voice recognition and simple information extraction functions have emerged.

[0003] However, traditional or existing basic intelligent systems typically only transcribe the caller's voice and extract keywords, which is a passive recording mode. When the caller is nervous, has limited language skills, or lacks complete information about the situation, their description is often vague, fragmented, and omits key elements (such as precise location, person's condition, and suspect characteristics). The system cannot proactively identify these information gaps and conduct strategic follow-up questions, resulting in incomplete and missing information in the generated police report. Furthermore, the police dispatch center, video surveillance platform, and historical police report database often operate independently, requiring dispatchers to manually switch between different systems, query, and compare information, which is inefficient. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent alarm receiving and automatic reporting device based on semantic analysis to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent alarm receiving and automatic reporting device based on semantic analysis, comprising a device body and a communication module and a processing module installed inside the device body, wherein the communication module and the processing module are connected, the communication module is used to receive alarm information, the alarm information including a voice alarm signal, and the processing module includes a context cognition module, a dynamic reporting module and a verification calculation module connected in sequence.

[0006] The main body of the device is also equipped with a multimodal fusion module connected to the processing module, which is used to access and integrate video streams, IoT sensor data and historical alarm data, and align, verify and complement the semantic information of alarm voice, and output multimodal fusion information to the context cognition module.

[0007] The contextual cognition module is used to perform real-time semantic analysis on alarm information, construct and dynamically update the contextual cognition network. The contextual cognition network includes a cognitive layer for understanding current information, an intent layer for predicting subsequent development, a contextual layer for associating global information, and a decision layer for generating action suggestions. The dynamic filling module is used to generate a dynamic filling interface and form content that matches the current alarm type and complexity based on the structured contextual information output by the contextual cognition network.

[0008] Preferably, the contextual cognition module includes an intelligent questioning unit, which is used to detect information gaps in the contextual cognition network in real time and initiate proactive questioning based on a preset questioning strategy library. The questioning strategy library includes clarification questioning strategies for vague descriptions, guidance questioning strategies for emotionally agitated callers, and confirmation questioning strategies for key information. The intelligent questioning unit is configured to dynamically select and combine different questioning strategies based on the type and urgency of the information gap and the caller's real-time emotional analysis results, and adopts a progressively converging questioning path.

[0009] Preferably, the dynamic reporting module includes an interface adaptation unit, which is used to dynamically adjust the information complexity and presentation density of the dynamic reporting interface according to the real-time assessment of the cognitive load status of the dispatcher. The assessment of the cognitive load status is based on at least one parameter among the real-time number of dispatches, the complexity level of the current alarm, and the dispatcher's operation frequency and number of decisions per unit time.

[0010] Preferably, the dynamic reporting module further includes a multi-view generation unit, used to generate differentiated form views for different roles in parallel based on the same structured contextual information. The differentiated form views include at least a detailed view for dispatchers that includes information completeness and analysis summary, an action view for on-site personnel that focuses on action lists and on-site risk warnings, and a decision view for commanders that integrates situational simulation and resource distribution. The action view can interact with the mobile terminals of on-site police officers, update the dispatch status and on-site feedback information in real time, and synchronize with the detailed view and decision view.

[0011] Preferably, the verification calculation module includes a distributed verification network for cross-verifying and consensus-confirming key elements in structured contextual information among multiple alarm receiving positions, on-site mobile terminals, and fixed monitoring nodes. The verification calculation module also includes a self-evolving knowledge graph unit for continuously optimizing alarm handling rules, entity relationships, and classification patterns based on the handling feedback results, and recording the source and update trajectory for each knowledge item in the graph.

[0012] Preferably, the self-evolving knowledge graph unit is configured to generate a traceable explanation chain for each rule, relationship, or pattern that the system automatically learns or optimizes during the knowledge update process. The explanation chain includes the decision basis, the data source used, and the event that triggers the learning. The verification calculation module also includes a knowledge review interface for receiving confirmation, correction, or rejection instructions from human operators for the knowledge items that the system automatically evolves, and updating the confidence weight of the knowledge graph according to the instructions.

[0013] Preferably, the context cognition module further includes a prediction unit, which is used to deduce multiple development paths of the alarm in the future period based on the currently constructed context cognition network, and evaluate the occurrence probability and resource impact of each path, forming a context evolution tree and outputting it to the dynamic reporting module. The prediction unit is connected to the multimodal fusion module and can dynamically correct and re-evaluate the context evolution tree based on new sensor data or on-site feedback acquired in real time.

[0014] Preferably, the multimodal fusion module includes a spatiotemporal context graph construction unit, used to construct the fused multi-source information into a graph containing entity nodes and their relational edges of "people, vehicles, objects, places, and times". The relations include spatial location relations, time sequence relations, and behavioral interaction relations. The multimodal fusion module also includes a conflict detection and resolution unit, used to resolve conflicts based on the historical reliability, real-time confidence level, and inherent attributes of each information source when there are conflicts in the information about the same entity provided by different information sources, and output the fusion result marked with confidence level and conflict state.

[0015] Preferably, the processing module further includes a speech recognition module and a confidence modeling unit. The confidence modeling unit is used to attach a confidence score and a source identifier to each piece of information and each inference result in the contextual cognition network, and to mark information with a confidence score lower than a preset threshold as high uncertainty information. When generating the form, the dynamic filling module visually highlights the content from the high uncertainty information and, together with the intelligent follow-up questioning unit, generates a secondary verification suggestion for the high uncertainty information.

[0016] The technical effects and advantages of this invention are as follows:

[0017] This invention utilizes a contextual cognition module to achieve deep semantic understanding of alarm voice messages, proactively identifying information gaps and dynamically adjusting follow-up questioning strategies based on the caller's state. This solves the problem of missing key information caused by vague and fragmented caller descriptions in traditional passive recording modes, improving the completeness of alarm information collection. Simultaneously, a multimodal fusion module integrates heterogeneous data from multiple sources, including alarm voice messages, video surveillance, and historical alarm data, achieving information alignment, verification, and complementarity. This eliminates the need for dispatchers to manually switch between multiple independent systems, improving information processing efficiency during alarm reception. Furthermore, a dynamic data entry module adjusts the interface information density based on the dispatcher's cognitive load, generating differentiated form views for dispatchers, on-site personnel, and command personnel. This meets the information needs of different roles, enabling cross-role information collaboration, avoiding information transmission distortion, and improving overall collaborative efficiency. Attached Figure Description

[0018] Figure 1 This is a detailed structural diagram of the communication module and processing module in an intelligent alarm receiving and automatic reporting device based on semantic analysis according to the present invention.

[0019] In the diagram: 1. Main body of the device; 2. Communication module; 3. Processing module. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] This invention provides, for example Figure 1 The illustrated intelligent alarm receiving and automatic reporting device based on semantic analysis includes a main body 1 and a communication module 2 and a processing module 3 installed inside the main body 1. The communication module 2 and the processing module 3 are connected. The communication module 2 is used to receive alarm information, including voice alarm signals. The processing module 3 includes a context cognition module, a dynamic reporting module, and a verification calculation module connected in sequence. The main body 1 is also equipped with a multimodal fusion module connected to the processing module 3, which is used to access and integrate video streams, IoT sensor data, and historical alarm data, and to align, verify, and complement the semantic information of the alarm voice, and output multimodal fusion information to the context cognition module.

[0022] The contextual cognition module performs real-time semantic analysis on alarm information, constructs and dynamically updates a contextual cognition network. This network includes a cognitive layer for understanding current information, an intent layer for predicting future developments, a contextual layer for relating global information, and a decision-making layer for generating action suggestions. The dynamic reporting module generates a dynamic reporting interface and form content that matches the type and complexity of the current alarm based on the structured contextual information output by the contextual cognition network. Through proactive semantic understanding of alarm information, rather than simple keyword extraction and transcription, it can proactively identify information gaps and initiate targeted follow-up questions, solving the problem of incomplete alarm information caused by vague and fragmented descriptions from alarm callers. At the same time, it integrates heterogeneous data from multiple sources, eliminating the need for dispatchers to manually switch and integrate data between multiple independent systems, thus improving alarm receiving efficiency and information accuracy.

[0023] The contextual cognition module includes an intelligent questioning unit, which detects information gaps in the contextual cognition network in real time and initiates proactive questioning based on a pre-set questioning strategy library. This library includes clarification strategies for vague descriptions, guidance strategies for emotionally agitated callers, and confirmation strategies for key information. The intelligent questioning unit is configured to dynamically select and combine different questioning strategies based on the type and urgency of the information gap and the caller's real-time emotional analysis results. It employs a progressively converging questioning path, flexibly adjusting the questioning method according to different scenarios and the caller's state. This avoids exacerbating the caller's emotional distress with overly dense questioning while gradually narrowing the information scope, quickly filling in key alarm elements, and improving the completeness and accuracy of information collection.

[0024] The dynamic data entry module includes an interface adaptation unit, which dynamically adjusts the information complexity and presentation density of the dynamic data entry interface based on the real-time assessment of the call receiver's cognitive load. The cognitive load assessment is based on at least one parameter among the real-time call volume, the complexity level of the current incident, and the call receiver's operation frequency and number of decisions per unit time. When the call receiving pressure is high and the call receiver's cognitive load is high, only the core key elements of the incident are highlighted to reduce unnecessary information interference. When the call receiving pressure is low and the cognitive load is low, more comprehensive related reference information is displayed to help the call receiver grasp the background of the incident more comprehensively. This optimizes the call receiver's work experience while ensuring call receiving efficiency and avoids the problems of information overload or insufficient information.

[0025] The dynamic data entry module also includes a multi-view generation unit, which generates differentiated form views for different roles in parallel based on the same structured contextual information. The differentiated form views include at least a detailed view for dispatchers, containing information completeness and analysis summary; an action view for on-site personnel, focusing on action lists and on-site risk warnings; and a decision view for commanders, integrating situational simulation and resource distribution. Among them, the action view can interact with the mobile terminals of on-site police officers, update the dispatch status and on-site feedback information in real time, and synchronize it to the detailed view and decision view. This meets the differentiated information needs of personnel in different positions, avoids the problem of information overload or missing key action information caused by a unified form, realizes cross-position information collaboration, avoids distortion and omissions in the information transmission process, and improves the overall collaborative efficiency of the response.

[0026] The verification and computation module includes a distributed verification network, which is used to cross-verify and confirm key elements in structured contextual information among multiple alarm receiving positions, on-site mobile terminals, and fixed monitoring nodes. The verification and computation module also includes a self-evolving knowledge graph unit, which is used to continuously optimize alarm handling rules, entity relationships, and classification patterns based on the feedback results of the handling, and to record the source and update trajectory of each knowledge item in the graph. This enables cross-verification of key alarm information, reduces judgment errors caused by a single information source, and can continuously iterate and optimize the system's cognitive ability based on actual handling results, so that the accuracy of the system's semantic analysis and information extraction can be continuously improved with use.

[0027] The self-evolving knowledge graph unit is configured to generate a traceable explanation chain for each rule, relationship, or pattern that the system automatically learns or optimizes during the knowledge update process. The explanation chain includes the decision basis, the data source used, and the event that triggered the learning. The verification and calculation module also includes a knowledge review interface, which is used to receive confirmation, correction, or rejection instructions from human operators for the knowledge items automatically evolved by the system, and update the confidence weight of the knowledge graph according to the instructions. This improves the interpretability and controllability of automatic knowledge updates, avoids unreliable erroneous knowledge generated by the system's autonomous learning, and allows for human intervention and correction, forming an iterative mechanism of "machine automatic learning + human review and calibration", which improves the reliability of the knowledge graph.

[0028] The situational cognition module also includes a prediction unit, which, based on the currently constructed situational cognition network, infers multiple development paths of the incident over a future period, assesses the probability of occurrence and resource impact of each path, and generates a situational evolution tree which is output to the dynamic reporting module. The prediction unit is connected to the multimodal fusion module and can dynamically correct and reassess the situational evolution tree based on new sensor data or on-site feedback acquired in real time. This helps alarm receivers and command personnel to predict the development direction of the incident in advance, prepare resource allocation and response plans in advance, avoid delays in response after the incident escalates, and adjust the prediction results in real time based on the latest on-site dynamics to ensure the adaptability and flexibility of the plan and improve the initiative of the overall emergency response.

[0029] The multimodal fusion module includes a spatiotemporal context map construction unit, which constructs a graph containing entity nodes and their relational edges of "people, vehicles, objects, places, and times" based on the fused multi-source information. The relationships include spatial location relationships, temporal sequence relationships, and behavioral interaction relationships. The multimodal fusion module also includes a conflict detection and resolution unit, which resolves conflicts based on the historical reliability, real-time confidence level, and inherent attributes of each information source when there are conflicts in the information about the same entity provided by different information sources. It outputs the fusion result marked with confidence level and conflict status, which can clearly sort out the correlation between multi-source heterogeneous information, intuitively present the complete spatiotemporal outline of the police scene, solve the problem of information conflicts from different data sources, improve the reliability of multi-source fused information, and provide an accurate and complete data foundation for subsequent contextual cognition and decision-making.

[0030] The processing module also includes a speech recognition module and a confidence modeling unit. The confidence modeling unit is used to attach a confidence score and a source identifier to each piece of information and each inference result in the contextual cognitive network, and to mark information with a confidence score below a preset threshold as high uncertainty information. When generating the form, the dynamic filling module visually highlights the content from high uncertainty information and, together with the intelligent follow-up questioning unit, generates secondary verification suggestions for the high uncertainty information. This allows dispatchers to quickly identify questionable uncertain information, clarify the content that needs to be verified, and avoid making handling decisions based on incorrect information. At the same time, combined with the intelligent follow-up questioning unit, it automatically initiates targeted verification, further improving the overall accuracy of the alarm information and reducing the handling risks caused by uncertain information.

[0031] It also includes a resource forward-looking orchestration module connected to processing module 3, which is used to generate pre-dispatch suggestions for police forces and pre-allocation plans for emergency resources at the alarm receiving stage based on predictive information output by the active contextual cognition engine and historical alarm patterns. The resource forward-looking orchestration module is connected to the geographic information system and the real-time police resource management system. Its generated pre-dispatch suggestions include the nearest patrol unit to be dispatched, the location of the suggested interception point, and the estimated arrival time.

[0032] When an alarm call comes into communication module 2, the system automatically associates the caller's number and quickly loads the number's historical alarm records, registered address, and other information in the sidebar. Simultaneously, based on the caller's base station location or keywords in the initial description, it automatically associates and retrieves public video surveillance streams and real-time data from IoT sensors in the vicinity of the incident location. Subsequently, the speech recognition module transcribes the call content in real time, and, in conjunction with the contextual cognition module, performs sentiment analysis and urgency assessment on the text, quickly extracting key entities and event types to initialize the contextual cognition network. Then, the intelligent questioning unit compares the information with the knowledge graph template in real time, detecting key gaps such as precise location and casualty details, and dynamically generates guided follow-up questions based on the caller's emotions, which are executed by the dispatcher or through system speech synthesis. During this process, newly acquired information is continuously injected into the cognitive network, building relationship chains between entities, thus completing the transformation from passive reception to actively constructing a dynamic contextual model within 60 seconds.

[0033] Subsequently, the multimodal fusion module deeply aligns and complements real-time video streams, sensor data, and speech semantics. For example, it associates "crowd gathering" identified in the video with "fighting" described in the speech. Simultaneously, its conflict detection unit verifies the consistency of multi-source information, marking conflicting content with confidence levels and pending verification status. Meanwhile, the prediction unit infers multiple future development paths and their probabilities based on the current context network, forming a context evolution tree. Next, the dynamic reporting module generates an adapted dynamic reporting interface based on context complexity and the real-time workload of the dispatcher, simultaneously presenting the dispatcher with a detailed view marked with conflict, pushing a focused action list mobile terminal view to on-site police officers, and generating a decision-making view integrating situational analysis and resource maps for command personnel. Furthermore, the resource forward-looking orchestration module operates concurrently, combining GIS and real-time police force data to generate pre-schedule suggestions and distributing key clues to relevant terminals for cross-verification through a collaborative verification network.

[0034] Then, on-site personnel report the dispatch status via mobile terminals, such as "Arrived, the hit-and-run vehicle is fleeing south." This information is transmitted back to the system core in real time. Subsequently, the situational awareness network and prediction model are dynamically updated, and key information such as the escape direction is corrected. At the same time, all related views (dispatcher's view, commander's view) are updated synchronously. In addition, key operations performed by on-duty personnel on the terminal (such as setting up roadblocks and sending patients to the hospital) are conveniently recorded. These status changes are synchronized to the interfaces of each view in real time, thereby ensuring that the three parties of receiving the alarm, command, and handling maintain coordination based on the same continuously evolving situational awareness.

[0035] Finally, after the incident was resolved, the police officers submitted a complete report containing the final results and resource consumption via the terminal. Subsequently, the verification and calculation module initiated a learning loop, comparing and analyzing the entire chain of data from this incident with earlier predictions. At the same time, the self-evolving knowledge graph unit extracted new patterns from the results, such as optimizing prediction weights in specific scenarios and adjusting the reliability scores of different information sources. The new knowledge was then submitted to the administrator for confirmation through the review interface. Finally, the confirmed knowledge was updated and integrated into the system's knowledge base and model to improve the accuracy and efficiency of handling similar incidents in the future, thus completing a complete closed loop from single incident handling to continuous system evolution.

[0036] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A semantic analysis-based intelligent alarm receiving and automatic reporting device, comprising a device body (1) and a communication module (2) and a processing module (3) installed inside the device body (1), wherein the communication module (2) and the processing module (3) are connected, characterized in that: The communication module (2) is used to receive alarm information, including voice alarm signals. The processing module (3) includes a context cognition module, a dynamic filling module, and a verification calculation module connected in sequence. The main body of the device (1) is also equipped with a multimodal fusion module connected to the processing module (3), which is used to access and integrate video streams, IoT sensor data and historical alarm data, and to align, verify and complement the semantic information of alarm voice, and output multimodal fusion information to the context cognition module.

2. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 1, characterized in that: The contextual cognition module is used to perform real-time semantic analysis on alarm information, construct and dynamically update the contextual cognition network. The contextual cognition network includes a cognitive layer for understanding current information, an intent layer for predicting subsequent development, a contextual layer for associating global information, and a decision layer for generating action suggestions. The dynamic filling module is used to generate a dynamic filling interface and form content that matches the current alarm type and complexity based on the structured contextual information output by the contextual cognition network.

3. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 2, characterized in that: The contextual cognition module includes an intelligent questioning unit, which is used to detect information gaps in the contextual cognition network in real time and initiate proactive questioning based on a preset questioning strategy library. The questioning strategy library includes clarification questioning strategies for vague descriptions, guidance questioning strategies for emotionally agitated callers, and confirmation questioning strategies for key information. The intelligent questioning unit is configured to dynamically select and combine different questioning strategies based on the type and urgency of the information gap and the caller's real-time emotional analysis results, and adopts a progressively converging questioning path.

4. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 2, characterized in that: The dynamic reporting module includes an interface adaptation unit, which is used to dynamically adjust the information complexity and presentation density of the dynamic reporting interface according to the real-time assessment of the cognitive load status of the dispatcher. The assessment of the cognitive load status is based on at least one parameter among the real-time number of dispatches, the complexity level of the current alarm, and the dispatcher's operation frequency and number of decisions per unit time.

5. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 4, characterized in that: The dynamic reporting module also includes a multi-view generation unit, which generates differentiated form views for different roles in parallel based on the same structured contextual information. The differentiated form views include at least a detailed view for dispatchers that includes information completeness and analysis summary, an action view for on-site personnel that focuses on action lists and on-site risk warnings, and a decision view for commanders that integrates situation simulation and resource distribution.

6. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 2, characterized in that: The verification calculation module includes a distributed verification network, which is used to cross-verify and confirm the key elements in the structured context information among multiple alarm receiving positions, on-site mobile terminals and fixed monitoring nodes. The verification calculation module also includes a self-evolving knowledge graph unit, which is used to continuously optimize the alarm handling rules, entity relationships and classification patterns based on the handling feedback results, and record the source and update trajectory of each knowledge item in the graph.

7. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 6, characterized in that: The self-evolving knowledge graph unit is configured to generate a traceable explanation chain for each rule, relationship, or pattern that the system automatically learns or optimizes during the knowledge update process. The explanation chain includes the decision basis, the data source used, and the event that triggers the learning. The verification calculation module also includes a knowledge review interface, which is used to receive confirmation, correction, or rejection instructions from human operators for the knowledge items that the system automatically evolves, and to update the confidence weight of the knowledge graph according to the instructions.

8. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 2, characterized in that: The context cognition module also includes a prediction unit, which is used to deduce multiple development paths of the alarm in the future period based on the currently constructed context cognition network, and to evaluate the probability of occurrence and resource impact of each path, forming a context evolution tree and outputting it to the dynamic reporting module. The prediction unit is connected to the multimodal fusion module and can dynamically correct and re-evaluate the context evolution tree based on new sensor data or on-site feedback acquired in real time.

9. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 1, characterized in that: The multimodal fusion module includes a spatiotemporal context graph construction unit, which is used to construct a graph of entity nodes and their relation edges from the fused multi-source information. The relations include spatial location relations, temporal sequence relations, and behavioral interaction relations. The multimodal fusion module also includes a conflict detection and resolution unit, which is used to resolve conflicts based on the historical reliability, real-time confidence level, and inherent attributes of each information source when there are conflicts in the information about the same entity provided by different information sources, and outputs a fusion result marked with confidence level and conflict state.

10. The intelligent alarm receiving and automatic reporting device based on semantic analysis according to claim 3, characterized in that: The processing module also includes a speech recognition module and a confidence modeling unit. The confidence modeling unit is used to add a confidence score and a source identifier to each piece of information and each inference result in the contextual cognitive network, and to mark information with a confidence score lower than a preset threshold as high uncertainty information. When generating the form, the dynamic filling module visually highlights the content from the high uncertainty information and, together with the intelligent follow-up questioning unit, generates a secondary verification suggestion for the high uncertainty information.

Images