Vehicle-mounted speech recognition law enforcement linkage system and method based on edge calculation

Abstract

The invention relates to the technical field of traffic law enforcement, in particular to a vehicle-mounted voice recognition law enforcement linkage system and method based on edge calculation. A data storage unit; a voice acquisition unit; the analysis unit determines law enforcement regulations of the current law enforcement area, obtains law enforcement intentions in the voice information, identifies and confirms a law enforcement target according to the law enforcement intentions and law enforcement scenes, and determines a current law enforcement type according to the law enforcement regulations, the law enforcement intentions, the law enforcement scenes and the law enforcement target. Obtaining a standardized operation guide corresponding to the current law enforcement type, pushing related typical cases, and sending an equipment control code according to the standardized operation guide; an equipment linkage unit; and an evidence chain management module. According to the invention, through linkage of voice recognition and equipment, automatic control based on voice information can be realized, a plurality of manual operation links in traditional law enforcement are integrated into a one-button flow triggered by voice information, the operation time is reduced, and the practicability is improved.

Description

Technical Field

[0001] The present invention relates to the field of traffic law enforcement technology, and in particular to an on-board voice recognition law enforcement linkage system and method based on edge computing. Background Art

[0002] With the advancement of the construction of a rule of law society and the increasing public demand for standardized law enforcement, the application of vehicle-mounted law enforcement equipment in field law enforcement scenarios such as traffic management and security patrols is becoming increasingly widespread.

[0003] Patent publication number CN103413438A discloses a vehicle-mounted mobile police enforcement platform, comprising: a license plate recognition system; a central data processing system that receives vehicle information collected by the license plate recognition system and compares it with a vehicle database; an alarm interception system that generates alarm information; an on-site processing system that performs on-site processing on vehicles based on the interception inspection results; and a wireless communication system that transmits information collected by the license plate recognition system to the central data processing system, transmits the compared information to the alarm interception system, and wirelessly connects the on-site processing system to various law enforcement systems. During the enforcement process, public officials on the vehicle-mounted mobile police enforcement platform must manually operate various on-board devices (warning lights, cameras, display screens, lift lights, etc.) and record enforcement data, which prevents rapid data collection and reduces enforcement efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide an on-board voice recognition law enforcement linkage system and method based on edge computing, so as to solve the problems in the prior art of needing to manually operate various on-board equipment (warning lights, cameras, display screens, lifting lights, etc.) and record law enforcement data, which makes it impossible to quickly collect law enforcement data and leads to low law enforcement efficiency.

[0005] The present invention provides a vehicle-mounted voice recognition and law enforcement linkage system based on edge computing, comprising:

[0006] A data storage unit for storing the law enforcement regulations of each law enforcement area, the standardized operating guidelines corresponding to each law enforcement type, and corresponding typical cases;

[0007] A positioning unit, used to obtain the location information of the law enforcement vehicle;

[0008] Voice collection unit, used to collect voice information of law enforcement personnel;

[0009] an analysis unit configured as an offline large language model, configured to determine the law enforcement regulations of the current law enforcement area based on the location information, locally recognize the voice information of the law enforcement personnel, obtain the law enforcement intention in the voice information, determine the law enforcement scene based on the real-time environmental perception data obtained by the vehicle-mounted sensor, identify and confirm the law enforcement target based on the law enforcement intention and the law enforcement scene, determine the current law enforcement type based on the law enforcement regulations, law enforcement intention, law enforcement scene and law enforcement target, obtain standardized operating instructions corresponding to the current law enforcement type and push relevant typical cases, and issue a device control code according to the standardized operating instructions;

[0010] A device linkage unit, configured to receive the device control code and control the vehicle-mounted law enforcement device to perform law enforcement and evidence collection according to the device control code, so as to obtain law enforcement data;

[0011] The evidence chain management unit is used to add time and space stamps to the audio and video and equipment operation records in the law enforcement data, and synchronize them to the blockchain evidence storage node after encryption.

[0012] As the preferred technical solution for the vehicle-mounted voice recognition law enforcement linkage system based on edge computing, the positioning unit is used to obtain the location information of the law enforcement vehicle, and obtain the road attribute information of the location of the law enforcement vehicle based on the location information. The road attribute information includes: the number of lanes, speed limit signs, one-way or two-way lanes, and road surface marking lines.

[0013] As the preferred technical solution for the vehicle-mounted voice recognition law enforcement linkage system based on edge computing, the voice collection unit has a voiceprint recognition function, which can identify the identities of different law enforcement personnel and match the corresponding law enforcement process operation permissions according to the personnel permissions.

[0014] As a preferred technical solution for the vehicle-mounted voice recognition and law enforcement linkage system based on edge computing, the analysis unit locally recognizes the voice information of the law enforcement personnel and obtains the law enforcement intention in the voice information, including:

[0015] Analyze the voice information to obtain key information from the voice information;

[0016] The key information includes: action keywords, law enforcement objects and limiting conditions. In response to the incompleteness of the key information, the key information is supplemented based on the law enforcement scenario, and the law enforcement intention is determined based on the key information.

[0017] As a preferred technical solution for the in-vehicle voice recognition law enforcement linkage system based on edge computing, the analysis unit determines the current enforcement type based on enforcement regulations, enforcement intent, enforcement scenario, and enforcement objectives, including:

[0018] Possible matching enforcement types are screened based on the enforcement intention, enforcement scenario, and enforcement target. In response to the existence of multiple possible enforcement types, the semantic similarity between the key information and the standard terms of each enforcement type is calculated, and the enforcement type with the highest similarity is selected as the current enforcement type.

[0019] As the preferred technical solution for the in-vehicle voice recognition law enforcement linkage system based on edge computing, the analysis unit adopts a dynamic weight algorithm trained based on historical law enforcement data when calculating the semantic similarity between key information and standard terms of law enforcement types, and dynamically adjusts the weight coefficient according to the importance of each keyword in different scenarios.

[0020] As the preferred technical solution for the in-vehicle voice recognition law enforcement linkage system based on edge computing, the analysis unit screens possible matching law enforcement types according to the law enforcement intention, law enforcement scenario and law enforcement target, and screens the candidate range in combination with the current time, weather and traffic flow.

[0021] The present invention also provides a vehicle-mounted voice recognition law enforcement linkage method based on edge computing, comprising:

[0022] Step S1, storing the law enforcement regulations of each law enforcement area, the standardized operating guidelines corresponding to each law enforcement type, and the corresponding typical cases;

[0023] Step S2, obtaining the location information of the law enforcement vehicle;

[0024] Step S3, collecting voice information of law enforcement personnel;

[0025] In step S4, the offline large language model determines the law enforcement regulations of the current law enforcement area based on the location information, locally recognizes the voice information of the law enforcement personnel, obtains the law enforcement intention in the voice information, determines the law enforcement scenario based on the real-time environmental perception data obtained by the vehicle-mounted sensor, comprehensively identifies and confirms the law enforcement target based on the law enforcement intention and law enforcement scenario, determines the current law enforcement type based on the law enforcement regulations, law enforcement intention, law enforcement scenario, and law enforcement target, obtains the standardized operation guide corresponding to the current law enforcement type and pushes relevant typical cases, and issues a device control code according to the standardized operation guide;

[0026] Step S5, receiving the device control code and controlling the vehicle-mounted law enforcement device to perform law enforcement and evidence collection according to the device control code to obtain law enforcement data;

[0027] Step S6: add time and space stamps to the audio and video and equipment operation records in the law enforcement data, encrypt them and synchronize them to the blockchain evidence storage node.

[0028] As a preferred technical solution for the vehicle-mounted voice recognition and law enforcement linkage method based on edge computing, local recognition of the voice information of the law enforcement personnel includes:

[0029] Extracting voiceprint features from the voice information and matching them with a pre-registered database of law enforcement officers’ voiceprints in real time;

[0030] If the voiceprint matching fails, the on-board surround view camera will be automatically activated to collect image information around the law enforcement vehicle;

[0031] The abnormal voiceprint records, encrypted image data and location information are uploaded to the command center security platform through the on-board 5G module.

[0032] As the preferred technical solution for the in-vehicle voice recognition law enforcement linkage method based on edge computing, the evidence chain adopts a multi-node blockchain synchronous evidence storage mechanism.

[0033] Compared with the existing technology, the beneficial effect of the present invention is that the system is freed from dependence on the network through the deployment of an offline large language model, and can still operate stably in areas without network coverage such as remote mountainous areas and tunnels. By linking voice recognition with equipment, automatic control based on voice information can be achieved, and multiple manual operation links in traditional law enforcement can be integrated into a one-click process triggered by voice information, thereby improving practicality.

[0034] Furthermore, edge computing is a distributed computing paradigm that transfers data processing and computing tasks from traditional centralized cloud computing platforms to local nodes closer to data generation devices. It has the characteristics of low latency, high bandwidth efficiency, high privacy and security. In the present invention, the data generation device is a vehicle-mounted law enforcement device, and edge computing is implemented using an offline large language model. The user voice is processed locally to avoid the slow response speed caused by multiple law enforcement vehicles connecting to the centralized cloud computing platform at the same time, thereby further increasing practicality.

[0035] Furthermore, through the application of data encryption and blockchain evidence storage technology, data can be processed directly in the car, eliminating the risk of law enforcement data being tampered with and ensuring the security of data during the law enforcement process.

[0036] Furthermore, this system can process instructions and response controls in real time without relying on a central server, and has the triple characteristics of anti-network loss, high response, and high automation, thereby further improving its practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a structural block diagram of an in-vehicle voice recognition and law enforcement linkage system based on edge computing according to an embodiment of the present invention;

[0038] Figure 2 This is a flowchart of the steps of the vehicle-mounted voice recognition law enforcement linkage method based on edge computing in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present application by illustrating the examples of the present application.

[0040] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, the elements defined by the phrase "comprising..." do not exclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0041] See also Figure 1 As shown in FIG, it is a structural block diagram of the vehicle-mounted voice recognition law enforcement linkage system based on edge computing according to an embodiment of the present invention, including:

[0042] A data storage unit for storing the law enforcement regulations of each law enforcement area, the standardized operating guidelines corresponding to each law enforcement type, and corresponding typical cases;

[0043] A positioning unit, used to obtain the location information of the law enforcement vehicle;

[0044] Voice collection unit, used to collect voice information of law enforcement personnel;

[0045] An analysis unit is configured to deploy an offline large language model on an on-board embedded platform, and is used to determine the law enforcement regulations of the current law enforcement area based on location information, locally recognize the voice information of law enforcement personnel, obtain the law enforcement intention in the voice information, determine the law enforcement scene based on real-time environmental perception data obtained by on-board sensors, comprehensively identify the law enforcement intention and law enforcement scene and confirm the law enforcement target, determine the current law enforcement type based on the law enforcement regulations, law enforcement intention, law enforcement scene and law enforcement target, obtain standardized operating instructions corresponding to the current law enforcement type and push relevant typical cases, and issue a device control code according to the standardized operating instructions;

[0046] The device linkage unit is used to receive the device control code and control the vehicle-mounted law enforcement equipment to perform law enforcement and evidence collection according to the device control code to obtain law enforcement data;

[0047] The evidence chain management unit is used to add time and space stamps to the audio and video and equipment operation records in the law enforcement data, and encrypt and synchronize them to the blockchain evidence storage node.

[0048] During implementation, the positioning unit uses a high-precision Beidou positioning module to obtain vehicle latitude and longitude, speed, direction and other location information in real time; the data storage unit adopts a hierarchical storage architecture, storing commonly used law enforcement regulations in a high-speed cache area and less commonly used regulations in a large-capacity hard disk; the voice acquisition unit uses a high-sensitivity directional microphone, which is installed on the shoulder of the law enforcement officer to ensure clear voice collection; the analysis unit uses a lightweight offline large language model deployed on the vehicle embedded platform, and adapts hardware resources through model compression and quantization technology; the equipment linkage unit is connected to the vehicle warning lights, cameras, recording equipment, etc. through a standardized interface; the evidence chain management unit uses distributed storage to synchronize data to multiple blockchain evidence nodes.

[0049] In detail, through high-precision positioning and real-time acquisition of road attribute information, the present invention allows law enforcement personnel to intuitively view a detailed map of the current law enforcement area on the vehicle-mounted display screen, including the number of lanes, speed limit signs, one-way or two-way lanes, and road surface marking lines, thereby further improving practicality.

[0050] Furthermore, the positioning unit is used to obtain the location information of the law enforcement vehicle, and obtain the road attribute information of the location of the law enforcement vehicle based on the location information. The road attribute information includes: the number of lanes, speed limit signs, one-way or two-way lanes, and road surface marking lines.

[0051] During implementation, the potential enforcement area will be divided into several zones. When an enforcement vehicle enters a new zone, road attribute data will be automatically updated. Onboard sensors will assist in detecting road surface markings. A road attribute information database will also be established and regularly updated and maintained, further enhancing the offline capabilities of enforcement vehicles. Accurate road attribute information will enable law enforcement officers to conduct targeted enforcement based on different road types and speed limits, providing appropriate enforcement standards and procedures, thereby improving the accuracy and effectiveness of law enforcement.

[0052] Specifically, the voice collection unit has a voiceprint recognition function, which can identify the identities of different law enforcement personnel and match the corresponding law enforcement process operation permissions according to the personnel's permissions.

[0053] In practice, the voice collection unit incorporates a deep learning-based voiceprint recognition chip, utilizing the i-vector and PLDA algorithms. After identity verification, law enforcement officers upload their voiceprint templates to the system management client. The system management client then assigns operational permissions based on the officer's position and responsibilities. These permissions are divided into three levels: officer, sergeant, and captain, each corresponding to a distinct function menu and scope of operation. Before each voice collection session, the system automatically performs voiceprint recognition. Only successful recognition allows subsequent operations. If recognition fails three times in a row, the system locks the device and sends an alert to the administrator.

[0054] Furthermore, the analysis unit locally recognizes the voice information of the law enforcement personnel, obtains the law enforcement intention in the voice information, parses the voice information, and obtains key information in the voice information, the key information including: action keywords, law enforcement targets, and limiting conditions. In response to incomplete key information, the key information is supplemented based on the law enforcement scenario, and the law enforcement intention is determined based on the key information, wherein:

[0055] Action keywords include: take photo, warn, etc.

[0056] Enforcement targets include: illegal parking, speeding, vehicles, etc.;

[0057] Restrictions include: immediate, full recording, etc.

[0058] During implementation, the analysis module uses a natural language processing model based on the Transformer architecture to perform word segmentation, part-of-speech tagging, and dependency syntax analysis on voice information, extracting key information such as action keywords, enforcement targets, and limiting conditions. The system establishes a dynamic keyword library containing more than 2,000 law enforcement-related words, and updates it by regularly collecting new law enforcement terms and descriptions of illegal behaviors; when key information is missing, it uses current positioning information, historical law enforcement records, and environmental data for semantic completion. For example, if the voice information is only "processing", the system automatically completes it with "processing illegally parked vehicles" based on the current location being a parking lot. The law enforcement intention classification model is trained based on a large amount of annotated historical voice information data, and uses a Softmax classifier to divide law enforcement intentions into several subcategories; the law enforcement type rule library is constructed by combining expert knowledge and machine learning. Each rule contains information such as the law enforcement type name, triggering conditions, evidence requirements, and disposal procedures.

[0059] Furthermore, the analysis unit determines the current enforcement type based on enforcement regulations, enforcement intentions, enforcement scenarios, and enforcement goals, including:

[0060] Possible matching enforcement types are screened based on enforcement intentions, enforcement scenarios, and enforcement objectives. In response to the existence of multiple possible enforcement types, the semantic similarity between key information and standard terms of each enforcement type is calculated, and the enforcement type with the highest similarity is selected as the current enforcement type.

[0061] Specifically, by parsing, analyzing, and supplementing law enforcement officers' voice messages, the system can receive diverse and ambiguous voice messages during law enforcement, effectively avoiding law enforcement errors caused by misunderstanding instructions. Intelligent completion and intent classification enable the system to quickly and accurately match enforcement types, reducing the time officers spend determining enforcement types and shortening the time required for a single enforcement action, thereby improving practicality.

[0062] Furthermore, when calculating the semantic similarity between key information and standard terms of law enforcement types, the analysis unit adopts a dynamic weighting algorithm trained based on historical law enforcement data, and dynamically adjusts the weight coefficient according to the importance of each keyword in different scenarios.

[0063] In detail, an embodiment of the present invention provides a method for calculating the semantic similarity between key information and standard terms of law enforcement types, including:

[0064] Step 1: Build a law enforcement scenario classification system

[0065] Based on historical law enforcement records, we created a law enforcement record dataset and divided law enforcement scenarios into three categories: daily patrol, traffic accident handling, and special rectification. Each category is further subdivided into multiple small scenarios:

[0066] Daily patrols include: vehicle inspections, pedestrian violation handling, illegal parking disposal and other small scenes.

[0067] Traffic accident handling includes small scenes such as on-site investigation, responsibility determination, and rescue of the injured.

[0068] Special rectification categories include: drunk driving investigation and punishment, overloading rectification, hazardous chemicals inspection and other small scenes.

[0069] Step 2: Create a scenario-based terminology library

[0070] For each small scenario, three key elements are extracted from historical law enforcement records:

[0071] Action keywords (e.g., inspect, intercept, seize)

[0072] Enforcement targets (e.g. trucks, pedestrians, driver's licenses)

[0073] Qualifiers (e.g., immediate, comprehensive, focused)

[0074] Set standard terms and semantically similar words for key elements in each small scenario. For example: Standard term: "carrying", Semantically similar words: "lifting" (similarity 0.92), "holding" (similarity 0.66).

[0075] Standard term: "van", Semantically similar words: "truck" (similarity 0.85), "box car" (similarity 0.95).

[0076] Standard term: "immediately", Semantically similar words: "at once" (similarity 0.98), "right away" (similarity 0.97).

[0077] For each type of key element in each small scenario, different weight coefficients are set. For example, in the small scenario of "immediately stop the truck", the weight coefficients for each type of key element are as follows: Action keyword weight coefficient: 0.25, Law enforcement object weight coefficient: 0.40, Limiting condition weight coefficient: 0.35. The specific values of the weight coefficients are determined by statistically analyzing historical law enforcement data, the experience of domain experts, and automatically optimizing the weights based on feedback data (for example, when it is found that the "seize" instruction is often corrected in the hazardous chemicals scenario, the object weight is automatically increased), etc. The methods for determining the weight values are all existing technologies and will not be elaborated here.

[0078] Step 3: Real-time semantic similarity calculation process

[0079] When receiving the voice instruction from the law enforcement officer, execute the following calculation process:

[0080] Extract three types of key elements from the voice instruction. For example: Action keyword: "stop", Law enforcement object: "box car", Limiting condition: "at once".

[0081] Query the similarity of each element in the corresponding scenario term library, that is:

[0082] The similarity between "stop" and the standard action word "intercept": 0.93.

[0083] The similarity between "box car" and the standard object word "van": 0.95.

[0084] The similarity between "at once" and the standard limiting word "immediately": 0.98.

[0085] Calculate the semantic similarity between the key information and the standard terms of the law enforcement type:

[0086] The semantic similarity between the key information and the standard terms of the law enforcement type = (action similarity × action weight + object similarity × object weight + limiting similarity × limiting weight) / total weight.

[0087] In detail, standard terms and semantically similar words are set for the key elements in each small scene. The domain BERT in the prior art can be combined to generate law enforcement object word vectors, and the similarity of standard terms and semantically similar words can be calculated based on the law enforcement object word vectors. Other calculation methods that can calculate the similarity of standard terms and semantically similar words in the embodiments of the present invention are also acceptable. The embodiments of the present invention do not impose specific restrictions on the calculation methods, and they are all prior art, so they will not be elaborated here.

[0088] The law enforcement record dataset includes voice commands, law enforcement types, processing results, and environmental data. The XGBoost algorithm is used to train a dynamic weight model to determine the importance of different keywords in law enforcement type matching.

[0089] Furthermore, the analysis unit screens possible matching enforcement types based on enforcement intentions, enforcement scenarios, and enforcement targets, and screens the candidate range in combination with the current time, weather, and traffic flow.

[0090] During implementation, the analysis module accesses the real-time data platforms of the China Meteorological Administration and the Ministry of Transportation through an API interface, obtaining data on environmental factors such as the current time, weather (temperature, humidity, rainfall, wind speed, etc.), and traffic flow (congestion index, vehicle volume). A rule library is established to associate environmental factors with enforcement types. For example, "When rainfall exceeds 50mm, drone evidence collection is not recommended," and "When the traffic congestion index exceeds 8, rapid diversion enforcement procedures are preferred." When selecting candidate enforcement types, the system first filters the rule library based on environmental factors, eliminating enforcement types that are unsuitable for the current environment. Further screening is then performed based on enforcement intent and other criteria.

[0091] In detail, the present invention makes law enforcement decisions more scientific and reasonable by incorporating environmental factors into the considerations of law enforcement type matching, effectively avoiding situations where law enforcement operations cannot be performed or the performance is poor due to environmental unsuitability. For example, in windy weather, the law enforcement plan using high-altitude camera equipment is automatically excluded to reduce the risk of equipment damage and law enforcement safety hazards. The law enforcement strategy is dynamically adjusted according to traffic flow, and traffic diversion is given priority during congested periods to improve road traffic efficiency and reduce the additional impact of law enforcement activities on traffic; during periods of low traffic volume, detailed law enforcement inspections are carried out to make full use of law enforcement resources and improve overall law enforcement efficiency.

[0092] Furthermore, the voice collection unit integrates a noise reduction microphone array, supports -20dB environmental noise suppression, and triggers real-time collection of voice information through the PTT button.

[0093] In practice, the voice collection unit's noise-cancelling microphone array utilizes adaptive beamforming technology, enabling real-time tracking of the sound source and suppressing ambient noise from other directions. A built-in DNN (deep neural network) noise reduction model performs secondary noise reduction on the collected voice signals, ensuring voice clarity even in -20dB noise levels. The PTT button features a military-grade waterproof and dustproof design, with an IP68 rating, ensuring operation in harsh environments such as heavy rain and dust. The button's surface features a non-slip texture, making it easy for officers to operate while wearing gloves. The system offers microphone sensitivity adjustment, supporting five levels of sensitivity. Officers can manually adjust the sensitivity based on the specific usage scenario, such as lowering the sensitivity in quiet indoor environments and increasing it on noisy streets.

[0094] Specifically, the device linkage unit controls the on-board equipment through the CAN bus. The on-board equipment includes: warning lights, on-board surround-view cameras and display screens.

[0095] During implementation, the device linkage unit uses the CAN bus as its primary communication method, adhering to the ISO11898-2 standard protocol and a communication rate of 500kbps to ensure real-time and stable data transmission. The RS485 interface is retained as a backup communication method for connecting devices that do not support the CAN bus. Each onboard device is assigned a unique 11-bit CAN device control code, which the device linkage unit uses to quickly identify and control the device. A device status monitoring mechanism is established, sending status query commands to the device every 100ms to obtain real-time operating status (such as power status, storage space, and operating mode). A strict device control priority strategy is established, with safety-related devices such as warning lights and emergency alarms given the highest priority, evidence collection devices such as cameras and recording equipment given the second highest priority, and auxiliary devices such as display screens and printers given normal priority. When multiple control commands arrive simultaneously, they are executed in order of priority.

[0096] See also Figure 2 As shown, it is a flowchart of the steps of the vehicle-mounted voice recognition law enforcement linkage method based on edge computing in an embodiment of the present invention, including:

[0097] Step S1, storing the law enforcement regulations of each law enforcement area, the standardized operating guidelines corresponding to each law enforcement type, and the corresponding typical cases.

[0098] Step S2: Obtain the location information of the law enforcement vehicle.

[0099] Step S3: collecting voice information of law enforcement personnel.

[0100] In step S4, the offline large language model determines the law enforcement regulations of the current law enforcement area based on the location information, locally recognizes the voice information of the law enforcement personnel, obtains the law enforcement intention in the voice information, determines the law enforcement scene based on the real-time environmental perception data obtained by the on-board sensors, comprehensively identifies the law enforcement intention and law enforcement scene, and confirms the law enforcement target. The current law enforcement type is determined based on the law enforcement regulations, law enforcement intention, law enforcement scene and law enforcement target, obtains the standardized operation guidelines corresponding to the current law enforcement type and pushes relevant typical cases, and issues the device control code according to the standardized operation guidelines.

[0101] Step S5: receiving the device control code and controlling the vehicle-mounted law enforcement device to perform law enforcement and evidence collection according to the device control code to obtain law enforcement data.

[0102] Step S6: Add time and space stamps to the audio and video and equipment operation records in the law enforcement data, encrypt them and synchronize them to the blockchain evidence storage node.

[0103] Furthermore, the large language model performs semantic recognition of speech information based on the transformer structure.

[0104] Furthermore, local recognition of the voice information of law enforcement officers is performed, including:

[0105] Extract voiceprint features from voice information and perform real-time matching with the pre-registered law enforcement officer voiceprint database;

[0106] If the voiceprint matching fails, the on-board surround view camera will be automatically activated to collect image information around the law enforcement vehicle;

[0107] The abnormal voiceprint records, encrypted image data and location information are uploaded to the command center security platform through the on-board 5G module.

[0108] Furthermore, the evidence chain adopts a multi-node blockchain synchronous evidence storage mechanism.

[0109] Furthermore, typical cases are pushed with graphic and text prompts that match the law enforcement type.

[0110] Furthermore, the device control code execution priority mechanism is designed, and the priority mechanism from high to low is warning light, camera and display screen.

[0111] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. A person skilled in the art would be able to make other variations or modifications based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A vehicle-mounted voice recognition law enforcement linkage system based on edge computing, characterized in that: include: A data storage unit for storing the law enforcement regulations of each law enforcement area, the standardized operating guidelines corresponding to each law enforcement type, and corresponding typical cases; A positioning unit, used to obtain the location information of the law enforcement vehicle; Voice collection unit, used to collect voice information of law enforcement personnel; an analysis unit configured as an offline large language model, configured to determine the law enforcement regulations of the current law enforcement area based on the location information, locally recognize the voice information of the law enforcement personnel, obtain the law enforcement intention in the voice information, determine the law enforcement scene based on the real-time environmental perception data obtained by the vehicle-mounted sensor, identify and confirm the law enforcement target based on the law enforcement intention and the law enforcement scene, determine the current law enforcement type based on the law enforcement regulations, law enforcement intention, law enforcement scene and law enforcement target, obtain standardized operating instructions corresponding to the current law enforcement type and push relevant typical cases, and issue a device control code according to the standardized operating instructions; A device linkage unit, configured to receive the device control code and control the vehicle-mounted law enforcement device to perform law enforcement and evidence collection according to the device control code, so as to obtain law enforcement data; The evidence chain management unit is used to add time and space stamps to the audio and video and equipment operation records in the law enforcement data, and synchronize them to the blockchain evidence storage node after encryption.

2. The vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to claim 1 is characterized in that: The positioning unit is used to obtain the location information of the law enforcement vehicle and obtain the road attribute information of the location of the law enforcement vehicle based on the location information. The road attribute information includes: the number of lanes, speed limit signs, one-way or two-way lanes, and road surface markings.

3. The vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to claim 2 is characterized in that: The voice collection unit has a voiceprint recognition function, which can identify the identities of different law enforcement personnel and match the corresponding law enforcement process operation permissions according to the personnel permissions.

4. The vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to claim 1 is characterized in that: The analyzing unit locally recognizes the voice information of the law enforcement officer to obtain the law enforcement intention in the voice information, including: Analyze the voice information to obtain key information from the voice information; The key information includes: action keywords, law enforcement objects and limiting conditions. In response to the incompleteness of the key information, the key information is supplemented based on the law enforcement scenario, and the law enforcement intention is determined based on the key information.

5. The vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to claim 4 is characterized in that: The analysis unit determines the current enforcement type based on enforcement regulations, enforcement intentions, enforcement scenarios, and enforcement targets, including: Possible matching enforcement types are screened based on the enforcement intention, enforcement scenario, and enforcement objectives. In response to the existence of multiple possible enforcement types, the semantic similarity between the key information and the standard terms of each enforcement type is calculated, and the enforcement type with the highest similarity is selected as the current enforcement type.

6. The vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to claim 5 is characterized in that: When calculating the semantic similarity between key information and standard terms of law enforcement types, the analysis unit adopts a dynamic weight algorithm trained based on historical law enforcement data, and dynamically adjusts the weight coefficient according to the importance of each keyword in different scenarios.

7. The vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to claim 5 is characterized in that: The analysis unit screens possible matching enforcement types based on the enforcement intention, enforcement scenario, and enforcement target, and screens the candidate range in combination with the current time, weather, and traffic flow.

8. A vehicle-mounted voice recognition and law enforcement linkage method based on edge computing, used to implement the vehicle-mounted voice recognition and law enforcement linkage system based on edge computing according to any one of claims 1 to 7, characterized in that: include: Step S1, storing the law enforcement regulations of each law enforcement area, the standardized operating guidelines corresponding to each law enforcement type, and the corresponding typical cases; Step S2, obtaining the location information of the law enforcement vehicle; Step S3, collecting voice information of law enforcement personnel; In step S4, the offline large language model determines the law enforcement regulations of the current law enforcement area based on the location information, locally recognizes the voice information of the law enforcement personnel, obtains the law enforcement intention in the voice information, determines the law enforcement scenario based on the real-time environmental perception data obtained by the vehicle-mounted sensor, comprehensively identifies and confirms the law enforcement target based on the law enforcement intention and law enforcement scenario, determines the current law enforcement type based on the law enforcement regulations, law enforcement intention, law enforcement scenario, and law enforcement target, obtains the standardized operation guide corresponding to the current law enforcement type and pushes relevant typical cases, and issues a device control code according to the standardized operation guide; Step S5, receiving the device control code and controlling the vehicle-mounted law enforcement device to perform law enforcement and evidence collection according to the device control code to obtain law enforcement data; Step S6: add time and space stamps to the audio and video and equipment operation records in the law enforcement data, encrypt them and synchronize them to the blockchain evidence storage node.

9. The vehicle-mounted voice recognition and law enforcement linkage method based on edge computing according to claim 8 is characterized in that: Performing local recognition of the voice information of the law enforcement officer, including: Extracting voiceprint features from the voice information and matching them with a pre-registered database of law enforcement officers’ voiceprints in real time; If the voiceprint matching fails, the on-board surround view camera will be automatically activated to collect image information around the law enforcement vehicle; The abnormal voiceprint records, encrypted image data and location information are uploaded to the command center security platform through the on-board 5G module.

10. The vehicle-mounted voice recognition and law enforcement linkage method based on edge computing according to claim 9 is characterized in that: The evidence chain adopts a multi-node blockchain synchronous evidence storage mechanism.

Images