Public network intercom terminal remote dispatching method and system based on AI voice interaction

Abstract

This application relates to the technical field of public network intercom, and discloses a remote scheduling method and system for public network intercom terminals based on AI voice interaction. The terminal collects the user's voice data stream; performs streaming recognition and intent parsing on the voice data stream; maps structured instructions to corresponding functional module APIs according to a preset extensible instruction table; verifies the user's role and current scene state through an RBAC scene lock model to determine whether the user has the operation permission to execute the API; if the permission verification passes, a priority queue mechanism is used to distribute the API to the corresponding functional module, controlling the parallel execution of operations by each functional module while arbitrating conflicts between concurrent tasks; otherwise, a permission insufficient prompt message is directly output to the terminal; after feedback and aggregation of the execution results, dual feedback of voice broadcast and interface prompt is output to the terminal, while recording the execution log, solving the problems of long operation chains and delayed feedback in the prior art.

Description

Technical Field

[0001] This application relates to the technical field of public network intercom, and in particular to a remote scheduling method and system for public network intercom terminals based on AI voice interaction. Background Technology

[0002] With its core advantages such as wide-area coverage, multimedia integration, and group call response, public network intercom systems have become a key support tool for collaborative group operations. In actual dispatching scenarios, especially in emergency response and sudden task response, the convenience and real-time nature of dispatching operations directly determine the effectiveness of task execution.

[0003] Existing public network intercom systems still suffer from numerous technical bottlenecks: The operation process is cumbersome, requiring multiple manual clicks on a PC to complete steps such as location, filtering, streaming, and group creation, resulting in long operation paths, high risk of accidental touches, and delayed response in emergencies; Voice control capabilities are limited, supporting only single-function triggers (such as "switch groups"), unable to parse cross-module composite commands (such as "filter online terminals in a specific area and enable multi-screen video"), and low coupling between commands and functions; Rigid permission management uses fixed permission configurations that cannot be dynamically adjusted according to the scenario, leaving ordinary members without necessary operational permissions in emergencies, while excessive permissions in non-emergency situations easily lead to misoperations; The feedback and resource scheduling mechanisms are imperfect, requiring manual refresh and querying of execution results, resulting in high feedback delays, and resource conflicts are prone to occur when multiple modules perform concurrent tasks, leading to low scheduling efficiency. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method and system for remote scheduling of public network intercom terminals based on AI voice interaction.

[0005] Firstly, this application provides a method for remotely dispatching a public network intercom terminal based on AI voice interaction, including the following steps: S1: The terminal collects the user's voice commands, preprocesses the voice commands, and outputs a standard format voice data stream. S2: Perform streaming recognition and intent parsing on the voice data stream, and output a structured instruction containing operation intent and slot information; S3: Based on a preset extensible instruction list, map the structured instructions to corresponding functional module APIs; the functional modules include a video scheduling module, a GIS positioning module, and a group management module; S4: Verify the user role and current scene state using the RBAC scene lock model to determine whether the user has the permission to execute the operation corresponding to the API; S5: If the permission verification passes, the API is distributed to the corresponding functional modules using a priority queue mechanism. The functional modules are controlled to perform operations in parallel while conflict arbitration is carried out on concurrent tasks. Otherwise, a permission insufficient prompt message is directly output to the terminal. S6: After feeding back and aggregating the execution results, output dual feedback to the terminal through voice broadcast and interface prompts, while recording the execution log.

[0006] By adopting the above technical solutions, the traditional multi-step manual operation is simplified into a cross-module composite scheduling through voice commands, avoiding the tediousness and delay of manual operation; the streaming recognition and parallel execution mechanism improves the real-time performance of scheduling, and the dual feedback and log recording ensure that the scheduling process is perceptible and traceable, solving the problems of long operation chains and delayed feedback in existing technologies.

[0007] Optionally, the sorting rule of the priority queue mechanism is: Basic priorities are set based on task type; video scheduling tasks have the highest basic priority, followed by GIS positioning and group management tasks, and ordinary voice intercom tasks have the lowest basic priority. An urgency level weighting coefficient is introduced; the task weighting coefficient in emergency mode is greater than that in non-urgent mode. The final priority is the product of the basic priority and the task weight coefficient. System resources are allocated from high to low according to the final priority.

[0008] Optionally, the geofencing algorithm of the GIS positioning module includes circular fence algorithm, rectangular fence algorithm and polygon fence algorithm.

[0009] Optionally, the RBAC scene lock model includes a role layer; The role layer categorizes user roles into ordinary members, group administrators, and system schedulers. The default permissions for regular members are voice calls and receiving dispatch instructions; The default permissions for the group administrator are voice calls, creating temporary groups, and single-module scheduling. The system scheduler's default permissions are full module scheduling, permission configuration, and scene lock triggering.

[0010] Optionally, the RBAC scene lock model includes a scene lock layer; The triggering conditions for the scene lock layer include the terminal reporting an emergency event and the platform receiving an emergency dispatch instruction. After the scene lock is triggered, ordinary members temporarily gain video scheduling and GIS location filtering permissions; The conditions for reclaiming the scene lock include the end of the event reported by the terminal, the administrator manually turning off the emergency mode, and no operation within the time limit.

[0011] Optionally, the logic of the dual feedback is as follows: TTS voice broadcast and interface prompts are triggered synchronously. The voice broadcast content is a concise description of the execution result, and the interface prompts include operation type icons, execution status text, and related data display. The execution log includes instruction content, execution time, execution module, result status, user information, and permission verification records. The log data is synchronized to the platform database for archiving via the MQTT channel.

[0012] Optionally, the parallel execution of operations by each functional module in step S5 can be achieved as follows: Each functional module is assigned an independent asynchronous execution thread, and each thread uses a callback function to report the execution progress and results to the engine. Set an execution status monitoring cycle to obtain the execution status of each module in real time. If a module times out, a retry mechanism will be triggered. If the retry fails after 2 attempts, the module task will be terminated and an exception message will be reported.

[0013] Optionally, the conflict arbitration rule in step S5 is: When a high-priority task and a low-priority task compete for the same system resources, a resource pre-emption mechanism is adopted. The low-priority task is suspended and the occupied resources are released. The low-priority task is resumed after the high-priority task is completed. When resource conflicts occur between tasks of the same priority, a time-slice round-robin mechanism is adopted to allocate resources sequentially according to the order in which tasks are received. The time slice occupied by each task for a single resource is a preset value.

[0014] Secondly, this application provides a remote dispatch system for public network intercom terminals based on AI voice interaction, adopting the following technical solution: A remote dispatch system for public network intercom terminals based on AI voice interaction includes: The terminal is used to collect user-input voice commands, preprocess the voice commands, and output a standard format voice data stream. The AI ​​speech recognition and parsing unit is loaded with a pre-trained end-to-end Transformer model, which is used to perform streaming recognition and intent parsing on the speech data stream and output structured instructions containing operation intent and slot information. An instruction mapper and classifier is used to map the structured instructions to corresponding functional module APIs according to a preset extensible instruction list. The functional modules include a video dispatch module, a GIS positioning module, and a group management module; The dynamic permission verification unit verifies the user role and current scene status based on the RBAC scene lock model to determine whether the user has the operation permission to execute the API. A multi-module scheduling engine is used to distribute the API to the corresponding functional modules using a priority queue mechanism when the permission verification is passed, control the parallel execution of operations by each functional module, and arbitrate conflicts between concurrent tasks. Each functional module feeds back the execution results to the multi-module scheduling engine. After aggregating the execution results, the multi-module scheduling engine outputs dual feedback to the terminal through the feedback module, including voice broadcast and interface prompts, while also recording the execution log. The feedback module is used to directly output a message indicating insufficient permissions to the terminal when the permission verification fails.

[0015] In summary, this application includes at least the following beneficial technical effects: This application reconstructs the remote dispatching logic of public network intercom terminals using AI voice interaction technology. It eliminates the need for complex manual operations, enabling single-module or cross-module composite dispatching solely through natural language commands. This significantly simplifies the operation process, lowers the dispatching threshold, and allows users at different levels to quickly initiate dispatching tasks. The end-to-end Transformer model's scenario-based training and streaming recognition mechanism ensure the accuracy and real-time performance of voice command parsing, providing core support for rapid dispatching. Priority queues and conflict arbitration strategies optimize system resource allocation, ensuring high-priority tasks are executed first and avoiding dispatching blockages caused by resource conflicts. The RBAC scene lock model, through role-based hierarchical and dynamic scene adaptation, balances system security and scene flexibility, overcoming the limitations of fixed permissions. A multi-module parallel execution mechanism shortens the time required for composite dispatching, while dual feedback and logging make the dispatching process perceptible and traceable, improving system controllability and maintainability. Furthermore, the GIS positioning module's various geofencing algorithms and the video dispatching module's multi-screen display function further expand scene adaptability, and the system's modular architecture facilitates future functional expansion and maintenance. Overall, this invention effectively solves the problems of cumbersome operation, low command coupling, high feedback delay, and rigid permissions in existing technologies, significantly improving the overall scheduling efficiency and practical value of public network intercom systems, and is suitable for various scheduling scenarios that require rapid collaboration. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the remote scheduling method for public network intercom terminals based on AI voice interaction in this application. Detailed Implementation

[0017] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0018] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0019] This application discloses a remote dispatching method for public network intercom terminals based on AI voice interaction, referring to... Figure 1 This includes the following steps: S1: The terminal collects the user's voice commands, preprocesses the voice commands, and outputs a standard format voice data stream. S2: Perform streaming recognition and intent parsing on the voice data stream, and output a structured instruction containing operation intent and slot information; S3: Based on a preset extensible instruction list, map the structured instructions to corresponding functional module APIs; the functional modules include a video scheduling module, a GIS positioning module, and a group management module; S4: Verify the user role and current scene state using the RBAC scene lock model to determine whether the user has the permission to execute the operation corresponding to the API; S5: If the permission verification passes, the API is distributed to the corresponding functional modules using a priority queue mechanism. The functional modules are controlled to perform operations in parallel while conflict arbitration is carried out on concurrent tasks. Otherwise, a permission insufficient prompt message is directly output to the terminal. S6: After feeding back and aggregating the execution results, output dual feedback to the terminal through voice broadcast and interface prompts, while recording the execution log.

[0020] A remote dispatch system for public network intercom terminals based on AI voice interaction includes: The terminal is used to collect user-input voice commands, preprocess the voice commands, and output a standard format voice data stream. The AI ​​speech recognition and parsing unit is loaded with a pre-trained end-to-end Transformer model, which is used to perform streaming recognition and intent parsing on the speech data stream and output structured instructions containing operation intent and slot information. An instruction mapper and classifier is used to map the structured instructions to corresponding functional module APIs according to a preset extensible instruction list. The functional modules include a video dispatch module, a GIS positioning module, and a group management module; The dynamic permission verification unit verifies the user role and current scene status based on the RBAC scene lock model to determine whether the user has the operation permission to execute the API. A multi-module scheduling engine is used to distribute the API to the corresponding functional modules using a priority queue mechanism when the permission verification is passed, control the parallel execution of operations by each functional module, and arbitrate conflicts between concurrent tasks. Each functional module feeds back the execution results to the multi-module scheduling engine. After aggregating the execution results, the multi-module scheduling engine outputs dual feedback to the terminal through the feedback module, including voice broadcast and interface prompts, while also recording the execution log. The feedback module is used to directly output a message indicating insufficient permissions to the terminal when the permission verification fails.

[0021] The end-to-end Transformer model is the core support for the method and system, and its key design is as follows: Model structure: 6-layer encoder (including 8-head self-attention mechanism, 2048-dimensional hidden layer feedforward neural network, layer normalization and residual connection), 3-layer decoder (including 8-head self-attention mechanism, 8-head cross-attention mechanism, layer normalization and residual connection). Training steps: ① Collect a public network intercom scenario voice dataset (including samples with different accents and noisy environments, labeled with intent and slot labels); ② The preprocessed data is converted to a 16kHz / 16bit PCM stream and divided into training / validation / test sets in a 7:2:1 ratio; ③ Initialize parameters (learning rate 1e-4, batch size 32, AdamW optimizer, cross-entropy loss function); ④ Iterative training continues until the intent recognition accuracy on the validation set is ≥95% and the slot filling accuracy is ≥93%. ⑤ The test set verifies real-time performance and accuracy; Adaptation to various scenarios: Scenario-based training ensures that the model can accurately parse natural language commands in the public network intercom field, and the streaming recognition mechanism meets real-time scheduling requirements.

[0022] The following describes the implementation process of the present invention in detail with reference to two typical embodiments, so as to ensure that those skilled in the art can fully reproduce it based on the description.

[0023] Example 1: AI voice command triggers single-module video scheduling.

[0024] The user (ordinary member) presses the terminal AI key and says the voice command "Pull up Li Si's video". The built-in microphone of the terminal collects the voice signal. First, spectral subtraction noise reduction is used (to remove environmental wind sounds and background human voices), and then VAD cutting (the trigger threshold is that the voice amplitude ≥ 3 times the background noise) is used to remove the silent segments, and a 16kHz / 16bit standard PCM voice data stream is output.

[0025] The terminal uploads the PCM data stream to the platform side. The AI voice recognition and parsing unit loads the pre-trained end-to-end Transformer model, and processes the voice data in real time through streaming recognition. It can start parsing without waiting for the complete instruction input, and finally outputs a structured instruction: operation intention = video scheduling, slot information = target user = Li Si.

[0026] The instruction mapping and classifier query the preset extensible instruction table, map the above structured instruction to the video scheduling module API: / api / video / call?target=Li Si, and carry the user role identifier.

[0027] The dynamic permission verification unit calls the RBAC scenario lock model: ① Identify that the user role is "ordinary member", and the default permission does not include video scheduling permission; ② Query that the current scenario status is "non-emergency mode" and the scenario lock is not triggered; ③ Determine that the permission verification fails.

[0028] The feedback module generates a permission insufficient prompt and pushes it to the terminal through the MQTT channel: The terminal speaker executes TTS voice broadcast "Insufficient permission to initiate video scheduling", and the display screen shows the text prompt "Insufficient permission" and a red prohibition icon; At the same time, record the execution log (instruction content: Pull up Li Si's video; execution time: xxxx-xx-xxxx:xx:xx; execution module: video scheduling module; result status: permission verification failed; user information: ordinary member; permission verification record: non-emergency mode, default no video scheduling permission), and synchronize the log to the platform database.

[0029] If the user (ordinary member) initiates the same instruction in an emergency scenario: ① The terminal reports an emergency event and triggers the scenario lock; ② The dynamic permission verification unit confirms that the scenario lock has been triggered, and the ordinary member temporarily obtains the video scheduling permission, and the permission verification passes; ③ The multi-module scheduling engine distributes the API to the video scheduling module and allocates an independent asynchronous thread; ④ If the terminal is currently conducting a regular voice intercom (low priority task), according to the conflict arbitration rules, the voice intercom task is paused and the bitstream resources are released, and the video scheduling task (high priority) is executed first. ⑤ The video scheduling module sends a WebRTC streaming command to Li Si's terminal, with the bitrate priority set to the highest. ⑥ After Li Si's terminal responds, the video stream is transmitted to the initiating terminal, the feedback module triggers a voice broadcast "Video connected", and the display shows 4 split screens (including Li Si's video stream). ⑦ After the event ends, the terminal reports the end of the event, the scene lock is reclaimed, and the video scheduling permissions of ordinary members automatically expire.

[0030] Example 2: AI voice command triggers multi-module composite scheduling.

[0031] The user (system dispatcher) inputs the voice command "With logistics park area A as the center, pull online terminals within 500 meters into a temporary group and start 4-way split-screen video". After noise reduction and VAD segmentation, the terminal outputs a standard PCM voice data stream.

[0032] The AI ​​voice recognition and parsing unit parses the structured instructions: Operation intention = composite scheduling, slot information = center point = logistics park area A, radius = 500 meters, group type = temporary, number of split screens = 4.

[0033] The instruction mapper and classifier map it into three functional module APIs: ①GIS positioning API ( / api / GIS / filter?center=Logistics Park A Area&radius=500&type=circle); ② Group Management API ( / api / group / create?type=temporary); ③ Video scheduling API ( / api / video / call?split=4).

[0034] The dynamic permission verification unit identifies the user role as "system scheduler," who by default has full module scheduling permissions, and the permission verification passes.

[0035] Priority allocation: The multi-module scheduling engine sets the basic priority of tasks (video scheduling API is the highest, followed by GIS positioning API and group management API). The current mode is non-urgent, with a weight coefficient of 1.0. The final priority is sorted according to the basic priority.

[0036] Parallel execution: Assign independent asynchronous execution threads to the three APIs, and the threads report progress through callback functions; set the execution status monitoring period to 100ms and the timeout threshold to 1500ms.

[0037] ①GIS positioning module: The circular fence algorithm is adopted. The distance d between the center point of area A of the logistics park (Lon0=120.12°, Lat0=30.35°) and the latitude and longitude of each online terminal is calculated by the Haversine formula. Four terminals (ID: T1, T2, T3, T4) with d≤500 meters are selected and the selection results are fed back through a callback function. ② Group Management Module: Create a temporary group, invite T1-T4 to join, set the permission mask to "only administrators can invite new members", and provide feedback that the group was created successfully; ③ Video scheduling module: Sends WebRTC streaming commands concurrently to T1-T4, configuring 4-way split-screen display.

[0038] Conflict arbitration: If the GIS positioning module and the group management module simultaneously request the database to query the terminal status (with the same priority), a time-slice round-robin mechanism is adopted, with a single time slice = 50ms. Resources are allocated sequentially according to the API reception order to ensure that both execute normally.

[0039] The multi-module scheduling engine aggregates the execution results (GIS filters 4 terminals, group creation successful, all 4 video feeds connected), and the feedback module generates a voice broadcast: "4 online terminals within 500 meters of Area A of the logistics park have been added to the temporary group, and 4 split-screen videos have been enabled." The terminal displays show the same message. ① Geographic fence boundary map (marking the center point and a 500-meter circle); ② Group member list (T1-T4); ③ 4-channel real-time video stream.

[0040] At the same time, execution logs are recorded and synchronized to the platform database via the MQTT channel. The logs contain complete instruction information, execution process and result data, which facilitates subsequent traceability.

[0041] This application reconstructs the remote dispatching logic of public network intercom terminals using AI voice interaction technology. It eliminates the need for complex manual operations, enabling single-module or cross-module composite dispatching solely through natural language commands. This significantly simplifies the operation process, lowers the dispatching threshold, and allows users at different levels to quickly initiate dispatching tasks. The end-to-end Transformer model's scenario-based training and streaming recognition mechanism ensure the accuracy and real-time performance of voice command parsing, providing core support for rapid dispatching. Priority queues and conflict arbitration strategies optimize system resource allocation, ensuring high-priority tasks are executed first and avoiding dispatching blockages caused by resource conflicts. The RBAC scene lock model, through role-based hierarchical and dynamic scene adaptation, balances system security and scene flexibility, overcoming the limitations of fixed permissions. A multi-module parallel execution mechanism shortens the time required for composite dispatching, while dual feedback and logging make the dispatching process perceptible and traceable, improving system controllability and maintainability. Furthermore, the GIS positioning module's various geofencing algorithms and the video dispatching module's multi-screen display function further expand scene adaptability, and the system's modular architecture facilitates future functional expansion and maintenance. Overall, this invention effectively solves the problems of cumbersome operation, low command coupling, high feedback delay, and rigid permissions in existing technologies, significantly improving the overall scheduling efficiency and practical value of public network intercom systems, and is suitable for various scheduling scenarios that require rapid collaboration.

[0042] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A remote dispatching method for public network intercom terminals based on AI voice interaction, characterized in that, Includes the following steps: S1: The terminal collects the user's voice commands, preprocesses the voice commands, and outputs a standard format voice data stream. S2: Perform streaming recognition and intent parsing on the voice data stream, and output a structured instruction containing operation intent and slot information; S3: Based on a preset extensible instruction list, map the structured instructions to corresponding functional module APIs; the functional modules include a video scheduling module, a GIS positioning module, and a group management module; S4: Verify the user role and current scene state using the RBAC scene lock model to determine whether the user has the permission to execute the operation corresponding to the API; S5: If the permission verification passes, the API is distributed to the corresponding functional modules using a priority queue mechanism. The functional modules are controlled to perform operations in parallel while conflict arbitration is carried out on concurrent tasks. Otherwise, a permission insufficient prompt message is directly output to the terminal. S6: After feeding back and aggregating the execution results, output dual feedback to the terminal through voice broadcast and interface prompts, while recording the execution log.

2. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 1, characterized in that, The sorting rules of the priority queue mechanism are as follows: Basic priorities are set based on task type; video scheduling tasks have the highest basic priority, followed by GIS positioning and group management tasks, and ordinary voice intercom tasks have the lowest basic priority. An urgency level weighting coefficient is introduced; the task weighting coefficient in emergency mode is greater than that in non-urgent mode. The final priority is the product of the basic priority and the task weight coefficient. System resources are allocated from high to low according to the final priority.

3. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 1, characterized in that, The geofencing algorithms of the GIS positioning module include circular fence algorithm, rectangular fence algorithm and polygon fence algorithm.

4. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 1, characterized in that, The RBAC scene lock model includes a role layer; The role layer categorizes user roles into ordinary members, group administrators, and system schedulers. The default permissions for regular members are voice calls and receiving dispatch instructions; The default permissions for the group administrator are voice calls, creating temporary groups, and single-module scheduling. The system scheduler's default permissions are full module scheduling, permission configuration, and scene lock triggering.

5. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 4, characterized in that, The RBAC scene lock model includes a scene lock layer; The triggering conditions for the scene lock layer include the terminal reporting an emergency event and the platform receiving an emergency dispatch instruction. After the scene lock is triggered, ordinary members temporarily gain video scheduling and GIS location filtering permissions; The conditions for reclaiming the scene lock include the end of the event reported by the terminal, the administrator manually turning off the emergency mode, and no operation within the time limit.

6. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 1, characterized in that, The logic of the dual feedback is as follows: TTS voice broadcast and interface prompts are triggered synchronously. The voice broadcast content is a concise description of the execution result, and the interface prompts include operation type icons, execution status text, and related data display. The execution log includes instruction content, execution time, execution module, result status, user information, and permission verification records. The log data is synchronized to the platform database for archiving via the MQTT channel.

7. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 1, characterized in that, The parallel execution of operations by each functional module in step S5 is implemented as follows: Each functional module is assigned an independent asynchronous execution thread, and each thread uses a callback function to report the execution progress and results to the engine. Set an execution status monitoring cycle to obtain the execution status of each module in real time. If a module times out, a retry mechanism will be triggered. If the retry fails after 2 attempts, the module task will be terminated and an exception message will be reported.

8. The remote dispatching method for public network intercom terminals based on AI voice interaction according to claim 1, characterized in that, The conflict arbitration rules in step S5 are as follows: When a high-priority task and a low-priority task compete for the same system resources, a resource pre-emption mechanism is adopted. The low-priority task is suspended and the occupied resources are released. The low-priority task is resumed after the high-priority task is completed. When resource conflicts occur between tasks of the same priority, a time-slice round-robin mechanism is adopted to allocate resources sequentially according to the order in which tasks are received. The time slice occupied by each task for a single resource is a preset value.

9. A remote dispatch system for public network intercom terminals based on AI voice interaction, characterized in that: include: The terminal is used to collect user-input voice commands, preprocess the voice commands, and output a standard format voice data stream. The AI ​​speech recognition and parsing unit is loaded with a pre-trained end-to-end Transformer model, which is used to perform streaming recognition and intent parsing on the speech data stream and output structured instructions containing operation intent and slot information. An instruction mapper and classifier is used to map the structured instructions to corresponding functional module APIs according to a preset extensible instruction list. The functional modules include a video dispatch module, a GIS positioning module, and a group management module; The dynamic permission verification unit verifies the user role and current scene status based on the RBAC scene lock model to determine whether the user has the operation permission to execute the API. A multi-module scheduling engine is used to distribute the API to the corresponding functional modules using a priority queue mechanism when the permission verification is passed, control the parallel execution of operations by each functional module, and arbitrate conflicts between concurrent tasks. Each functional module feeds back the execution results to the multi-module scheduling engine. After aggregating the execution results, the multi-module scheduling engine outputs dual feedback to the terminal through the feedback module, including voice broadcast and interface prompts, while also recording the execution log. The feedback module is used to directly output a message indicating insufficient permissions to the terminal when the permission verification fails.

Images