A noise intelligent processing system based on MCP protocol and large language model
By integrating an MCP Server and a cloud-based LLM intelligent orchestration center into the noise monitoring terminal, the noise intelligent processing system solves the problems of data silos, terminal intelligence, and system rigidity in the noise map system, achieving efficient, real-time updates and accuracy of the noise map, and reducing operation and maintenance costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN SANYUAN ENVIRONMENTAL GOVERNANCE CO LTD
- Filing Date
- 2025-11-04
- Publication Date
- 2026-07-14
AI Technical Summary
Existing noise map systems suffer from problems such as data silos, low terminal intelligence, system rigidity, and lagging noise map updates, resulting in low system efficiency, poor accuracy, poor scalability, and increased operation and maintenance costs.
A noise intelligent processing system is built using the MCP protocol and large language model (LLM). The perception layer terminal has a built-in lightweight MCP Server, and the edge-cloud collaboration layer deploys an LLM intelligent orchestration center to achieve data standardization, terminal intelligence and dynamic collaboration, and support flexible function upgrades.
It enables plug-and-play functionality for heterogeneous terminals, reduces invalid data transmission through terminal preprocessing, improves data quality and processing efficiency through cloud-based intelligent scheduling, allows the system to flexibly respond to complex needs, shortens update cycles, and enhances the accuracy and real-time performance of noise maps.
Smart Images
Figure CN121438841B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of noise processing technology, and in particular to a noise intelligent processing system based on the MCP protocol and a large language model. Background Technology
[0002] With the accelerating pace of urbanization, environmental noise pollution has become a key issue affecting residents' quality of life and health. Noise pollution not only disrupts daily life but can also lead to health risks such as sleep disorders and cardiovascular diseases. To effectively manage and control noise, the industry commonly employs the construction of "noise mapping" systems. These systems collect real-time noise data by deploying noise monitoring micro-stations and sub-stations throughout the city. Combined with auxiliary data such as Geographic Information Systems (GIS), traffic flow data, and urban planning information, the systems use algorithmic models to simulate and visualize regional noise distribution, thereby providing decision support for environmental management.
[0003] However, existing noise mapping systems have revealed several significant shortcomings in practical applications, which limit the system's efficiency, accuracy, and scalability. The following is a detailed analysis of existing technical solutions and an objective description of their deficiencies:
[0004] 1. Severe Data Silos and Heterogeneity Issues: The existing system relies on multiple heterogeneous data sources, including noise monitoring terminals from different manufacturers, meteorological data, and traffic flow data. These terminal devices differ significantly in data format, communication protocols, and sampling frequencies, leading to difficulties in data integration. For example, some terminals use custom binary protocols, while others may use JSON or XML formats, lacking a unified data access standard. This heterogeneity makes it difficult for the system to achieve seamless data fusion, forming "data silos" that affect the overall accuracy and consistency of the noise map.
[0005] 2. Low level of terminal intelligence and weak data preprocessing capabilities: Traditional noise monitoring terminals have limited functionality, typically only responsible for collecting and uploading raw data, lacking local intelligent processing capabilities. The terminals cannot perform preliminary cleaning, filtering, or analysis of the collected data, resulting in a large amount of invalid data (such as background noise and outliers) being directly transmitted to the cloud. This not only consumes valuable communication bandwidth but also increases the complexity of cloud data cleaning. For example, during peak hours, the terminal may upload a large amount of redundant data, requiring the cloud server to consume significant computing resources for noise reduction and verification, thus reducing the overall system efficiency.
[0006] 3. Rigid System and Difficult Function Expansion: The data processing workflow of existing systems (such as data verification, noise event identification, and spectrum analysis) is usually fixed and cannot flexibly adapt to new monitoring needs. When new functions need to be added (such as identifying specific construction noise or connecting new sensors), a full-chain upgrade from terminal firmware to cloud services is required, resulting in long development cycles and high costs. This rigid architecture makes it difficult for the system to respond quickly to policy changes or emerging noise sources (such as electric vehicle noise), limiting its practicality in dynamic environments.
[0007] 4. Delayed Noise Map Updates and Insufficient Accuracy: Due to long data processing chains and low efficiency in collaboration between terminals and the cloud, noise map updates often suffer from significant delays. Existing systems typically employ batch processing, where data collection and map generation can take hours or even days, failing to meet real-time monitoring requirements. Furthermore, the lack of intelligent preprocessing results in low-quality data upon which map generation models rely, leading to biases in noise distribution simulations. For example, in complex urban environments, mixed background noise and sudden noise may not be effectively separated, impacting map accuracy and guidance value.
[0008] In summary, existing noise mapping systems have significant shortcomings in data integration, terminal intelligence, system flexibility, and real-time performance. These deficiencies not only reduce the effectiveness of noise control but also increase operation and maintenance costs. Therefore, there is an urgent need for a new generation of noise processing systems capable of standardized access, intelligent processing, and dynamic collaboration to improve the accuracy, real-time performance, and scalability of noise maps. This invention is an innovative solution proposed to address the aforementioned problems. Summary of the Invention
[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide a noise intelligent processing system based on the MCP protocol and a large language model.
[0010] The objective of this invention is achieved through the following technical solution: a noise intelligent processing system based on the MCP protocol and a large language model, comprising a perception layer, wherein the perception layer is connected to an edge-cloud collaboration layer, and the edge-cloud collaboration layer is connected to an application layer;
[0011] The perception layer includes multiple noise monitoring terminals. Each noise monitoring terminal has a built-in lightweight MCP Server, which is used to encapsulate the various functions of the noise monitoring terminal into terminal tools that conform to the MCP standard and register them with the cloud.
[0012] The edge-cloud collaboration layer includes an LLM intelligent orchestration center deployed in the cloud, which receives user task instructions, dynamically parses tasks and generates execution plans, schedules terminal tools through the MCP protocol, and coordinates multiple noise monitoring terminals to perform data acquisition and preprocessing.
[0013] The application layer includes a noise map generation model, which receives standardized data processed by the LLM orchestration layer, generates a noise map, and performs multi-dimensional statistical analysis.
[0014] Preferably, the noise monitoring terminal includes a noise monitoring micro-station, a noise monitoring sub-station, a mobile monitoring device, and other monitoring terminals.
[0015] Preferably, the LLM intelligent orchestration center sends instructions to the noise monitoring terminal via the MCP protocol to control the noise monitoring terminal to adjust the sampling frequency, execute the noise separation algorithm, or enable collaborative diagnosis.
[0016] When the system needs to perform macro-level tasks, the LLM intelligent orchestration center performs dynamic planning, decomposing the macro-level tasks into multiple sub-tasks. Then, according to the task requirements of the sub-tasks, the LLM intelligent orchestration center sends instructions to the MCP Server of the noise monitoring terminal in the target area through the MCP protocol, while performing dynamic optimization. When the reported data of a noise monitoring terminal is abnormal, the LLM intelligent orchestration center enables the neighboring terminal to perform data compensation or sends a repair instruction to trigger the terminal self-test tool so that the abnormal terminal can perform self-diagnosis.
[0017] Preferably, the multi-dimensional statistical analysis includes decibel level statistical analysis, noise type proportion statistical analysis, and spatiotemporal change trend statistical analysis.
[0018] Preferably, the noise monitoring terminal has the following functions: noise data acquisition, spectrum analysis, equipment status self-test, and geographic location reporting.
[0019] Preferably, when a new noise recognition algorithm or sensor is added, its function is encapsulated into a new terminal tool and deployed to the noise monitoring terminal. The LLM intelligent orchestration center automatically discovers and calls the new terminal tool, realizing automatic upgrade of system functions.
[0020] The beneficial effects of this invention are:
[0021] 1) By embedding a server conforming to the MCP (Model Context Protocol) standard into each noise monitoring terminal, heterogeneous terminal capabilities (data acquisition and analysis tools) are encapsulated into a unified tool interface. This enables "plug-and-play" capabilities for terminal devices from different manufacturers and models, eliminating the need for the cloud system to develop specific adapters for each terminal and thus completely eliminating data barriers caused by differences in protocols and data formats. Based on a unified MCP data framework, the system can not only integrate noise data reported by various terminals but also naturally incorporate other auxiliary data sources such as geographic information, traffic flow, and meteorological data. When planning tasks, the LLM intelligent orchestration center can proactively call upon and associate these multimodal data, providing richer and more consistent contextual information for noise map generation and data analysis, significantly improving the utilization value of data resources and the comprehensiveness of decision support.
[0022] 2) The terminal transforms from a simple "data collector" into an "intelligent node" with preliminary analytical capabilities. By performing preprocessing tasks such as noise type separation, spectral feature extraction, and anomaly detection locally on the terminal, the computational burden of a large amount of raw audio data is absorbed at the edge. This directly reduces the amount of useless data uploaded by more than 70%, greatly saves communication bandwidth, and reduces the pressure on data cleaning and storage in the cloud; the large language model (LLM) in the cloud acts as the "intelligent brain" of the system. It can understand the user's natural language commands (such as "generate a traffic noise map of a certain area") and dynamically decompose tasks and intelligently schedule the most suitable terminal toolchain for execution. This task planning capability based on semantic understanding enables the system to flexibly respond to complex and ever-changing actual needs, realizing a leap from "pre-set processes" to "on-demand process generation," and a qualitative leap in the level of intelligence.
[0023] 3) When new analytical functions need to be added (such as identifying a new type of industrial noise) or new sensors need to be connected, only the corresponding MCP tool needs to be developed for that function and registered on the terminal or in the cloud. LLM can automatically discover and call new tools without requiring disruptive modifications to the main system architecture or a full-chain upgrade, greatly shortening the function launch cycle and reducing development and maintenance costs; the system's dynamic optimization mechanism makes it highly robust. When a terminal malfunctions or experiences data anomalies, LLM can respond quickly, deciding to activate nearby terminals for data compensation or instructing the faulty terminal to perform self-diagnosis, ensuring the continuous and effective operation of the entire monitoring network and improving the system's reliability and availability. Attached Figure Description
[0024] Figure 1 This is a system architecture diagram of the present invention. Detailed Implementation
[0025] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0026] See Figure 1 The present invention provides a technical solution: a noise intelligent processing system based on the MCP protocol and a large language model, including a perception layer, wherein the perception layer is connected to an edge-cloud collaboration layer, and the edge-cloud collaboration layer is connected to an application layer;
[0027] The perception layer includes multiple noise monitoring terminals. Each noise monitoring terminal has a built-in lightweight MCP Server, which is used to encapsulate the various functions of the noise monitoring terminal into terminal tools that conform to the MCP standard and register them with the cloud.
[0028] The edge-cloud collaboration layer includes an LLM intelligent orchestration center deployed in the cloud, which receives user task instructions, dynamically parses tasks and generates execution plans, schedules terminal tools through the MCP protocol, and coordinates multiple noise monitoring terminals to perform data acquisition and preprocessing.
[0029] The application layer includes a noise map generation model, which receives standardized data processed by the LLM orchestration layer, generates a noise map, and performs multi-dimensional statistical analysis.
[0030] In this embodiment, the present invention embeds the MCP Server service into terminal devices such as noise monitoring micro-stations and sub-stations, making them intelligent nodes in the system. This moves the application of the MCP protocol from a pure software service to the physical noise monitoring terminal, forming an end-to-cloud integrated intelligent architecture of "terminal MCP service-oriented architecture." Utilizing LLM as a global intelligent scheduler, various MCP tools distributed across the terminal and cloud are dynamically coordinated to complete complex noise data processing and map generation tasks, achieving a leap from "static process" to "dynamic intelligent orchestration."
[0031] Standardization and modularization of terminal data were achieved: By embedding an MCP Server in each terminal, all terminals interact using a unified "language," fundamentally solving the problems of data heterogeneity and protocol inconsistencies, laying a solid foundation for system collaboration. Data quality and transmission efficiency were improved: With preprocessing capabilities, terminals can filter invalid data and extract key features locally, significantly reducing the amount of data uploaded, saving bandwidth, and providing cleaner, higher-value data to the cloud. System flexibility and scalability were enhanced: When a new noise recognition algorithm or sensor is added, only a corresponding MCP tool needs to be developed and deployed to the terminal. LLM can automatically discover and call the new tool, achieving "hot-swappable" functional upgrades for the system. The accuracy and real-time performance of noise maps were significantly improved: Because LLM can intelligently schedule terminals for collaborative sampling and edge computing, the acquired data is more representative and of higher quality. At the same time, the streamlined data processing chain greatly shortens the cycle from data acquisition to map generation, enabling near real-time or on-demand rapid updates of noise maps.
[0032] Thanks to edge-side data preprocessing and the efficient task orchestration of LLM, the entire processing chain from data acquisition to map generation is significantly shortened. The system can achieve near real-time data processing and map updates (e.g., minute-level location tracking in the event of sudden noise incidents), completely overcoming the shortcomings of traditional systems' lag in updates and providing timely data for environmental monitoring and emergency response. Because the data transmitted to the cloud is high-quality, high-value feature data (such as "separated traffic noise decibel values") after intelligent preprocessing at the terminal, rather than the raw data containing a large amount of interference, the data quality relied upon by the cloud-based model for generating the noise map is fundamentally improved. Combined with the collaborative fusion of multi-terminal data and advanced interpolation algorithms, the accuracy of the final noise map in noise level assessment and noise source identification is substantially improved.
[0033] The "MCP+LLM" open collaborative architecture constructed in this invention is not only applicable to urban environmental noise monitoring, but its technical concept and implementation method can also be extended to other environmental monitoring fields (such as air quality monitoring, water quality monitoring, etc.). This architecture provides a general and reusable solution for realizing intelligent management and control of large-scale, heterogeneous, distributed environmental IoT, and has significant industry promotion value and forward-looking significance.
[0034] In some embodiments, the noise monitoring terminal includes a noise monitoring micro-station, a noise monitoring sub-station, a mobile monitoring device, and other monitoring terminals.
[0035] In this embodiment, the perception layer (terminal MCP service-oriented architecture) integrates a lightweight MCP Server within each noise monitoring micro-station, sub-station, and other terminal device. This server encapsulates the terminal's basic capabilities (such as noise data acquisition, spectrum analysis, device status self-checking, and geographic location reporting) into tools compliant with the MCP standard. The terminal is no longer a "dumb" data collector but an intelligent node with standardized self-description capabilities and preliminary data processing capabilities.
[0036] In some embodiments, the LLM intelligent orchestration center sends instructions to the noise monitoring terminal via the MCP protocol to control the noise monitoring terminal to adjust the sampling frequency, execute the noise separation algorithm, or enable collaborative diagnosis.
[0037] When the system needs to perform macro-level tasks, the LLM intelligent orchestration center performs dynamic planning, decomposing the macro-level tasks into multiple sub-tasks. Then, according to the task requirements of the sub-tasks, the LLM intelligent orchestration center sends instructions to the MCP Server of the noise monitoring terminal in the target area through the MCP protocol, while performing dynamic optimization. When the reported data of a noise monitoring terminal is abnormal, the LLM intelligent orchestration center enables the neighboring terminal to perform data compensation or sends a repair instruction to trigger the terminal self-test tool so that the abnormal terminal can perform self-diagnosis.
[0038] In this embodiment, the edge-cloud collaboration layer (LLM intelligent orchestration) involves deploying an LLM intelligent orchestration center in the cloud. This center possesses the capability descriptions of all registered terminal MCP Servers and their tools.
[0039] When the system needs to perform a task (such as "generating a daytime noise map of the city center area today"), LLM will perform dynamic programming: Step 1: Task decomposition. LLM decomposes the macro task into a series of sub-tasks, such as: [Get a list of terminals in a specified area] -> [Schedule terminals to perform high-quality data sampling] -> [Collect and merge data] -> [Execute noise interpolation algorithm] -> [Generate a visualization map].
[0040] Step Two: Intelligent Scheduling. LLM sends instructions directly to the MCP Server of the relevant terminals via the MCP protocol. For example, it can command terminals to increase the sampling rate during specific time periods or execute a real-time "traffic noise and background noise separation" algorithm, then upload the processed feature data instead of the original audio stream.
[0041] Step 3: Dynamic Optimization. If a terminal reports abnormal data, the LLM can decide to enable nearby terminals to perform data compensation or issue repair commands to enable the abnormal terminal to perform self-diagnosis.
[0042] In some embodiments, the multi-dimensional statistical analysis includes decibel level statistical analysis, noise type proportion statistical analysis, and spatiotemporal variation trend statistical analysis.
[0043] In this embodiment, the application layer (noise map generation and statistics) receives high-quality, standardized data processed by the LLM orchestration layer, calls a specialized noise map generation model, and quickly produces a high-precision noise map. Simultaneously, based on standard data from the MCP protocol, the system can easily perform multi-dimensional statistical analysis, such as decibel level statistics, noise type proportions, and spatiotemporal variation trends.
[0044] In some embodiments, the noise monitoring terminal has the following functions: noise data acquisition, spectrum analysis, device status self-test, and geographic location reporting.
[0045] In some embodiments, when a new noise recognition algorithm or sensor is added, its functionality is encapsulated as a new terminal tool and deployed to the noise monitoring terminal. The LLM intelligent orchestration center automatically discovers and calls the new terminal tool, thereby achieving automatic system function upgrades.
[0046] Two specific examples are given below:
[0047] Example 1: Rapid Generation of High-Precision Traffic Noise Maps
[0048] Scenario: Environmental protection departments need to quickly update noise maps of major traffic arteries during the evening rush hour.
[0049] Implementation process (in conjunction with appendix) Figure 1 ):
[0050] User instruction: The user sends an instruction to the system: "Generate a traffic noise map of Zhongshan East Road for the current time period, with high accuracy requirements."
[0051] LLM Analysis and Planning: The LLM Intelligent Orchestration Center analyzes the instructions and discovers from the MCP tool registry that multiple micro-stations (MCP Servers) deployed along Zhongshan East Road provide tools such as "high-frequency sampling," "A-weighted sound pressure level calculation," and "traffic noise separation." LLM Generation Plan: [Schedule relevant terminals to enable "high-frequency sampling" and "traffic noise separation"] -> [Collect processed feature data] -> [Call the "Kriging interpolation" algorithm in the cloud] -> [Generate map].
[0052] Dynamic execution: The execution engine sends instructions to the target terminal group via the MCP protocol. The terminals perform complex audio processing locally, uploading only lightweight data such as the separated "traffic noise decibel value" and "geographical location", rather than the massive original audio.
[0053] Results: Data transmission volume is reduced by 70%, cloud processing speed is doubled, and the generated noise map, based on cleaner "traffic noise" data, has far greater accuracy and guidance value than traditional methods.
[0054] Example 2: Location and Source Tracing of Sudden Noise Events
[0055] Scenario: At night, residents complained of construction noise from an unknown source in a certain area.
[0056] Implementation process:
[0057] User command: "Locate and identify sudden noise sources around XX residential area."
[0058] LLM Analysis and Planning: The LLM first schedules all terminals around the complaint area to enter "event listening mode". When the "burst noise recognition" function of the MCP tool of a terminal (Terminal A) is triggered, the LLM immediately receives the event.
[0059] Dynamic coordination: LLM then instructs terminal A to activate the "sound source orientation" tool, and simultaneously instructs the other three surrounding terminals (B, C, D) to activate high-precision sampling and orientation functions.
[0060] Data fusion and decision-making: The execution engine gathers directional data from four terminals, calls the cloud-based "triangulation" MCP tool to quickly calculate the specific location of the noise source (such as a construction site), and combines it with the audio spectrum characteristics uploaded by terminal A. The LLM determines that it is "heavy machinery construction noise".
[0061] Results: The system completed the entire process from event detection to precise location and identification within minutes, providing law enforcement agencies with accurate targets for handling the incident and demonstrating the system's real-time response and intelligent collaboration capabilities.
[0062] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.
Claims
1. A noise intelligent processing system based on the MCP protocol and a large-scale language model, characterized in that: It includes a perception layer, which is connected to an edge-cloud collaboration layer, and the edge-cloud collaboration layer is connected to an application layer; The perception layer includes multiple noise monitoring terminals. Each noise monitoring terminal has a built-in lightweight MCP Server, which is used to encapsulate the various functions of the noise monitoring terminal into terminal tools that conform to the MCP standard and register them with the cloud. The edge-cloud collaboration layer includes an LLM intelligent orchestration center deployed in the cloud, which receives user task instructions, dynamically parses tasks and generates execution plans, schedules terminal tools through the MCP protocol, and coordinates multiple noise monitoring terminals to perform data acquisition and preprocessing. The application layer includes a noise map generation model, which receives standardized data processed by the LLM orchestration layer, generates a noise map, and performs multi-dimensional statistical analysis. The LLM intelligent orchestration center sends instructions to the noise monitoring terminal via the MCP protocol to control the noise monitoring terminal to adjust the sampling frequency, execute the noise separation algorithm, or enable collaborative diagnosis. When the system needs to perform macro-level tasks, the LLM intelligent orchestration center performs dynamic planning, decomposing the macro-level tasks into multiple sub-tasks. Then, according to the task requirements of the sub-tasks, the LLM intelligent orchestration center sends instructions to the MCP Server of the noise monitoring terminal in the target area through the MCP protocol, while performing dynamic optimization. When the reported data of a noise monitoring terminal is abnormal, the LLM intelligent orchestration center enables the neighboring terminal to perform data compensation or sends a repair instruction to trigger the terminal self-test tool so that the abnormal terminal can perform self-diagnosis.
2. The noise intelligent processing system based on MCP protocol and large language model according to claim 1, characterized in that: The noise monitoring terminals include noise monitoring micro-stations, noise monitoring sub-stations, mobile monitoring devices, and other monitoring terminals.
3. The noise intelligent processing system based on MCP protocol and large language model according to claim 1, characterized in that: The multi-dimensional statistical analysis includes decibel level statistical analysis, noise type proportion statistical analysis, and spatiotemporal variation trend statistical analysis.
4. The noise intelligent processing system based on MCP protocol and large language model according to claim 1, characterized in that: The noise monitoring terminal has the following functions: noise data acquisition, spectrum analysis, equipment status self-test, and geographic location reporting.
5. The noise intelligent processing system based on the MCP protocol and a large language model according to any one of claims 1-4, characterized in that: When a new noise recognition algorithm or sensor is added, its functionality is encapsulated into a new terminal tool and deployed to the noise monitoring terminal. The LLM intelligent orchestration center automatically discovers and calls the new terminal tool, thereby achieving automatic system function upgrades.