Dynamic evaluation and management system for acoustic environment of urban functional area

Abstract

The invention relates to the technical field of urban sound environment evaluation and management, in particular to an urban functional area sound environment dynamic evaluation and management system which comprises a data acquisition module, a data processing module, a multi-source data fusion unit, a dynamic evaluation module, a management decision module and a visual interaction module. The data acquisition module acquires sound environment data including sound pressure level, frequency spectrum, timestamp and geographic position in real time through a sound sensor; the data processing module preprocesses the data and extracts characteristic parameters; the multi-source data fusion unit integrates auxiliary data such as traffic flow and weather; the dynamic evaluation module adopts a deep learning dynamic adaptive model and combines online learning to generate an acoustic environment evaluation index; the management decision module is used for realizing graded early warning, noise source automatic control and strategy optimization; and the visual interaction module displays a result in real time and provides a configuration interface. According to the invention, dynamic accurate evaluation and efficient management of the acoustic environment are realized, and the reliability and practicability of the system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban sound environment evaluation and management, in particular to a kind of urban functional area sound environment dynamic evaluation and management system. BACKGROUND

[0002] In the traditional urban sound environment evaluation and management technology, the evaluation model is mostly static model, cannot combine real-time traffic flow, meteorological data, urban planning and other auxiliary data, and cannot adjust parameters through online learning, it is difficult to adapt to the dynamic changes of urban sound environment, the evaluation result is one-sided and the accuracy is insufficient.

[0003] And the early warning mechanism of management level is imperfect, lacks grading early warning capability, and the control of noise source mostly depends on manual operation, the response efficiency is low, and the management strategy lacks iterative optimization based on historical effect, it is difficult to realize efficient management.In addition, there is no intuitive visual interactive interface, and management personnel is difficult to master sound environment evaluation result, management state and prediction trend in real time, which is not conducive to rapid decision-making, and cannot meet the demand of dynamic and accurate evaluation and efficient management of urban functional area sound environment, therefore, aiming at the above problems, a kind of urban functional area sound environment dynamic evaluation and management system is provided. SUMMARY

[0004] The purpose of the present application is to provide a kind of urban functional area sound environment dynamic evaluation and management system to solve the problems proposed in the above background.

[0005] To achieve the above purpose, the present application provides the following technical scheme: A kind of urban functional area sound environment dynamic evaluation and management system, comprising: Data acquisition module, for real-time acquisition of sound environment data by deploying multiple sound sensors in urban functional area, sound environment data includes sound pressure level, frequency spectrum, time stamp and geographic location information; Data processing module, connected to data acquisition module, for pre-processing sound environment data, including data cleaning, denoising and normalization, and extracting sound environment characteristic parameters, sound environment characteristic parameters include equivalent continuous A sound level, maximum sound level, sound event duration and spectral characteristics; Multi-source data fusion unit, for fusing auxiliary data from external system, auxiliary data includes traffic flow data, meteorological data, urban planning data and real-time activity data; Dynamic evaluation module, connected to data processing module and multi-source data fusion unit, for dynamic evaluation based on sound environment characteristic parameters and auxiliary data, using pre-trained sound environment evaluation model, to generate sound environment evaluation index, the sound environment evaluation model adopts dynamic self-adaptive model based on deep learning, and adjusts model parameters through online learning algorithm according to real-time data; The management decision module is connected to the dynamic evaluation module, and is configured to perform dynamic management operations according to the sound environment evaluation index and a preset management strategy library. The management decision module comprises: The early warning unit is configured to generate graded early warning information when the sound environment evaluation index exceeds a preset threshold. The control unit is configured to automatically control the noise source equipment through the Internet of Things platform. The strategy optimization unit is configured to optimize the management strategy based on historical management effect data through a reinforcement learning algorithm. The visualization interaction module is connected to the management decision module and the dynamic evaluation module, and is configured to display the sound environment evaluation results, management operation states and prediction trends in real time, and provide a parameter configuration interface.

[0006] As a preferred solution, the data collection module is configured to collect sound environment data in real time through a plurality of sound sensors deployed in urban functional areas, including: According to the sound environment monitoring requirements of the urban functional area, the sampling parameters and network connection parameters of the sound sensor are configured; The initialized sound sensor continuously captures environmental sound waves and generates analog sound electrical signals; The analog sound electrical signals are converted into digital sound signals through an analog-to-digital converter; The digital sound signals are analyzed in time domain and frequency domain, and the sound pressure level and frequency spectrum are calculated; The calculated sound pressure level and frequency spectrum are added with time stamp and geographic location information to form structured sound environment data; The structured sound environment data is transmitted in real time to the data processing module through a communication interface.

[0007] As a preferred solution, the data processing module is configured to preprocess the sound environment data, including data cleaning, denoising and normalization, and extract sound environment feature parameters, including: The sound environment data from the data collection module is received; The sound environment data is cleaned to remove invalid data and outliers, obtaining cleaned sound environment data; The cleaned sound environment data is denoised to eliminate environmental noise interference, obtaining denoised sound environment data; The denoised sound environment data is normalized to standardize the data to a preset range, obtaining normalized sound environment data; Based on the normalized sound environment data, the sound environment feature parameters are extracted, including equivalent continuous A sound level, maximum sound level, sound event duration and spectral characteristics.

[0008] As a preferred solution, the multi-source data fusion unit is configured to fuse auxiliary data from external systems, including: receiving auxiliary data from external systems, including traffic flow data, meteorological data, urban planning data, and real-time activity data; data preprocessing of the received auxiliary data, including data cleaning and format standardization, to remove invalid data and unify data formats, to obtain preprocessed auxiliary data; spatiotemporal alignment of the preprocessed auxiliary data with the acoustic environment data from the data processing module, matching based on timestamps and geographic location information, to obtain an aligned data set; data fusion processing of the aligned data set, integrating acoustic environment data and auxiliary data, to generate multi-source fusion data; outputting the multi-source fusion data to the dynamic evaluation module for use.

[0009] As a preferred solution, the dynamic evaluation module is configured to perform dynamic evaluation based on acoustic environment feature parameters and auxiliary data using a pre-trained acoustic environment evaluation model to generate an acoustic environment evaluation index, the acoustic environment evaluation model being a deep learning-based dynamic adaptive model that adjusts model parameters through an online learning algorithm according to real-time data, including: receiving acoustic environment feature parameters from the data processing module and auxiliary data from the multi-source data fusion unit; feature vector construction processing of the received acoustic environment feature parameters and auxiliary data, integrating different types and dimensions of parameters into a unified multi-dimensional feature vector; inputting the constructed multi-dimensional feature vector into the pre-trained acoustic environment evaluation model, the acoustic environment evaluation model including a multi-layer neural network structure for simulating the complex nonlinear relationship between acoustic environment quality and multi-source features; forward propagation calculation of the input multi-dimensional feature vector through the multi-layer neural network structure, extracting high-level abstract features layer by layer; regression analysis processing based on the high-level abstract features of the neural network output layer to generate an initial acoustic environment evaluation index; real-time monitoring of the deviation of newly input multi-dimensional feature vectors from the distribution of model training data, triggering a model update mechanism when the deviation exceeds a preset tolerance; using an online learning algorithm, calculating the loss function between the model prediction output and the expected output using real-time samples composed of newly arrived acoustic environment feature parameters and auxiliary data; based on the loss function, dynamically adjusting the connection weights and bias parameters of the multi-layer neural network in the acoustic environment evaluation model using a gradient backpropagation algorithm; updating the acoustic environment evaluation model using the adjusted model parameters to generate an optimized model with environmental adaptive capability; The updated sound environment evaluation model is used to re-calculate the multi-dimensional feature vector input in real time, and an optimized sound environment evaluation index is generated. The optimized sound environment evaluation index and historical evaluation data are output to the management decision module, and the data generated during the evaluation process are stored in the model training database.

[0010] As a preferred solution, the management decision module is configured to execute dynamic management operations based on the sound environment evaluation index and a pre-configured management strategy library, including: receiving the sound environment evaluation index from the dynamic evaluation module; comparing the received sound environment evaluation index with the threshold conditions in the pre-configured management strategy library to determine the current sound environment state level; based on the determined sound environment state level, retrieving a corresponding management strategy set from the management strategy library; based on the retrieved management strategy set, generating a set of dynamic management operation instructions in combination with the real-time sound environment evaluation index; executing the set of dynamic management operation instructions, including: based on the warning conditions in the set of dynamic management operation instructions, generating hierarchical warning information through the warning unit and sending it to the relevant terminal; based on the control logic in the set of dynamic management operation instructions, automatically adjusting the operating parameters of the noise source equipment through the control unit via the Internet of Things platform; based on the historical management effect data and the current sound environment state level, iteratively optimizing and updating the strategies in the management strategy library through the strategy optimization unit using reinforcement learning algorithm.

[0011] As a preferred solution, the visual interaction module is configured to display the sound environment evaluation results, management operation status and prediction trends in real time, and provide a parameter configuration interface, including: receiving the sound environment evaluation index from the dynamic evaluation module and the management operation status data from the management decision module; performing visual data conversion processing on the received sound environment evaluation index and management operation status data to generate graphical representation elements; based on the graphical representation elements, constructing real-time sound environment evaluation charts and management operation status display panels; integrating the prediction trend data from the dynamic evaluation module to generate a sound environment quality prediction curve through time series analysis processing; combining the real-time sound environment evaluation charts, management operation status display panels and sound environment quality prediction curve in the interface to form a comprehensive visual interface; rendering the comprehensive visual interface to output dynamic visual content on the display device; The parameter configuration interface processing is provided, and the management policy parameters and display preference parameters input by a user are received; According to the management policy parameters input by the user, the preset management policy library in the management decision module is updated; According to the display preference parameters input by the user, the display style and layout of the comprehensive visualization interface are adjusted.

[0012] As can be seen from the technical solutions provided by the above-mentioned application, the urban functional area sound environment dynamic evaluation and management system provided by the application has the beneficial effects that: The data acquisition and processing link can obtain multi-dimensional sound environment data containing sound pressure level, frequency spectrum and space-time information in real time, and through cleaning, denoising, normalization processing and feature parameter extraction, the real-time, integrity and high quality of the data are ensured, thereby laying a reliable data foundation for sound environment evaluation; The multi-source data fusion unit integrates sound environment data and auxiliary data such as traffic flow, weather, urban planning and the like, avoids one-sidedness of single sound environment data evaluation, makes the evaluation result more in line with the actual scene of the city, and improves the overall evaluation; The dynamic evaluation module adopts a dynamic adaptive model based on deep learning, and combines an online learning algorithm to adjust parameters in real time, can accurately capture the complex relationship between sound environment quality and multiple factors, and is suitable for environmental changes, thereby effectively improving the evaluation accuracy and environmental adaptability; The management decision module timely delivers sound environment abnormal information through hierarchical early warning, relies on the Internet of Things to automatically control noise source equipment to quickly dispose problems, and then iteratively optimizes the management strategy through reinforcement learning, thereby realizing rapid response and efficient disposal of sound environment abnormalities, and long-term improving management efficiency and accuracy; The visualization interaction module directly presents the sound environment evaluation result, management state and prediction trend, and simultaneously provides a parameter configuration function, thereby reducing the data understanding and operation threshold of management personnel, assisting in quickly grasping the sound environment situation and flexibly adjusting management and display settings, and assisting in efficient decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 FIG. 1 is a schematic diagram of the overall structure of the urban functional area sound environment dynamic evaluation and management system according to the application. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solutions and advantages of the application more clear, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application, and are not used to limit the application.

[0015] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in combination with the drawings and specific embodiments of the specification.

[0016] As Figure 1 shown in the figure, the embodiment of the present application provides a city functional area sound environment dynamic evaluation and management system, comprising: a data acquisition module for real-time acquisition of sound environment data through a plurality of sound sensors deployed in the city functional area, the sound environment data including sound pressure level, frequency spectrum, timestamp and geographic location information; a data processing module connected to the data acquisition module for preprocessing the sound environment data, including data cleaning, denoising and normalization, and extracting sound environment feature parameters, the sound environment feature parameters including equivalent continuous A sound level, maximum sound level, sound event duration and spectral characteristics; a multi-source data fusion unit for fusing auxiliary data from external systems, the auxiliary data including traffic flow data, meteorological data, urban planning data and real-time activity data; a dynamic evaluation module connected to the data processing module and the multi-source data fusion unit for dynamic evaluation based on the sound environment feature parameters and the auxiliary data using a pre-trained sound environment evaluation model, generating a sound environment evaluation index, the sound environment evaluation model using a dynamic self-adaptive model based on deep learning, adjusting model parameters through an online learning algorithm according to real-time data; a management decision module connected to the dynamic evaluation module for executing dynamic management operations according to the sound environment evaluation index in combination with a pre-set management strategy library, the management decision module including: a warning unit for generating graded warning information when the sound environment evaluation index exceeds a pre-set threshold; a control unit for automatically controlling noise source equipment through an Internet of Things platform; a strategy optimization unit for optimizing management strategies based on historical management effect data through a reinforcement learning algorithm; a visualization interaction module connected to the management decision module and the dynamic evaluation module for real-time display of sound environment evaluation results, management operation status and prediction trends, and providing a parameter configuration interface.

[0017] In this embodiment, the data acquisition module is mainly responsible for completing the whole process transformation from sound environment physical signals to structured digital data; first, according to the sound environment monitoring requirements of different functional areas of the city (such as commercial areas, residential areas, industrial areas, etc.), the sampling parameters and network connection parameters of the sound sensor are configured; then the sound sensor continuously captures the sound wave signals in the environment and converts them into calculable digital signals; then the digital signals are analyzed and the sound pressure level, frequency spectrum and other key sound environment parameters are extracted; then the time stamp and geographic location information are added to these parameters to form structured sound environment data; finally, the structured data is transmitted to the data processing module in real time through the communication interface, ensuring the integrity and timeliness of the data throughout the process, laying a data foundation for subsequent data analysis and evaluation of the system; including: Sensor parameter configuration unit: The core function of this unit is to set appropriate operating parameters for the sound sensor based on the acoustic environment characteristics and monitoring needs of different urban functional areas, ensuring that the data collected by the sensor accurately reflects the acoustic environment conditions of the target area. Sampling parameter configuration: Adjust parameters according to the noise characteristics of different functional areas; for example, in areas around busy roads, the sampling rate of the sensor should be set to a higher value (such as 44.1 kHz) to capture the instantaneous changes in traffic noise; in residential areas, the sampling rate can be appropriately reduced (such as 22.05 kHz) to balance data accuracy and energy consumption; at the same time, set the sound pressure level range (such as 30dB~130dB) to ensure that the sound pressure range that may occur in the target area is covered, avoiding data overflow or insufficient precision; Network connection parameter configuration: According to the network coverage of urban functional areas, configure the communication protocol of the sensor (such as LoRa, NB-IoT or Ethernet), among which LoRa protocol is suitable for long-distance, low-power deployment in suburban or open areas, and NB-IoT is suitable for stable transmission in densely populated urban areas; at the same time, set the data transmission frequency (such as every 10 seconds to transmit real-time data), to ensure that data can be uploaded in time, and to avoid waste of network resources; Sound wave capture and signal conversion unit: This unit is responsible for converting the physical signal of sound waves in the environment into digital signals that the system can process, and is the core conversion link of data acquisition; Sound wave capture: The initialized sound sensor captures the sound wave signal in the surrounding environment through the built-in sound-sensitive element (such as a capacitive microphone), and converts the vibration of the sound wave into a corresponding analog sound and electrical signal; this process needs to ensure the stability of the sensitivity of the sound-sensitive element, for example, in strong wind or vibration interference areas, the shock-absorbing design of the sensor shell reduces the impact of environmental interference on sound wave capture; Signal conversion: Through the built-in analog-to-digital converter of the sensor, the analog sound and electrical signal is converted into a digital sound signal; the number of bits of the analog-to-digital converter (such as 16 bits or 24 bits) directly affects the conversion accuracy, and a 24-bit analog-to-digital converter can more accurately restore the details of the analog signal, reducing signal distortion and ensuring the accuracy and reliability of subsequent acoustic environment parameter calculations; high-precision analog-to-digital conversion is achieved by high-precision analog-to-digital converters to convert analog sound and electrical signals into digital sound signals; analog sound and electrical signals are continuous voltage signals, while digital signals are discrete binary data; the analog-to-digital converter periodically samples the analog signal and quantizes the voltage value obtained by sampling to binary numbers, completing signal conversion; the higher the number of bits of the converter, the higher the quantization accuracy; Acoustic environment parameter calculation unit: This unit is responsible for analyzing digital sound signals and extracting key parameters required for acoustic environment monitoring, providing key indicators for subsequent data structuring; Time domain analysis calculation: Time domain analysis is performed on the digital sound signal, and the sound pressure level is obtained by calculating the root mean square value of sound pressure per unit time. Time domain analysis can reflect the strength and weakness of sound in the time dimension, such as capturing the instantaneous sound pressure level change of sudden noise (such as car horn); Frequency domain analysis calculation: Through frequency domain analysis (such as Fast Fourier Transform), the digital sound signal is converted from time domain to frequency domain, and the frequency spectrum is obtained. Frequency spectrum can reflect the energy distribution of different frequency components, such as distinguishing the low-frequency engine sound from the high-frequency tire friction sound in traffic noise, providing basis for subsequent extraction of spectral features; Data structure processing unit: This unit is responsible for adding spatial and temporal identifiers to the calculated sound environment parameters to form structured data, ensuring data traceability and spatial correlation; Timestamp addition: Through the clock module built-in the sensor, accurate timestamps are added to each set of calculated sound pressure level and frequency spectrum, with a precision of seconds, ensuring that the time dimension can be aligned during subsequent data processing and multi-source data fusion; Geographical location information addition: Based on the pre-set geographical location coordinates (such as latitude and longitude) during sensor deployment, corresponding geographical location information is added to each set of sound environment parameters. This operation ensures that data can be associated with the spatial distribution of urban functional areas, such as identifying that a set of sound pressure level data comes from a certain street or residential area, providing spatial basis for subsequent zonal evaluation and management; Structured integration: Sound pressure level, frequency spectrum, timestamp, and geographical location information are integrated into structured data (such as JSON format) with a unified format, making data have a standard field and format, and facilitating direct reading and parsing by data processing modules; Timestamp identification relies on the real-time clock module built-in the sensor, which synchronizes with the Network Time Protocol (NTP) server to ensure the accuracy of the timestamp, so that each set of data can correspond to the exact collection time. Geographical location information identification is based on the latitude and longitude coordinates entered during sensor deployment, which are obtained through GPS positioning or manual measurement, ensuring accurate association of data and spatial location. Structured integration arranges multiple parameters according to fixed fields through a pre-set data format template, forming standardized structured data, which facilitates subsequent modules to quickly read the required information through field parsing, reducing the complexity of data parsing; Real-time data transmission unit: This unit is responsible for transmitting structured sound environment data to the data processing module in real time through the communication interface, ensuring the timeliness and transmission stability of the data; Communication interface management: Manage the communication interface of the sensor (such as RS485, Ethernet port or wireless communication module), establish a stable communication link between the sensor and the data processing module according to the pre-set network connection parameters; Data real-time transmission: structured data is continuously transmitted to the data processing module at a set transmission frequency. During transmission, a data verification mechanism (such as CRC verification) is used to detect whether data transmission has errors. If errors are found, a retransmission mechanism is triggered to ensure that data is not lost or damaged during transmission. Transmission status feedback: real-time monitoring of data transmission status. If communication interruption or transmission delay occurs, timely feedback of sensor transmission abnormal information to the system to facilitate maintenance personnel to troubleshoot network faults and ensure data transmission continuity. Data real-time transmission ensures transmission reliability through communication protocol selection and data verification mechanism. In view of the possible network interference in urban environment, a communication protocol with strong anti-interference ability (such as LoRa protocol with long distance, low power consumption and anti-interference characteristics) is selected. At the same time, CRC verification mechanism is used to calculate CRC verification value of data before transmission. The receiving end compares the recalculated verification value with the verification value of the sending end to determine whether the data is damaged. If damaged, retransmission is triggered to ensure data integrity during transmission. In addition, the communication link state is monitored through the heartbeat packet mechanism. If no heartbeat packet is received for a long time, it is determined that the communication is interrupted and the fault information is fed back to ensure the stability of the transmission link.

[0018] In this embodiment, the core task of the data processing module is to realize the purification and refinement of the original sound environment data. First, structured sound environment data transmitted by the data acquisition module is received. Invalid and abnormal data is removed through data cleaning. Environmental interference on data is eliminated through denoising processing. Then the data is standardized to a unified range through normalization processing to eliminate dimension differences. On this basis, equivalent continuous A sound level, maximum sound level, sound event duration, spectral features and other core sound environment feature parameters are extracted from the processed data. Finally, the normalized sound environment data and the extracted feature parameters are output to the multi-source data fusion unit and the dynamic evaluation module respectively to provide high-quality data support for subsequent multi-source data integration and sound environment dynamic evaluation. It includes: Data receiving unit: The data receiving unit is the connection interface between the data processing module and the data acquisition module, mainly responsible for stable reception of original sound environment data and ensuring data integrity. Communication link maintenance: establish and maintain real-time communication link with data acquisition module, support multiple communication protocols (such as LoRa, NB-IoT, Ethernet), automatically adapt to corresponding protocol according to data acquisition module configuration, ensure smooth link; Data receiving and caching: real-time receiving of structured sound environment data (including sound pressure level, frequency spectrum, timestamp, geographic location information) according to the transmission frequency of the data acquisition module, and temporary caching of received data to the local buffer area to avoid data loss due to data transmission rate fluctuations; Data integrity check: Perform integrity check on the received data to check whether the data format conforms to the preset specification (such as whether the number of fields is complete, whether the data type is correct), and if the data format is found to be incorrect or the fields are missing, send a retransmission request to the data acquisition module to ensure that the data format is complete before entering the subsequent processing link; Data cleaning unit: The data cleaning unit is responsible for removing invalid data and outliers in the sound environment data, improving data accuracy, and laying a foundation for subsequent processing; Invalid data identification: Scan the cached sound environment data and detect the integrity of the key fields, including sound pressure level, frequency spectrum, timestamp, and geographic location information. If any of the fields is missing, empty, or has a format error (such as a timestamp format that does not conform to the year-month-day hour: minute: second specification), it is determined to be invalid data; Outlier identification: Identify abnormal data using a combination of statistical analysis and common sense judgment. First, calculate the mean and standard deviation of the sound pressure level data set, and mark the sound pressure level data that exceeds the mean ± 3 times the standard deviation range as potential outliers. At the same time, check the frequency spectrum data. If the frequency component is lower than 20Hz or higher than 20000Hz (outside the audible range of human ears and not in line with the requirements of urban sound environment monitoring), or the geographic location information exceeds the longitude and latitude range of the target city functional area, it is also determined to be an outlier; Data cleaning execution: Remove the identified invalid data and outliers from the data set, and record the key information during the cleaning process, including the number of data removed, the reason for invalidity (such as field missing), and the specific value of the outlier, to form a cleaning log for subsequent tracing and problem troubleshooting; Data denoising unit: The data denoising unit is responsible for eliminating interference noise mixed in the sound environment data and preserving valid sound environment signals to ensure that the data truly reflects the actual sound environment conditions; Noise type judgment: Analyze the characteristics of the cleaned data to determine the type of noise contained, including random noise generated by the sensor circuit (small amplitude, chaotic frequency), and external transient interference noise (such as sudden rise in sound pressure level caused by sudden electromagnetic interference); Denoising algorithm selection and execution: Select appropriate denoising algorithms for different noise types. For random noise, use wavelet transform denoising algorithm to decompose the sound signal into components of different frequency scales, retain the low-frequency components containing valid sound information, and remove the high-frequency components containing random noise. For transient interference noise, use sliding mean filter algorithm to smooth the transient abnormal values by calculating the mean of adjacent data points; Verification of denoising effect: After denoising, compare the sound pressure level change curve and frequency spectrum distribution before and after denoising. If the smoothness of the curve is improved and the effective sound signal characteristics (such as the specific frequency peak of traffic noise) are not lost, the denoising effect is qualified. If there is still obvious noise interference, adjust the denoising algorithm parameters again and process again. The core principle of the data denoising unit is signal decomposition and effective component retention. The wavelet transform denoising technology decomposes the sound signal into wavelet components of different frequency scales. Effective sound environment signals (such as traffic and industrial noise) are usually concentrated in low frequencies or specific frequency bands, and have stable amplitude. Random noise is usually distributed in high frequency bands and has small amplitude. By removing high frequency noise components and reconstructing low frequency effective components, denoising can be achieved. The sliding mean filter technology smooths the sudden data caused by transient interference by calculating the mean value of adjacent data points. It uses the time continuity of data to weaken the influence of isolated outliers. The combination of the two technologies can remove different types of noise while retaining the detailed characteristics of effective sound signals. Data normalization unit: The data normalization unit is responsible for standardizing sound environment data of different dimensions to a unified numerical range, eliminating the influence of dimension differences on subsequent analysis. Dimension difference analysis: Identify the dimension differences of each parameter in the denoised data, such as sound pressure level in dB (decibels) and frequency in Hz (hertz). Different dimensions will cause an imbalance in the weights of each parameter when constructing the feature vector, so normalization is needed to eliminate this effect. Normalization method execution: Use the min-max normalization method to map each parameter value to a unified range of 0 to 1. The specific process is as follows: first, calculate the minimum and maximum values of a certain parameter (such as sound pressure level) in the data set. Then, convert each data point of the parameter according to the rule of (current data value - minimum value) / (maximum value - minimum value) to ensure that the converted data is between 0 and 1. Normalization result check: Check if the normalized data is within the preset range (0 to 1). If there are data outside the range, recalculate the minimum and maximum values of the parameter to determine if the abnormal values have not been completely removed. After correction, perform the normalization operation again to ensure that all data meet the standardization requirements. The principle of data normalization is dimension unification and numerical mapping. The dimension differences of different sound environment parameters (such as dB and Hz) will cause the parameter with a large numerical range (such as frequency) to have a much greater impact on the result than the parameter with a small numerical range (such as sound pressure level) in subsequent multi-parameter analysis, violating the fairness of parameter weights. The min-max normalization method maps all parameters to a unified range of 0 to 1, making the numerical change range of each parameter consistent and eliminating the weight imbalance problem caused by dimension. This provides a fair parameter basis for subsequent feature vector construction and model calculation. Characteristic parameter extraction unit: The characteristic parameter extraction unit is the core of the data processing module, responsible for extracting core characteristic parameters that can reflect the quality of the acoustic environment from the normalized acoustic environment data; Equivalent continuous A sound level extraction: Set a time window (e.g. 1 minute), calculate the energy average of all sound pressure level data in the time window, which is the equivalent continuous A sound level; When calculating, the frequency response characteristics of the A weighting network should be considered to highlight the contribution of middle and high frequency noise sensitive to human ears, reflecting the overall noise level in the time window; Maximum sound level extraction: In the same time window, scan all sound pressure level data and select the maximum sound pressure level as the maximum sound level in the time window, reflecting the highest intensity of sudden noise (such as car horn, construction impact generated instantaneous high sound pressure) in the time window; Sound event duration extraction: First set the sound event judgment threshold (e.g. 5dB higher than the equivalent continuous A sound level), when the sound pressure level data continuously exceeds the threshold, it is judged as a sound event, and the time interval from the beginning to the end of the event is recorded as the sound event duration, reflecting the influence period of a specific high intensity noise event; Spectral feature extraction: Analyze the frequency spectrum data in the time window and extract key spectral features, including center frequency (frequency point with concentrated energy in frequency spectrum), frequency bandwidth (frequency range containing main energy), energy proportion of each octave band (such as 63Hz, 125Hz, 250Hz, etc. The proportion of energy in total energy), reflecting the frequency distribution characteristics of noise, helping to identify noise source types (such as low frequency mechanical noise, high frequency traffic noise); The principle of acoustic environment characteristic parameter extraction is data abstraction and key information extraction; The equivalent continuous A sound level is based on the principle of sound energy superposition, which converts the fluctuating sound pressure level into a single value by calculating the average value of sound energy in the time window, reflecting the overall noise level; The maximum sound level is based on the principle of extreme value screening, capturing the peak data in the time window, reflecting the intensity of sudden noise; The sound event duration is based on the principle of threshold triggering and time statistics, identifying specific sound events by setting a reasonable threshold and calculating their duration; The spectral feature is based on the principle of frequency energy distribution analysis, which extracts the frequency properties of noise by analyzing the energy concentration area and distribution proportion of the frequency spectrum. These characteristic parameters abstract the original data from different dimensions to form core indicators describing the quality of the acoustic environment; Data output unit: The data output unit is responsible for transmitting the processed standardized data and extracted characteristic parameters to the subsequent module, ensuring the timeliness and standardization of data transmission; Data arrangement: The normalized acoustic environment data and the extracted acoustic environment feature parameters are arranged into data sets in a predetermined format, respectively. The normalized data set includes timestamp, geographic location information, normalized sound pressure level, and frequency spectrum. The feature parameter data set includes timestamp, geographic location information, equivalent continuous A-weighted sound level, maximum sound level, sound event duration, and spectral features. Directional transmission: According to the system's preset data flow path, the normalized data set is transmitted to the multi-source data fusion unit for fusion processing with auxiliary data. The feature parameter data set is transmitted to the dynamic evaluation module as the core input data of the acoustic environment evaluation model. Transmission status feedback: Real-time monitoring of data transmission status, if data successfully arrives at the target module, record transmission time and data volume; if transmission fails (such as communication interruption), temporarily store data and try to retransmit, at the same time send transmission abnormal alarm to the system, ensure that data is not lost and can be delivered in time.

[0019] In this embodiment, the core task of the multi-source data fusion unit is to integrate two types of key data: one is the standardized acoustic environment data (including normalized sound pressure level, frequency spectrum, and extracted acoustic environment feature parameters) from the data processing module; the other is the auxiliary data (including traffic flow data, meteorological data, urban planning data, real-time activity data) from external systems. The workflow revolves around data reception-preprocessing-time and space alignment-fusion integration-output. Through cleaning and standardizing auxiliary data, the format differences and quality problems of different data sources are eliminated. Based on timestamp and geographic location information, accurate matching with acoustic environment data is realized, and multi-source fusion data containing acoustic environment features and external influencing factors are finally generated. This provides a data analysis basis for the dynamic evaluation module to build a more realistic acoustic environment scene, avoiding evaluation bias caused by single data. The multi-source data fusion unit includes: Auxiliary data reception: Auxiliary data reception is the connection interface between multi-source data fusion and external systems, responsible for stable access to various external auxiliary data, ensuring real-time and integrity of data acquisition. External system interface: For different types of auxiliary data, establish a dedicated data channel with the corresponding external system; for example, connect the traffic flow monitoring platform of the city traffic management department to obtain real-time vehicle flow and vehicle distribution data; connect the real-time weather service platform of the meteorological department to obtain temperature, humidity, wind speed, precipitation, and other meteorological parameters; connect the urban planning database of the natural resources department to obtain functional zoning, land use properties, building density, and other static data; connect the activity management system of large-scale event organizers to obtain the time, location, and number of participants of concerts, exhibitions, sports events, and other real-time activities. Multi-protocol adaptation: Support multiple data transmission protocols to be compatible with the output formats of different external systems, such as receiving structured traffic flow and weather data through API interface, receiving low-power and real-time activity data through MQTT protocol, and periodically obtaining low-update-frequency urban planning data through FTP protocol, to ensure smooth access of various auxiliary data; Receiving state monitoring: Real-time monitoring of data transmission link state with each external system, recording data receiving timestamp, data volume, and integrity identifier; if link interruption (such as weather data transmission timeout) or data loss (such as no traffic flow data received in a certain period) occurs, immediately trigger the reconnection mechanism or send a retransmission request to the external system to ensure continuous acquisition of auxiliary data; Auxiliary data preprocessing: Auxiliary data preprocessing is responsible for improving the quality of external auxiliary data, eliminating data noise and format differences, and providing standardized data basis for subsequent spatio-temporal alignment; Data cleaning: Eliminate invalid data and outliers through rule engine and statistical analysis; for invalid data, filter out records missing key fields (such as traffic flow data missing monitoring road section ID, statistical period, weather data missing observation station location, observation time) and directly eliminate; for outliers, based on domain knowledge and statistical threshold (such as temperature of weather data in urban environment exceeding -30℃ to 50℃, traffic flow data negative or far exceeding road design capacity in a certain period, real-time activity data participant number is 0 but activity status is in progress), replace the outliers with the historical same period mean or adjacent period mean of the indicator, to avoid abnormal data interference with the fusion result; Format standardization: Unify heterogeneous data output by different external systems into a system-compatible format; first, unify data storage format, convert auxiliary data in different formats such as JSON, CSV, XML to JSON format; second, unify field naming and data unit, such as traffic flow (vehicles / hour), hourly vehicle number in traffic flow data are unified named as traffic_flow and keep unit as vehicles / hour, temperature (℃), air temperature (℃) in weather data are unified named as temperature and keep unit as ℃, to ensure that fields can be directly matched during subsequent data processing; Spatio-temporal alignment: Spatio-temporal alignment is the core of realizing the association between sound environment data and auxiliary data, through matching in time and space dimensions, to ensure that both types of data correspond to the same spatio-temporal scenario, laying a foundation for fusion processing; Temporal dimension alignment: Based on the timestamps of the data, the temporal granularity is unified. If the auxiliary data and the sound environment data have the same temporal granularity (e.g., both are 1 minute per time), they are directly matched by timestamp. If there is a difference in temporal granularity (e.g., traffic flow data is 5 minutes per time, and sound environment data is 1 minute per time), an aggregation algorithm is used to aggregate multiple fine-grained sound environment data into coarse-grained data (e.g., the average of 5 one-minute equivalent continuous A sound levels is taken as the sound environment indicator for a 5-minute period). If the auxiliary data has a finer temporal granularity (e.g., real-time activity data is 30 seconds per time), an interpolation algorithm is used to interpolate the fine-grained auxiliary data to the same granularity as the sound environment data, ensuring that the data in the same period corresponds one-to-one. Spatial dimension alignment: Based on geographic location information, regional matching is achieved. The sensor latitude and longitude coordinates of the sound environment data are associated with the spatial identifiers of the auxiliary data, such as the monitoring road section latitude and longitude range in traffic flow data, the observation station latitude and longitude in meteorological data, the functional area boundary latitude and longitude in urban planning data, and the activity site latitude and longitude in real-time activity data. Through spatial coordinate mapping, it is determined whether the auxiliary data and the sound environment data belong to the same region (e.g., if the sensor latitude and longitude falls within the latitude and longitude range of a certain traffic road section, then the traffic flow data of that road section matches the sound environment data of that sensor), forming a data set with time and space dual-dimension association. Data fusion processing: Data fusion processing is responsible for deep integration of the sound environment data and auxiliary data after time and space alignment, generating multi-source fusion data with sound environment characteristics and external influencing factors. Data weight allocation: According to the influence degree and reliability of the data on the sound environment evaluation, different data types are assigned fusion weights. Core sound environment data (such as equivalent continuous A sound level, spectral characteristics) have higher weights (such as 0.6) because they directly reflect the sound environment quality. Auxiliary data with high correlation to noise (such as traffic flow data, which is directly related to traffic noise) have lower weights (such as 0.2). Auxiliary data with moderate correlation (such as wind speed, which affects noise propagation distance) have a weight of 0.1. Basic background auxiliary data (such as functional area types in urban planning, which determine noise limit standards) have a weight of 0.1, ensuring that the fusion results are more in line with core evaluation requirements. Feature layer fusion: Extract key features from various data and integrate them. From sound environment data, extract features such as equivalent continuous A sound level and maximum sound level. From traffic flow data, extract features such as peak period vehicle flow and large vehicle proportion. From meteorological data, extract features such as average wind speed and wind direction. From urban planning data, extract features such as functional area type and building shielding coefficient. From real-time activity data, extract features such as activity type and peak participation number. Integrate all features into a unified sound environment-auxiliary factor feature set, forming a multi-dimensional and comprehensive fusion data representation. Fusion data output: The fusion data output is responsible for directing the transmission of the processed multi-source fusion data to the dynamic evaluation module, while ensuring the integrity and timeliness of data transmission. Targeted data transmission: According to the system's preset data flow path, multi-source fusion data (including sound environment features, auxiliary data features, and space-time identifiers) are transmitted to the dynamic evaluation module as input data for the sound environment evaluation model, supporting the calculation of dynamic evaluation indexes. Transmission status feedback: Real-time monitoring of data transmission process, recording the time and data volume of successful transmission. If transmission fails (such as network interruption causing data loss), trigger the retry mechanism to retransmit the failed data. If multiple retries still fail, send a transmission exception alert to the system to remind maintenance personnel to troubleshoot link problems, ensuring that multi-source fusion data can be delivered to the dynamic evaluation module in a complete and timely manner.

[0020] In this embodiment, the core task of the dynamic evaluation module is to generate accurate sound environment evaluation indexes based on two types of input data: one is the sound environment feature parameters from the data processing module, including equivalent continuous A sound level, maximum sound level, sound event duration, and spectral features; the other is auxiliary data from the multi-source data fusion unit, including traffic flow data, weather data, urban planning data, and real-time activity data. The working logic revolves around data reception-feature integration-model calculation-deviation monitoring-model optimization-index output. First, different types and dimensions of input data are integrated into a unified multi-dimensional feature vector, which is input into a pre-trained deep learning model for forward propagation calculation to generate an initial sound environment evaluation index. Then, real-time monitoring of the distribution deviation between new data and model training data is performed. When the deviation exceeds the preset tolerance, the model parameters are adjusted using real-time samples through online learning algorithms, and the optimized evaluation index is recalculated after updating the model. Finally, the optimized index and historical evaluation data are output to the management decision module, and the current evaluation data is stored in the model training database to accumulate data for subsequent model iteration, realizing the dynamic cycle of evaluation-optimization-re-evaluation, including: Input data reception: Input data reception is the connection interface between the dynamic evaluation module and the data processing module and multi-source data fusion unit, responsible for stable acquisition of two types of core input data, ensuring the integrity and timeliness of data transmission. Dual-source data docking: Establish dedicated data channels with the data processing module and the multi-source data fusion unit respectively; Receive structured sound environment feature parameters from the data processing module, including the specific value of each feature parameter, the corresponding timestamp and geographic location information; Receive integrated auxiliary data from the multi-source data fusion unit, including traffic flow (such as vehicle flow per unit time, large vehicle proportion), meteorological parameters (such as temperature, wind speed), urban planning attributes (such as functional area type, land use property), real-time activity information (such as activity type, number of participants), and corresponding spatiotemporal identifiers; Data integrity check: Perform field integrity check on the received two types of data to confirm that the sound environment feature parameters have no missing (such as no equivalent continuous A sound level data), and the auxiliary data key fields have no omission (such as no timestamp for traffic flow data); If data is missing or format is incorrect, immediately send a retransmission request to the corresponding data source module to ensure that there is no missing key information in the input model data; Data caching and sorting: Temporarily store the received two types of data in the local cache according to the timestamp order to avoid time sequence confusion caused by differences in data transmission rate; When the sound environment feature parameters and auxiliary data of the same spatiotemporal dimension are both received, trigger the subsequent feature vector construction process to ensure the time sequence consistency of data processing; Multi-dimensional feature vector construction: Multi-dimensional feature vector construction is responsible for eliminating the type difference and dimension influence of input data, integrating sound environment feature parameters and auxiliary data into a unified format feature vector, and providing standardized input for model calculation; Data type unification: Convert different types of input data into numerical format; For example, convert the functional area type (residential area, commercial area, industrial area) in urban planning data into numerical coding (such as residential area coding as 1, commercial area coding as 2, industrial area coding as 3); Convert the activity type (concert, exhibition) in real-time activity data into corresponding numerical identifier (such as concert coding as 1, exhibition coding as 2), to ensure that non-numeric data can be recognized and calculated by the model; Dimension standardization: Standardize parameters of different dimensions; For example, the unit of equivalent continuous A sound level in sound environment feature parameters is decibel, and the unit of traffic flow in auxiliary data is vehicle / hour, and the unit of wind speed is meter / second. Through standardization algorithm, all parameter values are mapped to the same range (such as 0 to 1), to avoid the model being overly sensitive to certain parameters due to dimension differences, and to ensure fair weighting of all parameters in model calculation; Vector integration: splice the normalized sound environment feature parameters and auxiliary data into a multi-dimensional feature vector in a predetermined order; for example, the vector dimensions are equivalent continuous A sound level, maximum sound level, sound event duration, spectral features, traffic flow, wind speed, function area code, and activity type code, respectively, each dimension corresponds to a normalized parameter value, forming a unified data structure that can be directly input into the model; Pre-trained model calculation: The pre-trained model calculation is the core calculation link of the dynamic evaluation module. Through a multi-layer neural network structure, the complex relationship between sound environment quality and multi-source features is simulated to generate an initial sound environment evaluation index. Pre-trained model loading: load the pre-trained sound environment evaluation model at startup. This model is trained based on historical sound environment data and corresponding auxiliary data and includes a multi-layer neural network structure with input layer, hidden layer, and output layer. The number of input layer nodes is consistent with the dimension of the multi-dimensional feature vector. The hidden layer extracts hierarchical features through multiple neurons. The output layer outputs a single numerical initial sound environment evaluation index. Forward propagation calculation: input the constructed multi-dimensional feature vector into the model input layer. The data is calculated through each hidden layer neuron in turn. Each hidden layer neuron performs linear transformation and nonlinear activation processing on the input data, extracting high-level abstract features layer by layer. For example, the first hidden layer extracts the basic correlation features of equivalent continuous A sound level and traffic flow, and the deep hidden layer extracts the complex interaction features of spectral features, wind speed, and function area type, gradually uncovering the complex rules reflecting sound environment quality in the data. Initial index generation: after multiple forward propagations, the data reaches the model output layer. The output layer converts high-level abstract features into a single numerical initial sound environment evaluation index through regression analysis algorithm. The index value corresponds to the sound environment quality level (e.g., the lower the index, the better the sound environment quality, and the higher the index, the more serious the noise pollution), providing a basic result for subsequent evaluation and model optimization. Model update trigger: The model update trigger is responsible for monitoring the distribution difference between real-time data and model training data, determining whether to adjust model parameters, and ensuring the adaptability of the model to dynamic environments. Data distribution monitoring: real-time calculation of the distribution deviation of the new input multi-dimensional feature vector and the model training data set. Analyze the mean and standard deviation difference of the two types of data in each dimension parameter through statistical methods, for example, compare the mean deviation of traffic flow in real-time data and traffic flow in training data, and analyze the distribution range of equivalent continuous A sound level in real-time data and the coincidence degree with training data. Deviation threshold judgment: compare the calculated distribution deviation with the preset tolerance; if the deviation does not exceed the tolerance, it means that the distribution characteristics of real-time data and model training data are consistent, and the current model parameters can accurately evaluate the sound environment quality, and there is no need to update the model; if the deviation exceeds the tolerance (such as a sudden large-scale activity in a certain period of time leading to a too large difference between real-time activity data distribution and training data), the model updating mechanism is triggered, and the online learning optimization is started to adjust the model parameters; Trigger record and feedback: record the deviation monitoring result and the judgment of whether to trigger the update each time, and if the update is triggered, send the update instruction to the online learning optimization, and at the same time mark the key reason of this time triggering (such as abnormal distribution of real-time activity data), which provides direction for subsequent model optimization; Online learning optimization: Online learning optimization is responsible for adjusting the model parameters using real-time data when the model needs to be updated, generating an optimized model with environmental adaptability; Real-time sample construction: use the multi-dimensional feature vector at the time of triggering the update as the input sample, and combine the actual sound environment quality feedback (such as the standard index corresponding to the sound environment grade monitored by artificial monitoring) in the space-time scene as the expected output to construct the real-time training sample for model optimization; Loss function calculation: calculate the difference between the initial evaluation index of the model output and the expected output through the loss function; the loss function value reflects the deviation degree of the model prediction result and the actual situation, and the larger the value, the greater the deviation, which requires a larger adjustment of the model parameters; Gradient backpropagation: based on the loss function value, the gradient backpropagation algorithm is used to calculate the adjustment amount of the connection weight and bias parameter of each layer of neurons from the output layer to the input layer; for example, if the output layer loss is caused by the too high weight of a certain hidden layer neuron on the traffic flow parameter, the specific down-regulation amplitude of the weight is determined through gradient calculation, and the model prediction deviation is gradually reduced; Model parameter update: update the connection weight and bias parameter of each layer of the model according to the calculated adjustment amount, and generate the optimized sound environment evaluation model; the updated model can better adapt to the distribution characteristics of the current real-time data, and reduce the evaluation deviation caused by environmental changes; Evaluation index generation and output: Evaluation index generation and output is responsible for generating the final optimized evaluation index, and completing the directional output and storage of data, providing support for management decision and model iteration; Optimized index calculation: input the multi-dimensional feature vector before triggering the update into the updated optimized model, and calculate the optimized sound environment evaluation index through forward propagation; the index is more consistent with the current actual sound environment quality than the initial index, and eliminates the deviation caused by the inadaptability of the model parameters; Data-oriented output: The optimized sound environment evaluation index and the corresponding historical evaluation data (such as the evaluation index sequence within the last 1 hour) are packaged and transmitted to the management decision module to provide direct basis for early warning, control, and strategy optimization. At the same time, the multi-dimensional feature vector, initial index, optimized index, and model adjustment parameters generated during the evaluation process are stored in the model training database to accumulate samples for subsequent offline iterative training of the model; Output state monitoring: Real-time monitoring of data output process to confirm that the evaluation index is successfully sent to the management decision module and the training data is successfully stored in the database. If output fails (such as database connection interruption), a retry mechanism is triggered to ensure data is not lost and the integrity of the system data link is guaranteed.

[0021] In this embodiment, the management decision module is mainly responsible for three core tasks: first, determining the current sound environment state level based on the sound environment evaluation index and timely transmitting abnormal information through the hierarchical early warning mechanism; second, automatically controlling noise source equipment through the Internet of Things platform to reduce noise pollution according to the preset management strategy; third, iteratively optimizing the management strategy library using reinforcement learning algorithm relying on historical management effect data to improve management accuracy. Through the closed-loop operation of judgment-execution-optimization, the module realizes the dynamic, intelligent, and efficient management of sound environment, which not only ensures the rapid disposal of sudden noise events but also promotes the continuous improvement of long-term management strategies, providing adaptive sound environment management solutions for different urban functional areas (such as residential areas, commercial areas, and industrial areas), including: Early warning unit: Evaluation index reception and analysis: Real-time reception of sound environment evaluation index output by the dynamic evaluation module, synchronous acquisition of time stamp and geographic location information corresponding to the index, and analysis of the urban functional area (such as Dongcheng residential area, Xicheng commercial area) and monitoring period (such as daytime 8:00-22:00, night 22:00-8:00 next day) to which the current sound environment data belongs, providing scene basis for subsequent hierarchical early warning; Threshold comparison and level determination: Retrieve the sound environment threshold standards for the corresponding functional area and corresponding period in the preset management strategy library, compare the real-time evaluation index with the threshold, and determine the current sound environment state level. For example, the daytime threshold for residential areas is set to evaluation index 60 (corresponding to sound pressure level 55 decibels), and the nighttime threshold is set to evaluation index 50 (corresponding to sound pressure level 45 decibels). If the real-time index is 65, it is determined as level one warning; if the index is 62, it is determined as level two warning; if the index is 58, it is determined as level three warning; Hierarchical early warning information generation: generate differentiated early warning information according to the determined early warning level; the first level early warning information contains the exceeding range, the influence range, and the recommended emergency disposal measures (such as temporarily limiting vehicle access, suspending high-noise operation); the second level early warning information contains the exceeding reason analysis (such as the sudden increase of vehicle flow near the road section, sudden commercial activities), and the recommended regular disposal measures (such as increasing traffic diversion personnel, reminding the activity organizer to reduce the volume); the third level early warning information contains the attention period and the recommended monitoring frequency improvement scheme (such as adjusting from monitoring once every 10 minutes to once every 5 minutes); Early warning information distribution: distribute the generated early warning information to relevant terminals through multiple channels; for management personnel, send to mobile phone mobile APP and trigger pop-up window reminder; for city monitoring center, push to large screen display system and mark the abnormal area geographic location; for relevant responsible units (such as traffic management department, environmental protection law enforcement department), send instructive notice through special communication channel to ensure that the early warning information quickly reaches the responsible person; Control unit: Control logic retrieval and analysis: retrieve the corresponding noise source control logic from the management strategy library according to the current sound environment state level and the function area it belongs to; Internet of Things instruction generation: convert the retrieved control logic into standardized Internet of Things control instructions; the instructions contain target device identification (such as factory equipment number, traffic signal lamp ID, commercial audio terminal address), control parameters (such as power value, timing duration, on-off state), execution time limit (such as executing until the sound environment index falls below the threshold value, executing only during 18:00-21:00 period), to ensure that the instructions can be recognized by different types of noise source devices; Instruction issuing and execution monitoring: issue control instructions to target noise source devices through Internet of Things platform, and receive real-time feedback of instruction execution state (such as device has received instruction, power has been adjusted to 80%, instruction execution failure) from devices; if instruction execution failure occurs (such as device offline, parameter adjustment out of range), immediately generate backup instructions (such as switching to backup noise source control device, adjusting control parameter range) and reissue until the device successfully executes the control operation; Control effect feedback: after the execution of the control instruction, continuously receive the sound environment evaluation index output by the dynamic evaluation module to analyze the index change trend and evaluate the control effect; for example, after executing the instruction to reduce the power of factory equipment, if the sound environment index decreases from 68 to 58 within 15 minutes, it is determined that the control effect is good; if the index only decreases to 65, it is determined that the effect is not good, and the control strategy library needs to be re-retrieved to supplement additional control operations such as turning on additional sound insulation facilities; Strategy optimization unit: Historical management data collection: Two types of historical data are systematically collected. One is management operation data, including the warning level of each execution, control instruction content, execution time, and the functional area and equipment involved. The other is management effect data, including the change in sound environment evaluation index before and after the execution of management operations, the time length of the index falling below the threshold, and the change in resident complaint volume (obtained by connecting to the 12345 citizen hotline data). Both types of data are stored in chronological order and classified by functional area to form a historical management database. Reinforcement learning sample construction: Reinforcement learning samples are constructed based on historical management data. The sound environment state (such as the evaluation index, functional area type, and time period) is used as the sample input, the management strategy (such as the warning level and control instruction combination) is used as the sample action, and the management effect (such as the index decline amplitude, falling time length, and complaint volume reduction ratio) is used as the sample reward value. Policy iteration optimization: Reinforcement learning algorithms are used to iteratively optimize the management strategy library. The agent in the algorithm aims to maximize the long-term management effect by continuously learning from historical samples to adjust strategy parameters. For strategies with good management effects (such as a control instruction combination that causes the index to quickly fall), the probability of calling the strategy in the corresponding sound environment state is increased. For strategies with poor management effects (such as a warning level that does not timely trigger subsequent control, resulting in sustained index exceeding), the threshold settings or control instruction combination in the strategy are adjusted. Optimized strategy verification and update: The optimized management strategy is verified in a small range of functional areas, and the sound environment management effect (such as the number of exceedances and average falling time length) in the verification area and the non-verification area is compared. If the verification result shows that the management effect of the optimized strategy has improved by more than 15%, the strategy is officially updated to the management strategy library. If the effect does not meet expectations, more special scene data (such as sound environment management data under adverse weather conditions) is added to the sample construction stage, and iterative optimization is performed again to ensure that each strategy update improves management accuracy.

[0022] In this embodiment, the visual interaction module mainly undertakes four core functions: 1) data reception and conversion, real-time acquisition of sound environment evaluation index and management operation state data, and conversion into graphical elements; 2) visual content construction, generation of real-time evaluation charts, management state panels, and sound environment quality prediction curves; 3) comprehensive interface presentation, combination of multiple types of visual content into interfaces suitable for different terminals and rendering output; 4) parameter configuration and update, receiving of user input management strategy parameters and display preference parameters, and synchronous updating of management strategy library and interface style. Through the data-graphics-interaction process, the module eliminates data barriers, forms a visual closed loop of sound environment management state-operation-effect, supports daily monitoring and decision-making for management personnel, and provides intuitive data presentation for superior departments to view sound environment governance effectiveness. The visual interaction module includes: Data reception and analysis unit: Multi-source data acquisition: Real-time docking of dynamic evaluation module and management decision module, obtaining sound environment evaluation index, corresponding timestamp, geographic location information and prediction trend data from dynamic evaluation module; obtaining management operation state data from management decision module, including early warning information sending record (early warning level, receiving terminal, sending time), noise source control instruction execution situation (target equipment, control parameter, execution result) and strategy optimization record (optimization time, adjusted strategy content); Data integrity check: Field check on the acquired data, check if the sound environment evaluation index is missing value, if the management operation state data is missing execution result identifier, if the prediction trend data contains complete time series; if data is missing or format error is found, send a retransmission request to the corresponding module to ensure that the data for subsequent visualization processing is accurate and correct; Data classification and association: The received data is classified according to time-space dimension, for example, daily data is divided according to date, regional data is divided according to city function area; at the same time, data association relationship is established, sound environment evaluation index of a certain period in a certain area is bound with management operation (such as early warning sending, equipment control) in the same period to form state-operation associated data set, which provides support for subsequent visual display of linkage query; Visualization conversion unit: Numerical data graphical conversion: Convert numerical data such as sound environment evaluation index into graphical elements; for example, convert continuous time series evaluation index into line chart data points, the higher the index, the higher the point position in the line chart; convert real-time evaluation index of different functional areas into column chart data, each functional area corresponds to a column, and the column height represents the index size; convert the comparison relationship between evaluation index and threshold into dashboard elements, the dashboard scale corresponds to the index range, the pointer points to the current index position, and the threshold interval is marked with different colors (green for normal, yellow for mild over-standard, red for severe over-standard); State data symbolization conversion: Convert state data such as management operation state into symbolized elements; for example, in the early warning information sending record, the read state is represented by a blue check symbol, and the unread state is represented by a red exclamation mark symbol; in the noise source control instruction execution situation, execution success is represented by a green circle symbol, execution failure is represented by a red cross symbol, and execution in progress is represented by a yellow rotating loading symbol; in the strategy optimization record, optimization completion is represented by a purple pentagram symbol, and verification pending is represented by a gray question mark symbol; Spatial data mapping conversion: converting data with geographical location information into map mark elements; for example, marking the positions of sound sensors with different color dots on the city electronic map, and the color of the dot corresponds to the sound environment evaluation index level calculated by the data collected by the sound sensor (green dot for normal, yellow for third-level warning, and red for first-level warning); clicking the dot can pop up detailed data of the position, including real-time index, today's index change curve, and surrounding noise source distribution; Chart and panel building unit: Real-time sound environment evaluation chart building: based on the converted graphical elements, generate multiple types of real-time evaluation charts; time series chart shows the change of sound environment evaluation index in a certain period (such as 24 hours, 7 days) in a single functional area, with time on the horizontal axis and evaluation index on the vertical axis, and the threshold line of the corresponding period is marked; regional comparison chart shows the difference of evaluation index in different functional areas at the same time, with functional area name on the horizontal axis and evaluation index on the vertical axis, which facilitates quick identification of over-standard areas; noise source correlation chart shows the correlation between the evaluation index of a certain area and the surrounding noise sources (such as traffic intersection, factory, and shopping mall), connecting the index data and noise sources with lines, and the line thickness represents the influence degree; Management operation state panel building: integrate symbolic management operation data to generate a management operation state panel; the early warning state panel lists recent early warning records in chronological order, including early warning level, occurrence area, sending terminal, feedback state, and disposal result; the control operation panel displays noise source control instructions by device type, including device name, control parameter, execution time, execution result, and subsequent index change; the strategy optimization panel displays recent strategy adjustment content, including pre-optimization strategy, post-optimization strategy, optimization basis, and verification effect; Sound environment quality prediction curve building: integrate the prediction trend data of the dynamic evaluation module, combine with historical evaluation index, and generate a prediction curve through time series analysis; short-term prediction curve shows the predicted value of sound environment evaluation index in each period in the next 24 hours, and marks the period when over-standard may occur; long-term prediction curve shows the predicted value of daily average evaluation index in the next 7 days, compares with historical data of the same period, and predicts the trend of sound environment quality; the prediction curve is represented by a dashed line, with the prediction error range marked to ensure the reference of the prediction result; Comprehensive interface combination unit: Interface layout design: According to different display devices (such as monitoring center large screen, computer terminal for management personnel, mobile APP), design adaptive layout; Monitoring center large screen adopts multi-region split screen layout, left side is city electronic map (show real-time index mark of each region), middle is real-time evaluation time series chart and regional comparison chart, right side is management operation state panel; Computer terminal adopts up-down layered layout, upper part is real-time index overview (current index of each function area), middle is prediction curve and real-time evaluation chart switching area, lower part is management operation detail panel; Mobile terminal adopts vertical scrolling layout, showing map mark, real-time index, prediction curve, management operation record in turn, adapting to small screen view; Visual content association combination: Establish the linkage relationship of different visual content; Click on the prediction curve of a certain over-standard period, automatically pop up the management strategy suggestion corresponding to the period, realize the linkage query of figure-data-operation; Interface element integration: Real-time evaluation chart, management operation state panel, prediction curve and map mark are integrated into a unified comprehensive interface, add interface navigation bar (including data query, parameter configuration, history record, etc. Entrance), time selector (supporting data filtering by time period) and function area filter (supporting data viewing by region), ensure that the interface contains comprehensive visual content and convenient operation entrance; Interface rendering output unit: Multi-terminal adaptive rendering: According to the resolution and display ratio of display device, render and adjust the comprehensive interface; Monitoring center large screen is rendered by high resolution to ensure that the map mark and chart details are clear; Computer terminal is rendered by regular resolution to optimize chart font and panel layout; Mobile terminal is rendered adaptively by different screen sizes to adjust element size and spacing to avoid content overflow or incomplete display; Dynamic content real-time update: Set the update frequency of visual content, sound environment evaluation index and management operation state data are updated once a minute to ensure the real-time performance of the interface; Prediction curve is updated once every 2 hours to adjust the prediction result combined with the latest data; The regional index mark on the electronic map changes color with real-time data to ensure that the management personnel always see the latest state; Rendering effect optimization: Optimize the visual elements of the interface, charts use gradient color filling (such as light blue gradient filling under the fold line of time series chart), improve the visual level; Add dynamic effect to status symbol to enhance the interface interaction; Important data (such as over-standard index, failed control operation) is highlighted with flashing effect to remind the management personnel to pay attention; Parameter configuration unit: The configuration interface provides: setting parameter configuration entrance in the integrated interface navigation bar, and entering the configuration interface by clicking; the configuration interface is divided into two partitions, a management strategy parameter configuration area and a display preference parameter configuration area; the management strategy parameter configuration area provides threshold setting (sound environment evaluation index threshold of each functional area in each period), early warning rule setting (corresponding early warning level of different over-standard amplitudes), and control instruction template setting (default control parameters of different device types); the display preference parameter configuration area provides chart style setting (line chart color, column chart style, and instrument panel scale), interface layout setting (display order of each panel, whether to hide a certain type of panel), and data update frequency setting (real-time data update interval and prediction curve update interval); User input receiving and verification: receiving the parameters input by the user in the configuration interface, reasonably verifying the management strategy parameters, for example, the threshold setting needs to comply with the national sound environment quality standard, and the daytime threshold needs to be higher than the nighttime threshold; verifying the format of the display preference parameters; if the input parameters are unreasonable, a prompt information is popped up to guide the user to correct; Parameter synchronous updating: sending the verified management strategy parameters to the management decision module, updating the preset management strategy library, and ensuring that the subsequent sound environment evaluation and management operation are performed according to the new parameters; applying the verified display preference parameters to the integrated interface, and adjusting the chart style, interface layout, and data update frequency in real time, so that the user can see the configuration effect immediately.

[0023] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A dynamic evaluation and management system for the acoustic environment of urban functional areas, characterized in that: include: The data acquisition module is used to collect acoustic environment data in real time through multiple acoustic sensors deployed in urban functional areas. The acoustic environment data includes sound pressure level, frequency spectrum, timestamp and geographical location information. The data processing module, connected to the data acquisition module, is used to preprocess the acoustic environment data, including data cleaning, noise reduction and normalization, and to extract acoustic environment feature parameters, including equivalent continuous A-weighted sound level, maximum sound level, duration of sound events and spectral characteristics. A multi-source data fusion unit is used to fuse auxiliary data from external systems, including traffic flow data, meteorological data, urban planning data, and real-time activity data. The dynamic evaluation module, connected to the data processing module and the multi-source data fusion unit, is used to perform dynamic evaluation based on acoustic environment characteristic parameters and auxiliary data using a pre-trained acoustic environment evaluation model, and generate an acoustic environment evaluation index. The acoustic environment evaluation model adopts a deep learning-based dynamic adaptive model, which adjusts the model parameters according to real-time data through an online learning algorithm. A management decision module, connected to the dynamic evaluation module, is used to perform dynamic management operations based on the acoustic environment evaluation index and a pre-set management strategy library. The management decision module includes: The early warning unit is used to generate graded early warning information when the sound environment evaluation index exceeds a preset threshold. Control unit, used to automatically control noise source devices via an Internet of Things (IoT) platform; The strategy optimization unit is used to optimize management strategies based on historical management performance data and through reinforcement learning algorithms. The visualization and interaction module is connected to the management decision-making module and the dynamic evaluation module. It is used to display the acoustic environment evaluation results, management operation status and prediction trends in real time, and provides a parameter configuration interface.

2. The urban functional area acoustic environment dynamic evaluation and management system according to claim 1, characterized in that: The data acquisition module is used to collect real-time acoustic environment data through multiple acoustic sensors deployed in urban functional areas, including: Configure the sampling parameters and network connection parameters of the sound sensor according to the sound environment monitoring needs of urban functional areas; The ambient sound waves are continuously captured by the initialized sound sensor, and analog acoustic signals are generated. The analog acoustic signal is converted into a digital acoustic signal using an analog-to-digital converter; Perform time-domain and frequency-domain analysis on digital acoustic signals to calculate sound pressure level and frequency spectrum; Add timestamps and geographic location information to the calculated sound pressure level and frequency spectrum to form structured acoustic environment data; Structured acoustic environment data is transmitted to the data processing module in real time via a communication interface.

3. The urban functional area acoustic environment dynamic evaluation and management system according to claim 1, characterized in that: The data processing module is used to preprocess the acoustic environment data, including data cleaning, noise reduction, and normalization, and to extract acoustic environment feature parameters, including: Receive ambient acoustic data from the data acquisition module; Data cleaning is performed on the acoustic environment data to remove invalid data and outliers, resulting in cleaned acoustic environment data. The cleaned acoustic environment data is denoised to eliminate environmental noise interference, resulting in denoised acoustic environment data. The denoised acoustic environment data is normalized to standardize the data to a preset range, resulting in normalized acoustic environment data. Based on the normalized acoustic environment data, acoustic environment feature parameters are extracted, including equivalent continuous A-weighted sound level, maximum sound level, duration of sound events, and spectral characteristics.

4. The urban functional area acoustic environment dynamic evaluation and management system according to claim 1, characterized in that: The multi-source data fusion unit is used to fuse auxiliary data from external systems, including: Receive auxiliary data from external systems, including traffic flow data, meteorological data, urban planning data, and real-time activity data; The received auxiliary data is preprocessed, including data cleaning and format standardization, to remove invalid data and unify the data format, resulting in preprocessed auxiliary data. The preprocessed auxiliary data is spatiotemporally aligned with the acoustic environment data from the data processing module, and matched based on timestamps and geographic location information to obtain the aligned dataset. The aligned dataset is subjected to data fusion processing to integrate acoustic environment data and auxiliary data to generate multi-source fused data. The multi-source fusion data is output to the dynamic evaluation module for use.

5. The urban functional area acoustic environment dynamic evaluation and management system according to claim 1, characterized in that: The dynamic evaluation module is used to perform dynamic evaluation based on acoustic environment feature parameters and auxiliary data using a pre-trained acoustic environment evaluation model, generating an acoustic environment evaluation index. The acoustic environment evaluation model employs a deep learning-based dynamic adaptive model, adjusting model parameters through an online learning algorithm based on real-time data, including: It receives acoustic environment characteristic parameters from the data processing module and auxiliary data from the multi-source data fusion unit; The received acoustic environment characteristic parameters and auxiliary data are processed to construct feature vectors, integrating parameters of different types and dimensions into a unified multidimensional feature vector; The constructed multidimensional feature vector is input into the pre-trained acoustic environment evaluation model, which contains a multi-layer neural network structure to simulate the complex nonlinear relationship between acoustic environment quality and multi-source features. The multi-dimensional feature vector of the input is forward propagated through a multi-layer neural network structure to extract high-level abstract features layer by layer. Regression analysis is performed based on the high-level abstract features of the neural network output layer to generate an initial acoustic environment evaluation index. The deviation between the new input multidimensional feature vector and the model training data distribution is monitored in real time, and the model update mechanism is triggered when the deviation exceeds the preset tolerance. By using an online learning algorithm, and utilizing real-time samples composed of newly arrived acoustic environment feature parameters and auxiliary data, the loss function between the model's predicted output and the expected output is calculated. Based on the loss function, the gradient backpropagation algorithm is used to dynamically adjust the connection weights and bias parameters of the multilayer neural network in the acoustic environment evaluation model. The acoustic environment assessment model is updated using the adjusted model parameters to generate an optimized model with environmental adaptability. The updated acoustic environment assessment model is used to recalculate the multidimensional feature vectors input in real time to generate an optimized acoustic environment assessment index. The optimized acoustic environment evaluation index and historical evaluation data are output to the management decision-making module, while the data generated in this evaluation process are stored in the model training database.

6. The urban functional area acoustic environment dynamic evaluation and management system according to claim 1, characterized in that: The management decision module is used to perform dynamic management operations based on the acoustic environment evaluation index and a pre-set management strategy library, including: Receive the acoustic environment evaluation index from the dynamic evaluation module; The received acoustic environment evaluation index is compared with the threshold conditions in the preset management strategy library to determine the current acoustic environment status level. Based on the determined acoustic environment status level, the corresponding set of management strategies is retrieved from the management strategy library; Based on the retrieved set of management strategies and combined with the real-time acoustic environment evaluation index, a dynamic management operation instruction set is generated. Execute the dynamic management operation instruction set, including: The early warning unit generates tiered early warning information based on the early warning conditions in the dynamic management operation instruction set and sends it to relevant terminals. The operating parameters of the noise source device are automatically adjusted through the Internet of Things platform by the control unit, based on the control logic in the dynamic management operation instruction set. The strategy optimization unit uses reinforcement learning algorithms to iteratively optimize and update the strategies in the management strategy library based on historical management effect data and the current sound environment status level.

7. The urban functional area acoustic environment dynamic evaluation and management system according to claim 1, characterized in that: The visualization and interaction module is used to display the acoustic environment assessment results, management operation status, and predicted trends in real time, and provides a parameter configuration interface, including: Receives acoustic environment evaluation index from the dynamic evaluation module and management operation status data from the management decision module; The received acoustic environment evaluation index and management operation status data are processed for visualization data conversion to generate graphical representation elements; Based on graphical representation elements, a real-time acoustic environment evaluation chart and a management operation status display panel are constructed. Integrate the predicted trend data from the dynamic evaluation module and generate a sound environment quality prediction curve through time series analysis; The real-time acoustic environment evaluation chart, management operation status display panel and acoustic environment quality prediction curve are combined into a comprehensive visual interface. Render the comprehensive visualization interface and output dynamic visualization content on the display device; Provides a parameter configuration interface to receive user input of management strategy parameters and display preference parameters; Update the preset management strategy library in the management decision module based on the management strategy parameters input by the user; Adjust the display style and layout of the integrated visualization interface based on the display preference parameters input by the user.

Images