Regional noise monitoring and analysis method and system

Abstract

The present invention relates to a monitoring and analysis method and system for regional noise. The method: partition the target area according to the acoustic environment function standard, obtain the acoustic environment function area division data and determine the monitoring point distribution matrix; collect noise data and environmental parameter data, and align the timestamps to obtain a multi-dimensional associated data set; perform spectrum analysis and data compensation to obtain noise feature vectors, and at the same time make an over-standard judgment to generate audio recording data; input a voiceprint recognition model for classification to obtain a noise source type identifier, and perform positioning calculation to obtain the spatial coordinates of the noise source; perform spatiotemporal distribution statistics to obtain the functional area noise compliance rate and calculate the contribution of various noise sources; perform graded early warning judgments to generate alarm information and noise pollution prevention and control instructions. The implementation of the present invention realizes the full process automated management of regional noise from monitoring to early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of noise monitoring, and in particular to a method and system for monitoring and analyzing regional noise. Background Art

[0002] Traditional noise monitoring methods mainly rely on manual handheld devices for sampling and detection, which has problems such as low monitoring frequency, small coverage, and poor data continuity. At the same time, due to the temporary and decentralized characteristics of noise pollution, it is difficult to obtain evidence and the supervision efficiency is low. Although the existing automatic monitoring system can achieve 24-hour continuous monitoring, it is still relatively lacking in technical means in noise source identification, spatial positioning, and pollution source tracing. At the same time, due to the lack of scientific noise classification and identification and precise positioning methods, it is impossible to accurately determine the source and degree of noise pollution, which seriously affects the pertinence and effectiveness of noise pollution prevention and control. Summary of the invention

[0003] The main purpose of the present invention is to provide a method and system for monitoring and analyzing regional noise, which realizes the automatic management of the whole process from monitoring to early warning of regional noise.

[0004] To achieve the above object, the present invention provides a method for monitoring and analyzing regional noise, comprising the following steps:

[0005] The target area is zoned according to the acoustic environment function standard, the acoustic environment function zone division data is obtained and the monitoring point distribution matrix is ​​determined;

[0006] Based on the monitoring point distribution matrix, noise data and environmental parameter data are collected and timestamps are aligned to obtain a multi-dimensional associated data set;

[0007] Performing spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain a noise feature vector, and simultaneously performing an over-standard determination to generate audio recording data;

[0008] Inputting the audio recording data into a voiceprint recognition model for classification to obtain a noise source type identifier, and performing positioning calculation on the noise source type identifier to obtain the noise source spatial coordinates;

[0009] Performing spatiotemporal distribution statistics according to the noise feature vector and the noise source type identifier to obtain the noise compliance rate of the functional area, and calculating the contribution of various noise sources based on the spatial coordinates of the noise source;

[0010] A graded early warning judgment is made on the noise compliance rate of the functional area, an alarm message is generated, and a noise pollution prevention and control instruction is generated according to the contribution of each type of noise source.

[0011] The present invention also provides a regional noise monitoring and analysis system, comprising:

[0012] The zoning module is used to partition the target area according to the acoustic environment function standard, obtain the acoustic environment function area division data and determine the monitoring point distribution matrix;

[0013] A collection module, used to collect noise data and environmental parameter data based on the monitoring point distribution matrix, and perform time stamp alignment to obtain a multi-dimensional associated data set;

[0014] A determination module, used for performing spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain a noise feature vector, and at the same time, performing an over-standard determination to generate audio recording data;

[0015] A classification module, used for inputting the audio recording data into the voiceprint recognition model for classification, obtaining a noise source type identifier, and performing positioning calculation on the noise source type identifier to obtain the noise source spatial coordinates;

[0016] A calculation module, used to perform spatiotemporal distribution statistics according to the noise feature vector and the noise source type identifier, obtain the functional area noise compliance rate, and calculate the contribution of various noise sources based on the noise source spatial coordinates;

[0017] The generation module is used to make a graded early warning judgment on the noise compliance rate of the functional area, generate alarm information, and generate noise pollution prevention and control instructions according to the contribution of each type of noise source.

[0018] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0019] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0020] In summary, the technical solution provided by the present invention ensures the representativeness and accuracy of the monitoring data by establishing a scientific functional area division system and monitoring point layout method, combined with multi-dimensional data collection and timestamp alignment technology. The deep learning voiceprint recognition model is used to achieve accurate classification of five major types of noise sources; the three-dimensional spatial positioning of the noise source is achieved through the microphone array phase matching algorithm. Based on the spatiotemporal distribution statistics of noise feature vectors and noise source type identifiers, a quantitative evaluation method for the functional area noise compliance rate and noise source contribution is constructed. Through hierarchical early warning judgment and intelligent governance strategy matching, the whole process from monitoring, early warning to governance is automated. By establishing a noise event feature library and an intelligent analysis system, a scientific decision-making basis is provided for noise pollution prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the steps of a method for monitoring and analyzing regional noise in one embodiment of the present invention;

[0022] Figure 2 It is a structural block diagram of a monitoring and analysis system for regional noise in one embodiment of the present invention;

[0023] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0024] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0026] Reference Figure 1 This embodiment provides a method for monitoring and analyzing regional noise, comprising the following steps:

[0027] S1, zoning the target area according to the acoustic environment function standard, obtaining the acoustic environment function zone division data and determining the monitoring point distribution matrix;

[0028] Among them, the land use function analysis of the target area is carried out to clarify the specific functional types of different areas. Based on the analysis of data on different uses in the region, including the distribution of medical and health, culture and education, residential housing, commercial finance, industrial production and traffic arteries. By extracting the features of these functional data, the functional type data of the region is obtained. The regional functional type data is input into the zoning standard model, and the functional areas are marked according to the uses of different types of areas. The rehabilitation area is marked as Class 0 area, the residential area and medical and health area are marked as Class 1 area, the commercial and financial area is marked as Class 2 area, the industrial production area is marked as Class 3 area, and the areas on both sides of the traffic arteries are marked as Class 4 areas, and the acoustic environment functional area division data of the target area is obtained. Based on the functional area data, the number of permanent residents in the target area is read, and the urban scale level is determined by the population density calculation model. When the population in the target area is less than 500,000, the number of monitoring points is set to 7; when the population is between 500,000 and 1 million, the number of monitoring points is set to 10; for medium-sized cities with a population of 1 million to 5 million, the number of monitoring points is set to 15; when the population exceeds 5 million, the number of monitoring points is set to 20 to ensure sufficient coverage of noise monitoring in large-scale cities. Through this method, the minimum number of monitoring points in the target area is obtained to meet the effective monitoring needs of urban noise conditions. In order to optimize the layout of monitoring points, equivalent sound level measurements are performed on each functional area in the acoustic environment functional area division data to obtain the sound level data set in the functional area. The specific implementation process of the functional area equivalent sound level measurement is as follows: different sampling grid spacings are set according to the size of the functional area. A 50-meter grid is set for an area less than 1 square kilometer, a 100-meter grid is set for 1-5 square kilometers, and a 200-meter grid is set for an area greater than 5 square kilometers. A microphone is set at each grid intersection, 1.2 meters above the ground, away from reflective surfaces and fixed sound sources for more than 3 meters, and continuous sampling is carried out for 24 hours, with a sampling frequency of once per second. After the collection is completed, the equivalent sound level calculation formula for each measuring point is calculated: Leq = 10lg[1 / T∑(ti×10(Li / 10))], where Leq is the equivalent sound level (dB(A)), T is the total measurement time (s), ti is the measurement time interval (s), and Li is the sound level within the time interval (dB(A)). The calculation results are processed for outliers. When the difference between the sound level of the measuring point and the average value of the surrounding 8 adjacent points exceeds 10dB(A), it is necessary to re-measure. The equivalent sound levels of all grid points are arithmetic averaged to obtain the average equivalent sound level value of the functional area. Based on the sound level data set, the average equivalent sound level of each functional area is calculated, and this average equivalent sound level is used as the benchmark value for subsequent difference analysis. On this basis, the sound level difference between each measuring point and the reference value is calculated, and the measuring points whose sound level difference is less than the preset target value are selected as candidate monitoring points.Through the screening process, the monitoring points that meet the monitoring requirements are initially determined. The environmental conditions of the candidate monitoring points are evaluated to determine whether they meet the conditions for installing monitoring equipment. Environmental conditions include accessibility, safety, power supply conditions, and stability of data transmission at the measurement point. The specific standards for environmental condition assessment include four dimensions: the accessibility assessment requires that the monitoring point is no more than 100 meters away from the main road, has a 24-hour access for personnel, has a flat ground with a slope of no more than 15 degrees, and has no obstacles around that affect the installation of equipment; the safety assessment requires that the monitoring point is located in a closed area with a fence or wall, has anti-theft alarm and video surveillance coverage, is more than 20 meters away from dangerous sources such as high-voltage lines, has stable geology and no landslide collapse risks, and is more than 5 meters away from buildings; the power supply condition assessment requires that the mains power supply capacity is not less than 2kW, the voltage stability fluctuation does not exceed ±5%, has an independent meter and distribution box, a UPS backup power supply interface is reserved, and the grounding resistance is not greater than 4Ω; the data transmission stability assessment requires that the 4G signal strength is not less than -85dBm, the network delay does not exceed 100ms, the packet loss rate does not exceed 0.1%, there is an optical fiber or Ethernet backup channel, and there is no strong electromagnetic interference source. The four dimensions are weighted respectively: accessibility 0.2, safety 0.3, power supply conditions 0.25, data transmission 0.25, each dimension has a full score of 100, and the weighted average score is calculated. Points with a total score of 85 or more can be used as candidate monitoring points, and then these candidate points are sorted from high to low according to the total score, and the specific layout plan of the monitoring points is finally determined in combination with the minimum number of monitoring points. Through evaluation, the monitoring points that meet all installation conditions are screened out, and the final number allocation is carried out according to the minimum number of monitoring points determined in the previous article to obtain the number distribution table of monitoring points to ensure that the number of monitoring points in each functional area can meet the coverage requirements and monitoring accuracy. The point information in the number distribution table of monitoring points is mapped with the geographic information system (GIS) to generate a monitoring point distribution matrix containing the geographic coordinates of the monitoring points, functional area categories, and monitoring numbers.

[0029] S2, collects noise data and environmental parameter data based on the monitoring point distribution matrix, and aligns the timestamps to obtain a multi-dimensional correlation data set;

[0030] Specifically, all-weather outdoor microphones are arranged based on the coordinates of the points in the monitoring point distribution matrix. These microphones are used to collect sound level data of noise and obtain all-weather original sound level data. Audio parameters are calculated based on the original sound level data to obtain a noise parameter data set containing a variety of noise characteristic information. At the same time, based on each point in the monitoring point distribution matrix, a meteorological monitoring unit is arranged at each point to collect a variety of environmental parameters. These environmental parameters include wind speed, wind direction, temperature, humidity and atmospheric pressure, which have a direct impact on the propagation and change of noise. Through comprehensive monitoring of these meteorological conditions, an environmental parameter data set is obtained, which is combined with noise parameters to analyze the impact of the environment on noise characteristics. The noise parameter data set is verified for data validity to ensure that the collected data has good accuracy and consistency and form a valid data set. The process of data validity verification includes eliminating abnormal data points, eliminating errors generated by sensors, and excluding extreme values ​​caused by changes in environmental conditions. The specific implementation process of data validity verification for the noise parameter data set is as follows: a triple test method is used to eliminate abnormal data points: (1) a range test based on the physical characteristics of noise is used to mark data with instantaneous sound levels exceeding the range of [0dB, 130dB] as abnormal; (2) a statistical test is performed using the 3σ criterion to calculate the mean μ and standard deviation σ of the data within the sampling time window (usually 5 minutes), and data points that deviate from the mean by more than ±3σ are marked as abnormal; (3) a time continuity test is performed to calculate the rate of change of adjacent sampling points. When the rate of change exceeds 30dB / s, the data point is marked as abnormal. Secondly, the following steps are adopted to eliminate the sensor error: (1) Use a static calibrator to calibrate the microphone sensitivity, using a 94dB / 1kHz standard sound source. When the measured value deviates from the standard value by more than ±0.5dB, record the correction coefficient; (2) Perform temperature drift compensation, based on the sensor temperature coefficient (typical value -0.01dB / °C), and perform automatic correction when the temperature changes; (3) Perform frequency response correction, measure the sensor's frequency response curve in the range of 20Hz-20kHz using a 1 / 3 octave band analyzer, and establish a frequency correction function; (4) Perform nonlinear distortion compensation, measure the distortion every 10dB in the range of 70-130dB, and establish a piecewise linear correction model.The following methods are used to deal with extreme values ​​caused by environmental conditions: (1) Wind speed effect: When the wind speed exceeds 5m / s, correction is made based on the corresponding relationship between wind speed and noise increment (for every 1m / s increase in wind speed, the background noise increases by about 1.5dB); (2) Temperature and humidity effect: Data outside the temperature range of -10°C to 40°C and relative humidity of 20% to 90% need to be corrected for atmospheric absorption attenuation, and the correction amount is calculated according to the ISO 9613-1 standard; (3) Atmospheric pressure effect: When the air pressure deviates from the standard atmospheric pressure (101.325kPa) by more than ±5kPa, correction is required according to the proportional relationship between sound pressure and air pressure; (4) Precipitation effect: During precipitation, the pulse noise of raindrops hitting the microphone is identified and eliminated through the precipitation noise feature recognition algorithm (based on spectrum analysis). The data after the above processing is then verified for integrity, requiring the data collection rate to be no less than 95%, and finally an accurate and reliable valid data set is obtained. Based on the valid data set, the spectrum parameter analysis is performed to calculate the sound level values ​​of A-weighting, C-weighting and Z-weighting. A-weighting is the most commonly used noise weighting method, which simulates the human ear's perception of sounds of different frequencies. C-weighting is used to measure medium and low frequency noise, and Z-weighting is used to evaluate the overall spectrum of noise. While performing spectrum analysis on the noise, the time characteristics of the noise are integrated, including the time integral of fast characteristics (F), slow characteristics (S) and impulse characteristics (I). These characteristics represent different aspects of noise changes over time. Through the integration operation, the characteristic data of the sound level is obtained, reflecting the performance of noise at different time scales. The sound level characteristic data and the environmental parameter data set are synchronized in time. Through the timestamp alignment operation, the noise data and the environmental parameter data are compared and analyzed under the same time base to obtain time series correlation data. On the basis of time synchronization, the time series correlation data is checked for data integrity to ensure that there is no omission or damage in the data collection, transmission and storage process, and the integrity verification data is obtained. According to multiple dimensions such as monitoring point number, timestamp, noise characteristics and environmental parameters, the integrity verification data is organized into a multi-dimensional data structure to generate a multi-dimensional associated data set.

[0031] S3, performing spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain the noise feature vector, and at the same time making an over-standard determination to generate audio recording data;

[0032] It should be noted that the multi-dimensional correlation data set is transformed in the time-frequency domain, and the noise data is converted from the time domain to the frequency domain, so that the data analysis is performed under different frequency components to reveal more noise characteristic information. After the time-frequency domain transformation, the sound pressure level spectrum distribution is extracted to obtain the full-spectrum signal characteristics. The full-spectrum signal characteristics can fully reflect the distribution of noise in different frequency ranges. The full-spectrum signal characteristics are processed by segmented filtering, and the noise is divided into different frequency bands to obtain the sub-band sound spectrum data. The sub-band sound spectrum data is corrected for sound level according to three frequency weighting networks: A weighting, C weighting, and Z weighting. A weighting is used to simulate the human ear's perception of noise of different frequencies, C weighting is used to evaluate medium and low-frequency noise, and Z weighting is an unbiased evaluation of the noise spectrum. By performing multiple weighting corrections on these sub-band sound spectrum data, noise characteristic data that is more in line with human ear perception is obtained. On this basis, the time weighting processing of fast characteristics (F), slow characteristics (S), and pulse characteristics (I) is superimposed to better describe the performance of noise under different time characteristics. The time weighting processing reflects the dynamic characteristics of noise changing over time and obtains weighted correction data. The specific implementation process of noise time weighting processing is as follows: The calculation of weighted correction data adopts the exponential time weighting method to perform exponential integration of the sound pressure signal p (t) with different time constants. Among them, p (t) represents the sound pressure value at any time t, and the unit is Pascal (Pa).The fast characteristic (F) uses a time constant τF=125ms, which indicates the fast response time of the system to changes in sound pressure. The calculation formula is LF(t)=10lg[(1 / τF)∫(t-τF→t) p²(ξ)e((ξ-t) / τF)dξ / p0²], where LF(t) represents the fast characteristic sound level (dB) at time t, τF is the fast characteristic time constant (125ms), p0 is the reference sound pressure (20μPa, the minimum sound pressure audible to the human ear), ξ is the integral variable (s), and e((ξ-t) / τF) is an exponential decay function, which reflects the influence of early sound pressure on the current value. The slow characteristic (S) uses a time constant τS=1000ms, which indicates the slow response time of the system to changes in sound pressure. The calculation formula is LS(t)=10 lg[(1 / τS)∫(t-τS→t)p²(ξ)e((ξ-t) / τS)dξ / p0²], where LS(t) represents the slow characteristic sound level at time t (dB); the impulse characteristic (I) adopts an asymmetric exponential function with a rise time constant τI+=35ms and a fall time constant τI-=1500ms, where τI+ represents the response time of the system to a rapid rise in sound pressure, and τI- represents the response time of the system to a drop in sound pressure. The calculation formula is LI(t)=10lg[(1 / τI+)∫(t-τI+→t) p²(ξ)e((ξ-t) / τI+)dξ / p0²] is used for the rising section, LI(t)=10lg[(1 / τI-)∫(t-τI-→t) p²(ξ)e((ξ-t) / τI-)dξ / p0²] is used for the falling section, where LI(t) represents the pulse characteristic sound level (dB) at time t. In actual calculations, a digital processing method is used to convert the continuous time integral into the weighted sum of discrete time series. For discrete signals with a sampling frequency of fs (indicates the number of samples per second, in Hz), the recursive formula for the fast characteristic is LF(n)=αF·LF(n-1)+(1-αF)·20lg(p(n) / p0), where αF=exp(-1 / (fs·τF)) is the fast characteristic weighting coefficient, n represents the sequence number of the discrete time series, LF(n) represents the fast characteristic sound level at the nth sampling point, LF(n-1) represents the fast characteristic sound level at the previous sampling point, and p(n) represents the sound pressure value at the nth sampling point. The slow characteristic and the pulse characteristic use a similar recursive calculation method, except that the time constant is different. The final weighted correction data contains the calculation results of these three time characteristics, and a set of weighted correction data containing three values ​​of LF, LS, and LI (all in dB) is output every second for subsequent noise characteristic analysis and evaluation. The weighted correction data is calibrated and compensated by an electrostatic actuator to generate calibration parameter data. The collected noise data is accurately compensated through a physical calibration device to eliminate the deviation and measurement error of the sensor itself and ensure the accuracy of the data. The noise feature vector is constructed based on the calibration parameter data and weighted correction data.The noise feature vector is a comprehensive expression of the noise data in the monitoring area, including the frequency, intensity, time characteristics and other information of the noise. According to the functional area category of the monitoring point corresponding to the noise feature vector, different exceeding threshold sets are set. In the Class 0 area, the maximum allowable value is set to 50 decibels to ensure noise control in extremely quiet areas such as rehabilitation and convalescent areas; the maximum allowable value of the Class 1 area is 55 decibels, which is suitable for residential and medical and health areas to ensure the quality of life of residents; the Class 2 area is the commercial and financial area, with a maximum allowable value of 60 decibels; the Class 3 area is the industrial production area, with a maximum allowable value of 65 decibels; the Class 4 area is on both sides of the traffic artery, and the maximum allowable value is set to 70 decibels. Through these settings, the noise monitoring of different functional areas is provided with evaluation standards that meet actual needs, and a exceeding threshold set is formed. The noise feature vector is threshold-compared. When the measured sound level exceeds the exceeding threshold of the corresponding functional area, the system will automatically trigger the FLC (lossless compression algorithm) to record and record the exceeding of the noise. This record includes data such as the time when the exceeding occurs, the duration, and the magnitude of the exceeding, forming a complete exceeding record data. The excessive noise record data and the environmental parameter data at the corresponding time are associated and packaged to generate audio record data, including the recording of the excessive noise and the environmental parameters at that time (such as wind speed, temperature, humidity, etc.).

[0033] S4, inputting the audio recording data into the voiceprint recognition model for classification, obtaining a noise source type identifier, and performing positioning calculation on the noise source type identifier to obtain the noise source spatial coordinates;

[0034] Specifically, voiceprint features are extracted from audio recording data, and the collected noise data is divided into five categories: life noise, construction noise, industrial noise, traffic noise, and natural noise. Typical sub-class noises are analyzed, such as knocking sounds in life noise, insect sounds, wind sounds, and thunder sounds in natural noise, to generate audio feature vectors. The audio feature vectors are input into the machine learning training model to obtain the voiceprint recognition model. During the training process, a large amount of labeled audio data is used to supervise the model, and through continuous iteration and optimization, the model can accurately identify different types of noise. In order to ensure that the model has good generalization ability and high precision, the voiceprint recognition model will be used as a benchmark model after it is built, and used to optimize the algorithm of manually labeled audio data, and add new audio to the voiceprint training library, so as to build and continuously expand and improve the noise source type identification library. Through continuous optimization and expansion, the voiceprint recognition model can more accurately classify different types of noise. After completing the noise classification, based on the noise source type identification, the omnidirectional acoustic radar is started to locate the sound source. The omnidirectional acoustic radar samples the sound field of the noise to help determine the location of the sound source. According to the identified noise pollution type, different sound field sampling intervals are selected to more effectively adapt to the characteristics of different noise scenes. The specific implementation of the sound field sampling interval selection and sampling process for different noise types is as follows: the sampling parameters are determined based on the time-frequency characteristics of the noise type. For industrial noise (such as mechanical equipment operation), because it has relatively stable spectral characteristics, the sampling interval is set to 1000ms and the sampling duration is 60s to form a measurement cycle; for construction noise (such as piling, cutting, etc.), due to its sudden and high-frequency characteristics, the sampling interval is set to 100ms and the sampling duration is 30s to capture transient changes; for traffic noise (such as vehicle passing), considering the influence of the Doppler effect, the sampling interval is set to 250ms and the sampling duration is 45s; for life noise (such as crowd activities), because of its strong randomness, the sampling interval is set to 500ms and the sampling duration is 40s; for natural noise (such as wind and rain), due to its slow changes, the sampling interval is set to 2000ms and the sampling duration is 120s. During the sampling process, the segmented sliding window technology was used, with the window length equal to the sampling duration and the overlap rate set to 50% to ensure the continuity of the data. To improve the sampling accuracy, three repeated measurements were performed at each sampling point, and the arithmetic mean was taken. At the same time, the appropriate sampling frequency was selected according to the frequency range of the noise source. The sampling frequency for industrial noise was set to 48kHz, construction noise 96kHz, traffic noise 44.1kHz, life noise 32kHz, and natural noise 22.05kHz to meet the requirements of the Nyquist sampling theorem. Environmental parameters (temperature, humidity, wind speed, and air pressure) were recorded synchronously during the sampling process for later data correction.When the environmental parameters exceed the preset range (temperature <-10℃ or >40℃, relative humidity >90%, wind speed >5m / s), the system automatically adjusts the sampling interval and increases the sampling time by 50% to obtain more data samples for averaging and reduce the impact of environmental factors. The sampled data is processed by an anti-aliasing filter, and the cutoff frequency is set to 0.45 times the sampling frequency to ensure the validity of the sampled signal. The sampled data is quantized, and a 24-bit AD conversion is used with a dynamic range of 120dB to ensure sampling accuracy. The sound field sampling is completed by the microphone array arranged in the area, and the sound pressure data and phase data of each microphone are collected to generate array sampling data. The array sampling data is signal preprocessed to remove redundant and unstable components in the noise data. The preprocessing includes signal filtering and signal amplitude normalization processing, so that all collected signals are compared under the same benchmark to obtain the filtered array signal. The filtered array signal is input into the phase matching positioning algorithm to calculate the time delay and phase difference of the sound wave reaching the adjacent microphones. The specific implementation process of the phase matching positioning algorithm is as follows: First, a spatial geometric array model consisting of M microphones is established, where M represents the total number of microphones (generally 8-16 to ensure positioning accuracy), and the three-dimensional spatial coordinates of each microphone are marked as (xm, ym, zm), where m is the microphone number (m=1,2,...,M), in meters. The spacing d between adjacent microphones is set to 0.1 meters (this spacing is based on the spatial sampling theorem at a sampling frequency of 20kHz). Assuming that the sound source is in the far field (the distance from the sound source to the center of the array is greater than twice the maximum size of the array), the incident sound wave can be approximated as a plane wave, and the spatial direction of the sound source is represented by the azimuth angle θ (the angle with the north direction in the horizontal plane, ranging from 0° to 360°) and the pitch angle φ (the angle with the horizontal plane, ranging from -90° to +90°). The time delay τm (in seconds) of the sound wave propagating from the sound source to the mth microphone is calculated by the formula τm=(xm·sinθ·cosφ + ym·sinθ·sinφ + zm·cosφ) / c, where c is the speed of sound (about 340m / s at standard atmospheric pressure and temperature 20°C). For any two adjacent microphones i and j, the time delay difference Δτij=τi-τj (in seconds) of the received sound waves. In the frequency domain, the phase difference Δφij(ω)=ω·Δτij corresponding to this time delay difference, where ω is the angular frequency (in radians / second), and the calculation range is 0-20kHz. The actual calculation adopts the generalized cross-correlation method (GCC-PHAT), where Xi(ω) and Xj(ω) represent the Fourier transform complex values ​​of the signals received by the two microphones, respectively, and * represents the complex conjugate operation. The cross-correlation function Rij(τ) calculates the correlation between the two signals, and the integration range covers all analysis frequencies.The measured phase difference Δφijmeasured (ω) is matched with the theoretically calculated phase difference Δφijtheoretical (ω, θ, φ) to construct the least squares optimization objective function J (θ, φ). The gradient descent method is used to iteratively solve the optimal azimuth angle θ and elevation angle φ, and the iteration step is 0.1° (a compromise between ensuring accuracy and computational efficiency). The iteration is stopped when the angle change between two adjacent iterations is less than the convergence threshold of 0.01°. The final positioning results include: azimuth angle θ (accuracy is ±1°), elevation angle φ (accuracy is ±2°), and positioning confidence coefficient C. Among them, C=exp(-Jmin / σ²), Jmin is the minimum objective function value after optimization, σ is the standard deviation of the measurement noise (obtained through system calibration), and a C value greater than 0.9 indicates that the positioning result is reliable. Based on these calculation results, the sound source direction estimation equation is established to obtain the azimuth data of the sound source. The relative orientation of the noise source can be accurately inferred through the differences in the propagation of sound waves in the array. Based on the sound source orientation data, the triangulation method is used to establish a sound source depth estimation model. The triangulation method combines the positional relationship of each sensor in the microphone array and the propagation characteristics of the sound wave. By calculating the propagation attenuation characteristics of the sound wave in the air, the distance from the sound source to the center of the microphone array is estimated, and the three-dimensional spatial position data of the sound source is generated, including the specific coordinates of the noise source in the three dimensions of x, y, and z. The generated three-dimensional spatial position data is synchronized and spatially mapped with the image frames collected by the surveillance camera. By jointly analyzing the sound source positioning data and the image frame, the three-dimensional spatial position of the noise source is marked in the image, so that the position of the noise source is intuitively displayed on the surveillance image. A surveillance image containing the spatial coordinates of the noise source is generated.

[0035] S5, perform spatiotemporal distribution statistics based on noise feature vectors and noise source type identifiers to obtain the noise compliance rate of the functional area, and calculate the contribution of various noise sources based on the spatial coordinates of the noise source;

[0036] Among them, the noise feature vector is divided into statistical divisions according to the category of functional area, and the noise information in different categories of areas is classified to obtain the distribution data of noise types. Through this step, the noise types existing in different functional areas and their respective proportions in the area are clarified. The noise type distribution data is associated with the time dimension so as to analyze the noise compliance of each type of functional area during the daytime and nighttime periods. By calculating the energy average of the noise data collected in different time periods, the noise level of each functional area during the daytime and nighttime is obtained. According to these calculation results, it is analyzed whether the noise of the functional area in different time periods meets the standard, and the functional area time period compliance data is formed. The functional area time period compliance data is hierarchically counted, and the noise compliance results are divided into three levels: full-day compliance, daytime compliance and nighttime compliance. The ratio of the number of compliance points in each level to the total number of monitoring points is calculated to obtain the noise compliance rate of the functional area. By calculating the compliance status of different levels, the noise control status of the functional area in different time periods is evaluated. The spatial diffusion analysis of noise pollution is carried out based on the spatial coordinates of the noise source. The noise contour lines are superimposed on the regional map by color code mapping to construct a noise distribution heat map. Through visualization, the diffusion and agglomeration of noise pollution in different regions can be reflected, and the hot spots of noise pollution can be identified, so as to facilitate intervention measures for these key areas. The cumulative sound energy of various noise sources in the noise distribution heat map is spatially superimposed according to the logarithmic relationship, and the sound energy contribution values ​​of different noise sources at each monitoring point are calculated to obtain the sound energy distribution data of the noise source. The specific calculation process of the sound energy contribution value of the noise source is as follows: First, define the monitoring point P (x, y) to be measured in the spatial coordinate system, where x and y represent the east and north coordinates of the point in the plane rectangular coordinate system (unit: meter). Classify and count the noise sources, and use Si to represent the i-th type of noise source (i=1,2,...,N, where N is the total number of noise source types, for example, N=5 means that the noise is divided into five categories: industry, transportation, life, construction and nature). Each type of noise source Si can be further divided into Mi sub-source points sij (j=1,2,...,Mi, Mi represents the number of sub-sources contained in the i-th type of noise source). The sound power level of each sub-source point sij is represented by LWij (unit: dB, representing the radiation capacity of the sound source), which is obtained by measuring the sound pressure level and reverse calculation.In the calculation of sound wave propagation attenuation, three attenuation mechanisms are involved: geometric divergence attenuation ΔL1=20lg(r / r0), where r is the actual distance from the sound source to the receiving point (unit: meter), and r0 is the reference distance (taken as 1 meter); air absorption attenuation ΔL2=αr / 100, where α is the air absorption coefficient (unit: dB / 100m, related to temperature, humidity and frequency, about 1.1dB / 100m at 1kHz frequency under standard atmospheric conditions); ground effect attenuation ΔL3=4.8-(2hm / r)[17+(300 / r)], where hm is the average height of the sound wave propagation path (unit: meter). The total attenuation ΔLij of the sub-sound source sij is the algebraic sum of the three attenuation values. At the monitoring point P (x, y), the sound pressure level contribution LP,ij (unit: dB) of the sub-sound source sij is equal to its sound power level minus the propagation attenuation. Logarithmic energy superposition is performed on all sub-sound sources of the same type to obtain the total sound energy contribution value LEi of this type of noise source. Taking into account the intermittent nature of noise, a time correction coefficient Ti is introduced, and its value range is 0-1 (for example, if a certain type of noise lasts for 30 minutes within 1 hour, then Ti=0.5). Finally, the corrected sound energy contribution value LEicorr is obtained, and the relative contribution Ci is calculated (expressed as a percentage, indicating the degree of influence of this type of noise source on the overall sound environment). The data sampling frequency during the entire calculation process is set to once per second, and the observation period is usually 24 hours to ensure the statistical representativeness of the data. Based on the sound energy distribution data of the noise source, a sound source dominance evaluation system is established. The influence factor of the noise source is obtained by weighted calculation of the sound pressure level and duration of the noise.The calculation of noise source impact factor involves multiple key parameters, as follows: Noise Impact Factor (NIF) is a comprehensive indicator that characterizes the degree of impact of noise sources. It is composed of parameters in three dimensions: The first dimension is the sound pressure level intensity (SPL), using the A-weighted sound pressure level LA as the basic measurement value (unit: dB (A), reflecting the sensitivity of the human ear to sounds of different frequencies), the sound pressure level weight coefficient Ws = 10 [(LA-L0) / 10], where LA is the measured A-weighted sound pressure level, and L0 is the benchmark sound level value of the functional area (the benchmark values ​​for different functional areas are: 50dB for Class 0 areas for special quiet areas, 55dB for Class 1 areas for residential areas, 60dB for Class 2 areas for commercial areas, 65dB for Class 3 areas for industrial areas, and 70dB for Class 4 areas for traffic arterial areas); the second dimension is Duration (T) uses the normalized time weight coefficient Wt=(T / T0)0.5, where T is the duration of a single noise event (in seconds, the time period from the start to the end of the noise), T0 is the reference time (3600 seconds, or 1 hour, is taken as the standard reference period), and the exponent 0.5 is a time response characteristic parameter determined based on the Weber-Fechner law (which describes the logarithmic relationship between human perception and physical stimulation); the third dimension is the frequency of occurrence (F), using the frequency weight coefficient Wf=lg(1+F / F0), where F is the total number of occurrences of this type of noise within 24 hours (the number is an integer), and F0 is the base frequency (the value is 24, indicating a standard frequency of once an hour on average). The three weight coefficients are multiplied to obtain the noise impact factor: NIF=Ws×Wt×Wf=10[(LA-L0) / 10]×(T / 3600)0.5×lg(1+F / 24). In order to make different noise sources comparable, the standardized noise impact factor SNIF=NIF / NIFmax×100 is introduced, where NIFmax is the maximum value of NIF among all noise sources involved in the evaluation. The value range of SNIF is 0-100 (dimensionless). The larger the value, the more significant the impact of the noise source. Finally, the normalization coefficient K is introduced to adjust the SNIF value to the range of 0-100. The K value is determined by statistical analysis of measured data. The impact factor reflects the degree of influence of a certain type of noise source on the regional sound environment over a period of time and is the core indicator for measuring the impact of the noise source. In order to comprehensively evaluate the impact of the noise source, the impact factor is accumulated over time to obtain the spatiotemporal characteristic vector of the noise source. Based on the spatiotemporal characteristic vector of the noise source, the proportion of the impact of various noise sources on the regional sound environment quality is calculated. By establishing a contribution scoring model, various noise sources are quantitatively weighted to obtain the contribution of each type of noise source. The contribution scoring model comprehensively considers multiple factors such as the sound pressure level, duration, and spatial distribution of the noise source to quantify the degree of influence of various noise sources on the regional sound environment. The process of establishing the noise source contribution scoring model involves the calculation of indicators in four dimensions.The first dimension is the sound pressure level index (SI), which is calculated as SI=Σ(LAeq,i-L0)×ti / T, where LAeq,i represents the equivalent sound level of the ith statistical period (unit: dB(A), reflecting the average level of sound energy during the period), L0 is the noise limit of the functional area (unit: dB(A), the limits for different functional areas are 50dB for Class 0 area, 55dB for Class 1 area, 60dB for Class 2 area, 65dB for Class 3 area, and 70dB for Class 4 area), ti is the duration of exceeding the standard (unit: hours, recording the cumulative time when the sound level exceeds the limit), and T is the total statistical time (24 hours is taken as the evaluation period). The second dimension is the temporal index (TI), which is calculated as TI=(Σ(Ldn,i×wi)) / N, where Ldn,i is the day and night equivalent sound level of the ith monitoring point (unit: dB(A), a comprehensive value considering the difference in day and night sound levels), wi is the time period weight coefficient (1.0 for the daytime period 07:00-22:00, and 1.4 for the nighttime period 22:00-07:00 the next day, reflecting the aggravation of the noise impact at night), and N is the total number of monitoring points. The third dimension is the spatial distribution index (DI), which is calculated as DI=(Sa / Sr)×Ka, where Sa is the area affected by the noise exceeding the standard (unit: square meters, the area of ​​the area where the noise exceeds the limit), Sr is the total area of ​​the evaluation area (unit: square meters), and Ka is the area type correction coefficient (2.0 for Class 0 areas, 1.6 for Class 1 areas, 1.2 for Class 2 areas, 1.0 for Class 3 areas, and 0.8 for Class 4 areas, reflecting the sensitivity of different functional areas). The fourth dimension is the sensitivity index (RI, Receptor Index), which is calculated as RI=Σ(Pi×Ci) / Pt, where Pi is the number of people affected by the i-th sensitive target (unit: person), Ci is the sensitivity level coefficient (2.0 for hospitals, 1.8 for schools, 1.5 for residential areas, 1.2 for office areas, and 1.0 for others, reflecting the sensitivity of different types of receptors), and Pt is the total population of the region. The final comprehensive contribution score (CS, Contribution Score) is calculated by the weighted average method: CS=α×SI+β×TI+γ×DI+δ×RI, where the weight coefficients α=0.4, β=0.3, γ=0.2, and δ=0.1 (the relative importance of each dimension is determined by the hierarchical analysis method). The calculation formula for the standardized contribution score (NCS) is NCS=CS / CSmax×100, where CSmax is the maximum CS value of all evaluated noise sources. The NCS score range is 0-100 and is used for the graded evaluation of noise sources.

[0037] S6, conducts graded early warning judgment on the noise compliance rate of functional areas, generates alarm information, and generates noise pollution prevention and control instructions based on the contribution of various noise sources.

[0038] Specifically, three levels of alarm thresholds are set based on the noise compliance rate of functional areas to effectively evaluate the degree of noise pollution. Combined with the night compliance rate target of the acoustic environment functional area, areas with a compliance rate greater than 85% are marked as mild warnings, indicating that although the acoustic environment is affected to a certain extent, it is still relatively acceptable overall; areas with a compliance rate between 75% and 85% are marked as moderate warnings, indicating that there is a significant noise impact in the area and close attention is required; and areas with a compliance rate less than 75% are marked as severe warnings, which means that noise pollution has had a serious impact on the regional acoustic environment and emergency intervention measures are required. Through the early warning mechanism, clear hierarchical early warning data are obtained. Abnormal pattern analysis is performed on the hierarchical early warning data and the contribution of noise sources, and abnormal noise events are detected for different functional area types. By analyzing the characteristics and contribution of noise sources, noise behaviors that do not conform to the normal pattern are identified, such as high-intensity industrial noise that suddenly appears at night in residential areas. For each detected abnormal event, its warning level, noise event type, time of occurrence, and area involved are entered into the noise event database to form noise event feature data. The noise event characteristic data is input into the exceeding standard alarm system to realize the real-time alarm and processing of the noise exceeding standard. In this system, the monitored noise decibel value is displayed in real time through the LED display screen, and the green, yellow and red three-color alarm lights are triggered according to different warning levels to intuitively reflect the severity of noise pollution. When the noise exceeds the standard, the green light indicates a slight exceeding standard, the yellow light indicates a moderate exceeding standard, and the red light indicates a serious exceeding standard. At the same time, the alarm information is pushed to the mobile terminal of the relevant management personnel synchronously to ensure that the noise exceeding standard can be quickly known at any time and place and the alarm information can be obtained. Based on the alarm information, multiple data sources are integrated, including audio recording data, monitoring video, noise exceeding standard data, and environmental parameter data. By associating these data with timestamps, it is ensured that all relevant data can be analyzed and compared synchronously when the noise exceeds the standard. For example, the specific type of noise is confirmed by audio recording data, the monitoring video helps to lock the specific location of the noise source and its behavioral characteristics, and the environmental parameter data is used to evaluate the impact of external conditions on noise propagation. Based on the analysis of noise events, the corresponding governance strategies are automatically matched for different types of noise sources. Noise control strategies mainly include three technical measures: noise source isolation. For sources that are likely to cause significant noise pollution, such as industrial equipment or construction sites, noise leakage can be reduced by setting up sound barriers or improving the sound insulation of equipment; propagation path blocking, for example, by setting up sound barriers or vegetation isolation belts on the noise propagation path to reduce the spread of noise in the area; sensitive point protection. For noise-sensitive locations, such as schools, hospitals or residential areas, local noise reduction measures are taken, such as installing soundproof windows or strengthening the sound insulation of the indoor environment.By automatically matching these control measures, the impact of noise on the surrounding environment and sensitive groups can be effectively reduced, and specific noise pollution prevention and control instructions can be generated.

[0039] In one example, the target area is zoned according to the acoustic environment function standard to obtain acoustic environment function zone division data and determine the monitoring point distribution matrix, including: land use function analysis of the target area, feature extraction based on the distribution data of medical care, culture and education, residential housing, commercial finance, industrial production, and traffic arteries, to obtain regional function type data; the regional function type data is input into the zoning standard model, the rehabilitation area is marked as Class 0 area, the residential housing and medical care area are marked as Class 1 area, the commercial and financial area is marked as Class 2 area, the industrial production area is marked as Class 3 area, and the area on both sides of the traffic arteries is marked as Class 4 area, to obtain acoustic environment function zone division data; based on the acoustic environment function zone division data, the number of permanent residents in the target area is read, and the city scale level is determined by the population density calculation model. When the population is less than 500,000, the number of monitoring points is set to 7, and when the population is 500,000-1 million, the number of monitoring points is set to 1. 10, when the population is 1-5 million, the number of monitoring points is set to 15, and when the population is greater than 5 million, the number of monitoring points is set to 20, to obtain the minimum number of monitoring points; perform equivalent sound level measurement on each functional area in the sound environment functional area division data to obtain a functional area sound level data set, and calculate the average equivalent sound level of each functional area based on the functional area sound level data set, take the average equivalent sound level as the benchmark value, calculate the sound level difference between each measuring point and the benchmark value, and select the measuring points with a sound level difference less than the preset target value as candidate monitoring points; evaluate the environmental conditions of the candidate monitoring points to obtain monitoring points that meet the installation conditions, and allocate the number of monitoring points that meet the installation conditions according to the minimum number of monitoring points to obtain a monitoring point quantity distribution table; coordinate the point information in the monitoring point quantity distribution table with the geographic information system to generate a monitoring point distribution matrix containing the geographic coordinates of the monitoring points, functional area categories, and monitoring numbers.

[0040] In this example, different land use data in the region are collected and analyzed, including the distribution data of medical and health, culture and education, residential housing, commercial finance, industrial production, and transportation trunk lines. These data are input into the feature extraction system, and different types of land use are quantitatively analyzed through feature extraction to generate regional functional type data. The feature extraction process is actually to identify and classify the main use of each sub-region based on big data, such as identifying hospitals and clinics as medical and health, schools and educational facilities as cultural and educational, and residential areas as residential housing, etc. Through this analysis process, the land use type in each plot and region is clearly labeled, and a regional functional type data set is generated. The regional functional type data is input into the zoning standard model to divide the acoustic environment functional zones. According to the noise sensitivity of different land uses, the areas are labeled and classified: rehabilitation and convalescent areas are very sensitive to environmental noise and need to be marked as Class 0 areas; residential and medical and health areas also require a quiet environment and are marked as Class 1 areas; commercial and financial areas are relatively less sensitive to noise and are marked as Class 2 areas; the noise in industrial production areas is relatively large, so they are marked as Class 3 areas; the areas on both sides of traffic arteries usually have more significant noise and are marked as Class 4 areas. Through the classification and labeling method, the division data of different acoustic environment functional zones in a region is obtained. Based on the acoustic environment functional zone division data, the number of permanent residents in the target area is read to determine the city scale level. Population size is a key factor in judging the size of a city and the number of monitoring points to be deployed. Use the population density calculation model to determine the city scale of the region. When the population size 500,000, set the number of monitoring points to 7; when When , the number of monitoring points is set to 10; when When , the number of monitoring points is set to 15; when When the number of monitoring points is set to 20, the minimum number of monitoring points required to be deployed in the area is obtained according to the scale of the city. The equivalent sound level of each sound environment functional zone is measured to obtain the sound level data set of each functional zone. For the selected candidate monitoring points, the environmental conditions are evaluated to ensure that these monitoring points meet the conditions for installing monitoring equipment. The evaluation of environmental conditions includes the accessibility, safety, power supply conditions and stability of data transmission of the measurement point. Through this process, the monitoring points that meet the installation conditions are screened out. According to the minimum number of monitoring points, the number of monitoring points that meet the conditions is reasonably allocated to obtain the number distribution table of monitoring points. The point information in the number distribution table of monitoring points is mapped with the geographic information system. Through the geographic information system, the geographical coordinates of each monitoring point are matched with the functional area category and monitoring number to which it belongs, and a monitoring point distribution matrix is ​​generated, which includes the geographical coordinates of the monitoring point, the functional area category and the monitoring number.

[0041] In one example, noise data and environmental parameter data are collected based on a monitoring point distribution matrix, and timestamp alignment is performed to obtain a multi-dimensional associated data set, including: sound level data of an all-weather outdoor microphone is collected according to the point coordinates in the monitoring point distribution matrix to obtain original sound level data, and audio parameters are calculated based on the original sound level data to obtain a noise parameter data set; wind speed, wind direction, temperature, humidity, and atmospheric pressure data are collected from the meteorological monitoring unit of each point in the monitoring point distribution matrix to obtain an environmental parameter data set, and data validity verification is performed on the noise parameter data set to obtain a valid data set; spectrum parameter analysis is performed based on the valid data set to calculate A-weighted, C-weighted, and Z-weighted sound level values, and time integration of fast characteristics F, slow characteristics S, and pulse characteristics I is performed simultaneously to obtain sound level feature data; time synchronization is performed on the sound level feature data and the environmental parameter data set to obtain time series associated data, and data integrity check is performed on the time series associated data to obtain integrity verification data; according to the monitoring point number, timestamp, noise characteristics, and environmental parameters, the integrity verification data is organized into a multi-dimensional data structure to generate a multi-dimensional associated data set.

[0042] In this example, the sound level data of the all-weather outdoor microphones are collected according to the point coordinates in the monitoring point distribution matrix, and the microphones are deployed at the specific locations of the monitoring points. After the microphones are installed, the noise in the target area is continuously monitored around the clock, and the noise data collected is the original sound level data. These original sound level data are a direct reflection of the regional acoustic environment, recording the sound pressure level at each moment, that is, the intensity of the sound. The audio parameters of the original sound level data are calculated to generate a noise parameter data set. The original sound level data is obtained through the microphone and is the sound pressure level. Over time function. At the same time, the meteorological monitoring unit at each monitoring point collects environmental parameter data, including wind speed, wind direction, temperature, humidity, and atmospheric pressure. These meteorological conditions have a significant impact on the propagation and attenuation of noise, and are combined with noise data for analysis. For example, wind speed and wind direction have a direct impact on the propagation speed and direction of sound, while temperature and humidity affect the attenuation coefficient of sound waves. By collecting meteorological parameter data, a complete environmental parameter data set is obtained. The noise parameter data set is verified for data validity to ensure the accuracy and reliability of the data. During the data collection process, the sensor is subject to various environmental interferences or failures, resulting in abnormal or invalid data collected. Through statistical analysis and anomaly detection methods, extreme values ​​and invalid data are eliminated. For example, a threshold range is set based on statistical methods. Detect abnormal data points. If the sound pressure value at a certain measurement point If the data exceeds the range, it is considered abnormal data and is removed. Through validity verification, a reliable and valid data set is obtained. Based on the valid data set, spectrum parameter analysis is performed to calculate the sound level values ​​under different frequency weightings, including A weighting, C weighting, and Z weighting. The sound level feature data is synchronized with the environmental parameter data set to obtain time series correlation data. For example, if the timestamp of the noise data is , the timestamp of the environmental parameter data is ,if (satisfying a small error range), the two data are considered to be synchronized, and they are combined to form time series correlation data. Perform data integrity checks on time series correlation data to ensure the consistency and completeness of the entire data set. For example, check whether there are corresponding noise data and environmental data for each time period. If data is missing in some time periods, compensate for it through interpolation methods. The result of the data integrity check is an integrity verification data set, which contains all verified and compensated valid data to ensure that there will be no deviation in the analysis results due to missing data in subsequent analysis. According to the dimensions of monitoring point number, timestamp, noise characteristics and environmental parameters, the integrity verification data is organized into a multi-dimensional data structure to generate a multi-dimensional correlation data set.

[0043] In one example, spectrum analysis and data compensation are performed on a multi-dimensional associated data set to obtain a noise feature vector, and an over-standard judgment is performed to generate audio recording data, including: performing time-frequency domain transformation and sound pressure level spectrum distribution extraction on the multi-dimensional associated data set to obtain full-spectrum signal features, and performing segmented filtering processing on the full-spectrum signal features to obtain frequency-band sound spectrum data; performing sound level correction on the frequency-band sound spectrum data according to three frequency weighting networks of A weighting, C weighting, and Z weighting, and superimposing time weighting processing of fast characteristic F, slow characteristic S, and pulse characteristic I to obtain weighted correction data; calibrating and compensating the weighted correction data through an electrostatic actuator to generate calibration parameter data, and based on the calibration parameter data The noise characteristic vector is constructed by using the noise characteristic vector and the weighted correction data. According to the functional area category of the monitoring point corresponding to the noise characteristic vector, the maximum allowable value of the Class 0 area is set to 50 decibels, the maximum allowable value of the Class 1 area is set to 55 decibels, the maximum allowable value of the Class 2 area is set to 60 decibels, the maximum allowable value of the Class 3 area is set to 65 decibels, and the maximum allowable value of the Class 4 area is set to 70 decibels, and the exceeding threshold set is obtained. The noise characteristic vector is threshold-matched. When the measured sound level exceeds the corresponding functional area threshold in the exceeding threshold set, the FLC lossless compression algorithm is triggered to record, and the exceeding time, duration, and magnitude are marked to obtain the exceeding record data. The exceeding record data and the environmental parameter data at the corresponding time are associated and packaged to generate audio recording data.

[0044] In this example, a time-frequency domain transformation is performed on a multi-dimensional correlation data set to reveal its characteristics at different frequencies. The time-frequency domain transformation uses fast Fourier transform to convert the original time domain signal into the frequency domain to obtain the amplitude and phase of each frequency component. Suppose the original time domain signal is , whose Fourier transform is:

[0045] ;

[0046] in, Indicates that the signal is at frequency The amplitude of is an imaginary unit, is the time. Then we get the frequency domain representation After that, the sound pressure level spectrum distribution is calculated to obtain the full spectrum signal characteristics. The sound pressure level (SPL) calculation formula is:

[0047] ;

[0048] in, Indicates frequency The sound pressure level at For frequency The sound pressure at is the reference sound pressure, which is generally taken as Through this process, the full spectrum signal characteristics are obtained to describe the noise intensity distribution at different frequencies. The full spectrum signal characteristics are processed by segmented filtering, and the spectrum is divided into multiple frequency bands to obtain frequency spectrum data. For example, the spectrum is divided into different frequency bands such as low frequency (20-200Hz), medium frequency (200-2000Hz) and high frequency (2000-20000Hz). Indicates The sound pressure level of each frequency band is divided into multiple frequency bands through the processing of band-pass filters to obtain the frequency spectrum data. The sound level of the sub-band sound spectrum data is corrected according to the three frequency weighting networks of A weighting, C weighting and Z weighting. A weighting simulates the sensitivity of the human ear to noise of different frequencies at medium sound pressure levels, C weighting is used for noise evaluation at higher sound pressure levels, and Weighted A does not add weights, and the entire spectrum is balanced. The correction of A-weighted sound level is expressed as:

[0049] ;

[0050] in, For the The A-weighted sound level in each frequency band, For the frequency Similarly, the correction data of C-weighted and Z-weighted are obtained, which are and After frequency weighting correction, the time characteristic processing is superimposed to obtain the weighted data of fast characteristic (F), slow characteristic (S) and pulse characteristic (I). The fast characteristic and slow characteristic correspond to different response speeds to signal fluctuations, and the pulse characteristic is used to reflect sudden noise. , and Respectively The fast characteristic, slow characteristic and impulse characteristic sound levels of each frequency band are obtained by time-weighted integration of the weighted corrected sound pressure level. For example, the fast characteristic integral is:

[0051] ;

[0052] in, Indicates the time constant of the fast characteristic response, For the The A-weighted sound pressure of each frequency band. Through this step, the corrected data after frequency weighting and time weighting are obtained to describe the noise level under different frequency and time characteristics. After completing the weighting correction, in order to ensure the accuracy of the measurement results, the weighted correction data is calibrated and compensated. The calibration compensation is performed through an electrostatic actuator to eliminate the systematic errors in the sensor and the measurement system. Assume that the calibration parameters are , then the calibrated sound level data is expressed as:

[0053] ;

[0054] Through calibration, a more accurate noise feature data set is obtained. On this basis, the weighted correction data is combined with the calibration parameters to construct the noise feature vector :

[0055] ;

[0056] in, , , , , , Respectively represent the sound level data under different frequency and time characteristics, Represents the calibration compensation value. Noise characteristic vector It is a detailed description of the overall picture of noise. Based on the noise feature vector, different noise exceeding thresholds are set according to the functional area category of the monitoring point. For example, for the rehabilitation area (category 0 area), the maximum allowable sound level is set to 50 decibels; for residential and medical and health areas (category 1 area), the maximum allowable sound level is 55 decibels; for commercial and financial areas (category 2 areas), it is 60 decibels; for industrial production areas (category 3 areas), it is 65 decibels; and for areas on both sides of traffic arteries (category 4 areas), it is 70 decibels. Let the exceeding threshold set be , then it is expressed as:

[0057] ;

[0058] For each monitoring point, according to the functional area to which it belongs, the corresponding threshold is selected from the threshold set, and the noise feature vector is compared with the threshold. Exceeds the threshold of the corresponding functional area When the FLC (lossless compression) algorithm is triggered to record, the exceeding event is recorded. The specific record includes the time when the exceeding event occurs. , Duration And the extent of excess

[0059] ;

[0060] Through these data, complete exceeding standard record data is obtained, including the time, duration and magnitude of each exceeding standard. The exceeding standard record data is associated and packaged with the environmental parameter data at the corresponding time to generate audio record data.

[0061] In one example, audio recording data is input into a voiceprint recognition model for classification to obtain a noise source type identifier, and the noise source type identifier is located and calculated to obtain the spatial coordinates of the noise source, including: extracting voiceprint features from the audio recording data, dividing the noise into five categories: living noise, construction noise, industrial noise, traffic noise, and natural noise, and generating an audio feature vector by analyzing sub-class noises such as knocking sounds, insect sounds, wind sounds, and thunder sounds; performing machine learning training on the audio feature vector to obtain a voiceprint recognition model, and using the voiceprint recognition model as a baseline model, performing algorithm optimization on manually annotated audio data, adding the trained new audio to the voiceprint training library, and constructing a noise source type identifier; starting an omnidirectional acoustic radar based on the noise source type identifier, and locating and locating the noise source according to the noise pollution type. The sound field sampling interval is selected, the sound pressure data and phase data of the microphone array are collected, and the array sampling data is generated; the array sampling data is preprocessed and the signal amplitude is normalized to obtain the filtered array signal; the filtered array signal is input into the phase matching positioning algorithm, the time delay and phase difference of the sound wave reaching the adjacent microphones are calculated, and the sound source direction estimation equation is established to obtain the sound source azimuth data; based on the sound source azimuth data, a sound source depth estimation model is established by using the triangulation method, and the distance from the sound source to the array center is calculated in combination with the propagation attenuation characteristics of the sound wave in the air, and the three-dimensional spatial position data is generated; the three-dimensional spatial position data is synchronized in time and spatially mapped with the image frames collected by the surveillance camera, the noise source position is marked on the image, and the noise source spatial coordinates are generated.

[0062] In this example, voiceprint feature extraction is performed on audio recording data, and noise is divided into five categories: life noise, construction noise, industrial noise, traffic noise, and natural noise. Signal processing is performed on audio data to extract features. Audio signal After preprocessing, including denoising and enhancement of specific frequency bands, the time-frequency domain is converted. Mel frequency cepstral coefficients (MFCC) are used to extract the acoustic features of audio. MFCC can effectively capture the human ear's perception characteristics of sounds of different frequencies. The MFCC feature extraction process divides the audio signal into frames, usually with a frame length of 25ms and a frame shift of 10ms, and then Fourier transforms each frame to obtain frequency domain information. The frequency domain data is processed through a set of triangular filters to obtain a logarithmic energy spectrum, and finally the energy spectrum is discrete cosine transformed to obtain Mel frequency cepstral coefficients. In this way, a vector describing the audio features is obtained. Based on these voiceprint features, the noise is classified and the audio is divided into five categories: life noise (such as knocking sounds), construction noise (such as excavation sounds), industrial noise (such as mechanical operation sounds), traffic noise (such as car horns), and natural noise (such as insect sounds, wind sounds, and thunder sounds). Machine learning training is performed on the audio feature vector to build a voiceprint recognition model. Machine learning training uses supervised learning methods. First, a large number of labeled audio samples are prepared. These samples are used to train the model so that it can automatically identify different types of noise. Models such as support vector machines (SVM) and convolutional neural networks (CNN) are used to implement classification tasks. During the training process, the feature vector is input into the model, the goal of the model is to learn a classification function ,in is a noise type label, such as living noise, traffic noise, etc. After the model training is completed, the voiceprint recognition model is used as the baseline model, and it is optimized with new manually annotated audio to gradually improve its recognition accuracy and generalization ability. For each optimized model, the generated new audio data will be added to the voiceprint training library, and the training data set will be continuously expanded and improved to improve the adaptability and robustness of the model. After an iterative process, a voiceprint recognition system that can identify noise types with high precision is formed to construct a noise source type identification. Based on the noise source type identification, the noise source is located. Based on the identified noise type, the omnidirectional acoustic radar is started to accurately locate the spatial position of the sound source. The omnidirectional acoustic radar is an array composed of multiple microphones that can infer the direction and distance of the sound source by sampling the sound field. Depending on the type of noise pollution, the omnidirectional acoustic radar selects different sound field sampling intervals for collection. For example, for relatively stable industrial noise, a longer sampling interval is selected, while for sudden construction noise, a shorter sampling interval is required to capture its changing characteristics. Each microphone in the microphone array collects sound pressure data And phase information , these data constitute the array sampling data. The array sampling data is subjected to signal preprocessing. Signal preprocessing includes denoising, filtering and signal amplitude normalization. Amplitude normalization can eliminate the difference in sound pressure intensity caused by distance and direction of different microphones, so that the signal amplitudes of all microphones can be compared under the same benchmark. Through normalization, unnecessary amplitude differences caused by the measurement system are eliminated to ensure signal consistency. The filtered array signal is input into the phase matching positioning algorithm to calculate the time delay and phase difference of the sound wave reaching the adjacent microphone. Similar calculations are performed on all microphones to obtain multiple azimuth information, so as to accurately estimate the direction of the sound source. After obtaining the azimuth data of the sound source, the triangulation method is used to establish a sound source depth estimation model. The triangulation method calculates the distance from the sound source to the center of the array by combining the position and azimuth data of each microphone in the known microphone array with the propagation path of the sound wave. Combined with the propagation attenuation characteristics of sound waves in the air, the distance estimation is adjusted to obtain a more accurate sound source location. During the propagation of sound waves, due to the scattering and absorption of energy, its sound pressure level will decay with increasing distance. By compensating for the attenuation characteristics, the three-dimensional spatial position data of the sound source can be obtained. . Time synchronization and spatial mapping of the three-dimensional spatial position data with the image frames collected by the surveillance camera. In the surveillance area, the camera establishes a spatial mapping relationship with the acoustic array through preset calibration parameters. The location of the sound source is marked in the image frame of the camera to visualize the location of the sound source. By marking the location of the sound source on the image, a surveillance image containing the spatial coordinates of the noise source is generated.

[0063] Among them, the audio feature vector is trained by machine learning to obtain a voiceprint recognition model, and the voiceprint recognition model is used as a benchmark model. The algorithm optimization is performed on the manually annotated audio data, and the trained new audio is added to the voiceprint training library to construct a noise source type identification, including: preprocessing the audio feature vector, extracting the spectrogram features by short-time Fourier transform, and extracting the voiceprint spectrum features by Mel-frequency cepstrum coefficient analysis to generate a feature matrix; constructing a voiceprint convolutional neural network model based on the feature matrix, including an input layer, five convolution modules, two fully connected layers and an output layer. The convolution module includes a convolution layer, a batch normalization layer and an activation function layer. The output layer uses a Softmax classifier to obtain network structure parameters; the audio feature vector is divided into a data set according to five categories of labels: life noise, construction noise, industrial noise, traffic noise, and natural noise, and is divided into a training set and a validation set in a ratio of 8:2 to obtain a training data set; data enhancement is performed on the training data set, and samples are expanded by adding Gaussian noise, time stretching, amplitude scaling, etc., and the enhanced data is normalized to obtain an enhanced data set; the enhanced data set is input into the voiceprint The first convolution module of the convolutional neural network model contains 32 3×3 convolution kernels, uses the ReLU activation function and the maximum pooling layer, outputs a 32-channel feature map, and obtains the first layer of features; the first layer of features is input into the second to fifth convolution modules, and 64, 128, 256, and 512 convolution kernels are used for feature extraction, and a Dropout layer is used after each module to prevent overfitting to obtain deep features; the deep features are input into two fully connected layers, the number of nodes in the first fully connected layer is 1024, and the number of nodes in the second fully connected layer is 512, using tan h activation function to obtain the feature vector; input the feature vector into the Softmax classification layer for multi-classification, calculate the probability distribution of each category, and when the classification accuracy reaches 95%, save the model parameters to obtain the voiceprint recognition model; use the voiceprint recognition model to perform classification prediction on the newly collected manually annotated audio data, add the samples with a prediction accuracy higher than 98% to the training set, retrain the model parameters, and obtain the optimized model parameters; update the optimized model parameters to the voiceprint recognition model, and add the corresponding audio data and its features to the voiceprint training library to construct the noise source type identification.

[0064] In one example, the spatiotemporal distribution statistics are performed according to the noise characteristic vector and the noise source type identifier to obtain the functional area noise compliance rate, and the contribution of each type of noise source is calculated based on the spatial coordinates of the noise source, including: performing zoning statistics on the noise characteristic vector according to the functional area category to obtain the noise type distribution data; associating the noise type distribution data with the time dimension, performing noise compliance analysis during the daytime and nighttime periods for each type of functional area, averaging the energy of noise data at different points, and obtaining the functional area period compliance data; performing hierarchical statistics on the functional area period compliance data, dividing the noise compliance results into three levels: full-day compliance, daytime compliance, and nighttime compliance, and calculating the ratio of the number of compliance points at each level to the total number of monitoring points to obtain the functional area noise compliance rate; based on the noise source spatial coordinates The spatial diffusion analysis of noise pollution is carried out by using spatial coordinates, and the noise contour lines are superimposed on the regional map by color mapping to construct a noise distribution heat map. The cumulative sound energy of various noise sources in the noise distribution heat map is spatially superimposed according to a logarithmic relationship, and the sound energy contribution values ​​of different noise sources at each monitoring point are calculated to obtain the sound energy distribution data of the noise sources. A sound source dominance evaluation system is established based on the sound energy distribution data of noise sources, and the noise source impact factor is obtained by weighted calculation of sound pressure level and duration. The noise source impact factor is accumulated over time to obtain the spatiotemporal characteristic vector of the noise source. Based on the spatiotemporal characteristic vector of the noise source, the proportion of the impact of various noise sources on the regional sound environment quality is calculated, and the contribution of various noise sources is quantitatively weighted through the contribution scoring model to obtain the contribution of various noise sources.

[0065] In this example, the noise feature vectors collected from each monitoring point are classified and counted according to the pre-defined functional areas. Functional area categories include residential areas, industrial areas, commercial and financial areas, both sides of traffic arteries, and rehabilitation and convalescent areas. Each category has different sensitivity to noise and the range of allowed noise. Through zoning statistics, the noise data in different functional areas are effectively aggregated to facilitate subsequent analysis. The noise type distribution data is associated with the time dimension to distinguish the noise characteristics during the day and at night. The compliance requirements for noise during the day and at night are different, and the noise restrictions at night are more stringent. Through energy averaging, the noise levels of the functional areas during the day and at night are obtained. The compliance data of the functional areas during the time period are hierarchically counted, and the noise compliance results are divided into three levels: full-day compliance, daytime compliance, and nighttime compliance. For each level, the ratio of the number of compliance points to the total number of monitoring points is calculated. By calculating these ratios, the noise control situation of each functional area in different time periods is evaluated, so as to better evaluate the quality of the regional sound environment. Based on the spatial coordinates of the noise source, the spatial diffusion analysis of noise pollution is carried out to show the distribution of noise in the region in a more intuitive way. The noise contour lines are superimposed on the regional map by color mapping to construct a noise distribution heat map. The calculation of the noise contour lines is based on the sound pressure level data of each monitoring point. The noise level between the monitoring points is inferred by the interpolation algorithm. Through this interpolation method, the noise distribution of the entire area is obtained and superimposed on the regional map to form a heat map to reflect the noise intensity distribution in different areas. After constructing the noise distribution heat map, the cumulative sound energy of various noise sources is spatially superimposed according to the logarithmic relationship to calculate the sound energy contribution value of different noise sources at each monitoring point. For multiple noise sources, the cumulative sound energy is calculated by logarithmic superposition of each sound energy contribution value. In this way, the sound energy contribution value of different noise sources at each monitoring point is calculated to obtain the sound energy distribution data of the noise source. According to the sound energy distribution data of the noise source, a sound source dominance evaluation system is established, and the influence factor of the noise source is obtained by weighted calculation of the sound pressure level and duration. In order to better analyze the changes of noise sources in time and space, the influencing factors of noise sources are accumulated over time to obtain the spatiotemporal characteristic vector of noise sources. By calculating the cumulative influencing factors, the spatiotemporal characteristic vector of noise sources is obtained to describe the changing trend of noise sources in time and space. Based on the spatiotemporal characteristic vector of noise sources, the impact proportion of various noise sources on the regional sound environment quality is calculated. Various noise sources are quantitatively weighted through the contribution scoring model to obtain the contribution of various noise sources. In this way, the degree of impact of various noise sources on the regional sound environment is clarified, helping decision makers to formulate targeted noise control strategies.

[0066] In one example, a graded warning judgment is made on the noise compliance rate of the functional area, an alarm information is generated, and a noise pollution prevention and control instruction is generated according to the contribution of various noise sources, including: setting a three-level alarm threshold based on the noise compliance rate of the functional area, and combining the night-time compliance rate target of the sound environment functional area, marking a compliance rate greater than 85% as a mild warning, a compliance rate of 75%-85% as a moderate warning, and a compliance rate less than 75% as a severe warning, to obtain graded warning data; performing abnormal pattern analysis on the graded warning data and the contribution of noise sources, detecting abnormal noise events for different functional area types, and calculating the warning level, noise event type, The time of the incident and the area involved are entered into the noise event database to obtain the characteristic data of the noise event; the characteristic data of the noise event is input into the excessive alarm system, and the noise decibel value is displayed in real time on the LED display screen. The corresponding warning levels trigger the green, yellow and red alarm lights respectively, and push them to the mobile terminal at the same time to obtain the alarm information; based on the alarm information, the audio recording data, monitoring video, noise exceeding standard data, and environmental parameter data are timestamped and associated, and the governance strategy is automatically matched for different noise source types. The governance strategy includes technical measures such as noise source isolation, transmission path blocking, and sensitive point protection to generate noise pollution prevention and control instructions.

[0067] In this example, the noise compliance rate in each functional area is determined. , which means the proportion of monitoring points whose noise levels meet the specified standards to the total number of monitoring points in a certain period of time. The noise pollution status of different functional areas is evaluated by calculation. Three alarm thresholds are set according to the noise compliance rate: When When When , it is marked as a severe warning. By analyzing the noise compliance rate of the functional area, the graded warning data of the area is obtained. The abnormal pattern analysis of the graded warning data and the contribution of the noise source is performed to detect the existence of abnormal noise events. The goal of the abnormal pattern analysis is to identify whether the noise event in the functional area conforms to the normal noise distribution pattern. Suppose the contribution of the noise source is , whose value indicates the The proportion of the impact of this type of noise source on the total noise. The time series of noise sources can be used to identify abnormal changes in the contribution of noise sources. The noise event feature data is input into the over-standard alarm system to achieve real-time alarm of noise anomalies. The alarm system includes physical display devices such as LED display screens and alarm lights. The current noise decibel value is displayed in real time on the LED display screen. , to provide intuitive monitoring data. When the noise event reaches different warning levels, the corresponding color alarm lights are triggered respectively: mild warning corresponds to green light, moderate warning corresponds to yellow light, and severe warning triggers red light, indicating that emergency treatment is required. The alarm information is pushed to the mobile terminals of relevant managers to ensure that they can receive the noise exceeding the standard alarm in time. In this way, the warning information of noise exceeding the standard is presented in real time and intuitively, which helps relevant managers to take countermeasures quickly. Based on the exceeding standard alarm information, the noise exceeding standard event is comprehensively analyzed and the various data related to it are integrated. These data include audio recording data, monitoring video, noise exceeding standard data and environmental parameter data. In order to realize the integrated analysis of data, all data are timestamped and the timestamps of each data are aligned to ensure that the information from different data sources can be synchronized. After completing the data integration, the corresponding governance strategy is automatically matched for different noise source types to effectively control noise pollution. The choice of governance strategy depends on the type of noise source and its impact on the functional area, usually including technical measures such as noise source isolation, propagation path blocking and sensitive point protection. For noise source isolation, the impact on the surrounding environment can be reduced by setting up sound barriers or adjusting the operating time of the noise source. The propagation path blocking is to reduce the propagation of noise in the area by setting up sound barriers or adding green belts. For example, efficient sound barriers are set up on both sides of traffic arteries to reduce the propagation of vehicle noise. The measures for sensitive point protection are aimed at noise-sensitive areas, such as schools and hospitals, by installing soundproof windows in buildings or adding indoor sound-absorbing materials to reduce the impact of noise. Through these technical measures, the pollution impact of noise on functional areas can be effectively reduced, and the goal of improving the quality of the regional sound environment can be achieved. After completing the matching of the governance strategy, a noise pollution prevention and control instruction is generated and sent to the relevant management department or control system for the implementation of governance measures.

[0068] Reference Figure 2 This embodiment provides a regional noise monitoring and analysis system, including:

[0069] Zoning module 1 is used to perform zoning processing on the target area according to the acoustic environment function standard, obtain acoustic environment function zone division data and determine the monitoring point distribution matrix;

[0070] Collection module 2, used to collect noise data and environmental parameter data based on the monitoring point distribution matrix, and align timestamps to obtain a multi-dimensional correlation data set;

[0071] The determination module 3 is used to perform spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain the noise feature vector, and at the same time perform over-standard determination to generate audio recording data;

[0072] Classification module 4, used for inputting audio recording data into the voiceprint recognition model for classification, obtaining the noise source type identification, and performing positioning calculation on the noise source type identification to obtain the noise source spatial coordinates;

[0073] Calculation module 5, used to perform spatiotemporal distribution statistics according to the noise feature vector and the noise source type identifier, obtain the noise compliance rate of the functional area, and calculate the contribution of various noise sources based on the spatial coordinates of the noise source;

[0074] The generation module 6 is used to make graded early warning judgments on the noise compliance rate of the functional area, generate warning information, and generate noise pollution prevention and control instructions according to the contribution of various noise sources.

[0075] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0076] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0077] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0078] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0079] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0080] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0081] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for monitoring and analyzing regional noise, characterized in that: The following steps are involved: The target area is divided into zones according to the acoustic environment function standards, the acoustic environment function zone division data is obtained and the monitoring point distribution matrix is ​​determined; Based on the monitoring point distribution matrix, noise data and environmental parameter data are collected and timestamps are aligned to obtain a multi-dimensional associated data set; Performing spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain a noise feature vector, and simultaneously performing an over-standard determination to generate audio recording data; Inputting the audio recording data into a voiceprint recognition model for classification to obtain a noise source type identifier, and performing positioning calculation on the noise source type identifier to obtain the noise source spatial coordinates; Performing spatiotemporal distribution statistics according to the noise feature vector and the noise source type identifier to obtain the noise compliance rate of the functional area, and calculating the contribution of various noise sources based on the spatial coordinates of the noise source; A graded early warning judgment is made on the noise compliance rate of the functional area, an alarm message is generated, and a noise pollution prevention and control instruction is generated according to the contribution of each type of noise source.

2. The method for monitoring and analyzing regional noise according to claim 1, characterized in that: The target area is partitioned according to the acoustic environment function standard to obtain acoustic environment function area division data and determine the monitoring point distribution matrix, including: Conduct land use function analysis on the target area, extract features based on the distribution data of medical care, culture and education, residential housing, commercial finance, industrial production, and transportation trunk lines, and obtain regional function type data; The regional functional type data is input into the zoning standard model, the rehabilitation area is marked as Class 0 area, the residential area and medical and health area are marked as Class 1 area, the commercial and financial area is marked as Class 2 area, the industrial production area is marked as Class 3 area, and the area on both sides of the traffic artery is marked as Class 4 area, so as to obtain the acoustic environment functional area division data; Based on the acoustic environment functional zone division data, the number of permanent residents in the target area is read, and the city scale level is determined by the population density calculation model. When the population is less than 500,000, the number of monitoring points is set to 7, when the population is 500,000-1 million, the number of monitoring points is set to 10, when the population is 1 million-5 million, the number of monitoring points is set to 15, and when the population is greater than 5 million, the number of monitoring points is set to 20, and the minimum number of monitoring points is obtained; Perform equivalent sound level measurement on each functional area in the acoustic environment functional area division data to obtain a functional area sound level data set, and calculate the average equivalent sound level of each functional area according to the functional area sound level data set, take the average equivalent sound level as a reference value, calculate the sound level difference between each measuring point and the reference value, and select the measuring points whose sound level difference is less than a preset target value as candidate monitoring points; Evaluate the environmental conditions of the candidate monitoring points to obtain monitoring points that meet the installation conditions, and allocate the number of monitoring points that meet the installation conditions according to the minimum number of monitoring points to obtain a monitoring point quantity distribution table; The point information in the monitoring point quantity distribution table is mapped with the geographic information system to generate a monitoring point distribution matrix including the geographic coordinates of the monitoring points, functional area categories, and monitoring numbers.

3. The method for monitoring and analyzing regional noise according to claim 2, characterized in that: The noise data and environmental parameter data are collected based on the monitoring point distribution matrix, and timestamps are aligned to obtain a multi-dimensional associated data set, including: According to the point coordinates in the monitoring point distribution matrix, sound level data of the all-weather outdoor microphone is collected to obtain original sound level data, and audio parameters are calculated based on the original sound level data to obtain a noise parameter data set; Collecting wind speed, wind direction, temperature, humidity, and atmospheric pressure data from the meteorological monitoring unit at each point in the monitoring point distribution matrix to obtain an environmental parameter data set, and performing data validity verification on the noise parameter data set to obtain a valid data set; Perform spectrum parameter analysis based on the valid data set, calculate A-weighted, C-weighted, and Z-weighted sound level values, and simultaneously perform time integration of fast characteristic F, slow characteristic S, and impulse characteristic I to obtain sound level feature data; Performing time synchronization on the sound level feature data and the environmental parameter data set to obtain time series associated data, and performing data integrity check on the time series associated data to obtain integrity verification data; According to the monitoring point number, timestamp, noise characteristics and environmental parameters, the integrity verification data is organized into a multi-dimensional data structure to generate a multi-dimensional associated data set.

4. The method for monitoring and analyzing regional noise according to claim 3, characterized in that: The performing spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain a noise feature vector, and performing an over-standard determination to generate audio recording data, includes: Performing time-frequency domain transformation and sound pressure level spectrum distribution extraction on the multi-dimensional correlation data set to obtain full-spectrum signal features, and performing segmented filtering on the full-spectrum signal features to obtain frequency-band sound spectrum data; The frequency-divided sound spectrum data is corrected for sound level according to three frequency weighting networks: A weighting, C weighting, and Z weighting, and time weighting processing of fast characteristic F, slow characteristic S, and pulse characteristic I is superimposed to obtain weighted correction data; Calibrate and compensate the weighted correction data through an electrostatic actuator to generate calibration parameter data, and construct a noise feature vector based on the calibration parameter data and the weighted correction data; According to the functional area category of the monitoring point corresponding to the noise characteristic vector, the maximum allowable value of the Class 0 area is set to 50 decibels, the maximum allowable value of the Class 1 area is set to 55 decibels, the maximum allowable value of the Class 2 area is set to 60 decibels, the maximum allowable value of the Class 3 area is set to 65 decibels, and the maximum allowable value of the Class 4 area is set to 70 decibels, and a set of exceeding threshold values ​​is obtained; Perform threshold comparison on the noise feature vector, and when the measured sound level exceeds the threshold of the corresponding functional area in the exceeding threshold set, trigger the FLC lossless compression algorithm to record, and mark the exceeding time, duration, and amplitude to obtain exceeding record data; The exceeding-standard record data and the environmental parameter data at the corresponding time are associated and packaged to generate audio record data.

5. The method for monitoring and analyzing regional noise according to claim 4, characterized in that: The step of inputting the audio recording data into a voiceprint recognition model for classification to obtain a noise source type identifier, and performing positioning calculation on the noise source type identifier to obtain the noise source spatial coordinates includes: Extracting voiceprint features from the audio recording data, classifying the noise into five categories: life noise, construction noise, industrial noise, traffic noise, and natural noise, and generating audio feature vectors by analyzing sub-class noises of knocking, insect sounds, wind, and thunder; Perform machine learning training on the audio feature vector to obtain a voiceprint recognition model, and use the voiceprint recognition model as a benchmark model to perform algorithm optimization on manually annotated audio data, add the trained new audio to the voiceprint training library, and construct a noise source type identifier; Based on the noise source type identifier, the omnidirectional acoustic radar is started, a sound field sampling interval is selected according to the noise pollution type, sound pressure data and phase data of the microphone array are collected, and array sampling data is generated; Performing signal preprocessing on the array sampling data and performing signal amplitude normalization processing to obtain a filtered array signal; Inputting the filtered array signal into a phase matching positioning algorithm, calculating the time delay and phase difference of the sound wave reaching adjacent microphones, establishing a sound source direction estimation equation, and obtaining sound source azimuth data; Based on the sound source orientation data, a sound source depth estimation model is established by using a triangulation method, and the distance from the sound source to the array center is calculated in combination with the propagation attenuation characteristics of sound waves in the air to generate three-dimensional spatial position data; The three-dimensional spatial position data is time-synchronized and spatially mapped with the image frames collected by the monitoring camera, the position of the noise source is marked on the image, and the spatial coordinates of the noise source are generated.

6. The method for monitoring and analyzing regional noise according to claim 5, characterized in that: The performing of spatiotemporal distribution statistics according to the noise feature vector and the noise source type identifier to obtain the noise compliance rate of the functional area, and calculating the contribution of various noise sources based on the spatial coordinates of the noise source, includes: Performing zoning statistics on the noise feature vector according to functional area categories to obtain noise type distribution data; The noise type distribution data is associated with the time dimension, and the noise compliance analysis is performed for each type of functional area during the day and night periods. The energy of the noise data at different points is averaged to obtain the functional area period compliance data; The compliance data of the functional areas during the time periods are statistically graded, and the noise compliance results are divided into three levels: full-day compliance, daytime compliance, and nighttime compliance. The ratio of the number of compliance points at each level to the total number of monitoring points is calculated to obtain the noise compliance rate of the functional areas; Based on the spatial coordinates of the noise source, a spatial diffusion analysis of noise pollution is performed, and noise contours are superimposed on the regional map by color mapping to construct a noise distribution heat map; The accumulated sound energy of various noise sources in the noise distribution thermodynamic map is spatially superimposed according to a logarithmic relationship, and the sound energy contribution values ​​of different noise sources at each monitoring point are calculated to obtain noise source sound energy distribution data; A sound source dominance evaluation system is established according to the sound energy distribution data of the noise source, a noise source influence factor is obtained by weighted calculation of the sound pressure level and the duration, and the noise source influence factor is cumulatively counted over time to obtain a noise source spatiotemporal characteristic vector; The influence proportion of each noise source on the regional acoustic environment quality is calculated based on the spatiotemporal characteristic vector of the noise source, and each noise source is quantitatively weighted through a contribution scoring model to obtain the contribution of each noise source.

7. The method for monitoring and analyzing regional noise according to claim 6, characterized in that: The step of performing graded early warning determination on the noise compliance rate of the functional area, generating warning information, and generating noise pollution prevention and control instructions according to the contribution of various noise sources includes: A three-level alarm threshold is set based on the noise compliance rate of the functional area, and combined with the target of the night compliance rate of the acoustic environment functional area, a compliance rate greater than 85% is marked as a mild warning, a compliance rate of 75%-85% is marked as a moderate warning, and a compliance rate less than 75% is marked as a severe warning, to obtain graded warning data; An abnormal pattern analysis is performed on the graded warning data and the contribution of the noise source, abnormal noise events are detected for different functional area types, and the warning level, noise event type, event time, and involved area are entered into a noise event database to obtain noise event feature data; The noise event characteristic data is input into the excessive alarm system, and the noise decibel value is displayed in real time on the LED display screen. The corresponding warning levels trigger the green, yellow and red three-color alarm lights respectively, and are pushed to the mobile terminal at the same time to obtain the alarm information; Based on the alarm information, the audio recording data, monitoring video, noise excess data, and environmental parameter data are timestamped and associated, and governance strategies are automatically matched for different noise source types. The governance strategies include technical measures such as noise source isolation, propagation path blocking, and sensitive point protection, and noise pollution prevention and control instructions are generated.

8. A regional noise monitoring and analysis system, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: The zoning module is used to partition the target area according to the acoustic environment function standard, obtain the acoustic environment function area division data and determine the monitoring point distribution matrix; A collection module, used to collect noise data and environmental parameter data based on the monitoring point distribution matrix, and perform timestamp alignment to obtain a multi-dimensional correlation data set; A determination module, used for performing spectrum analysis and data compensation on the multi-dimensional correlation data set to obtain a noise feature vector, and at the same time, performing an over-standard determination to generate audio recording data; A classification module, used for inputting the audio recording data into the voiceprint recognition model for classification, obtaining a noise source type identifier, and performing positioning calculation on the noise source type identifier to obtain the noise source spatial coordinates; A calculation module, used to perform spatiotemporal distribution statistics according to the noise feature vector and the noise source type identifier, obtain the functional area noise compliance rate, and calculate the contribution of various noise sources based on the noise source spatial coordinates; The generation module is used to make a graded early warning judgment on the noise compliance rate of the functional area, generate alarm information, and generate noise pollution prevention and control instructions according to the contribution of each type of noise source.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images