Noise monitoring device based on autonomous network and noise source tracing optimization method
By deploying autonomous network noise monitoring devices and improved Seq2Seq algorithms in various areas of the city, acoustic soundprint library is built and multi-model noise monitoring network architecture is used to solve the problem that traditional technology cannot monitor noise accurately and trace noise in real time, realizing noise monitoring Comprehensive and real-time, and solve the problem of noise pollution in a timely manner.
Patent Information
- Application Number
- CN202410092943.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2023-12-08
- Filing Date
- 2024-01-23
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-01-23
AI Technical Summary
Traditional urban noise monitoring technology cannot fully capture noise conditions, it is difficult to monitor noise events in real time, and it is impossible to accurately trace the source of noise, resulting in the noise pollution problem not being solved in a timely manner.
Using the noise monitoring device based on autonomous network and the improved Seq2Seq algorithm, the noise monitoring device is deployed to collect and process noise data in various areas of the city, and acoustic acoustic print library is built, and noise traceability is used to use multi-model Seq2Seq noise monitoring network architecture and cosine calculation. , quickly determine the source of noise.
It realizes comprehensive monitoring and real-time response to urban noise, accurately traces the noise source, improves the accuracy and operability of noise monitoring, and can solve the problem of noise pollution in a timely manner.
Smart Images

Figure CN117928722B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of big data and acoustic technology, and in particular to a noise monitoring device based on an autonomous network and a noise source tracing optimization method. Background Art
[0002] As the urbanization process continues to accelerate, urban noise pollution has become a serious environmental problem, which has had a serious negative impact on people's health and quality of life. It has become increasingly urgent to strictly control the generation and spread of noise, and monitoring and identifying noise during the control process is particularly important. Traditional urban noise monitoring technology has obvious defects in the face of this challenge. It can only monitor the noise level of the regional environment, but cannot specifically identify the source of the noise and how it is generated. Traditional noise monitoring methods usually rely on a limited number of distributed sensors. These sensors have limited coverage and it is difficult to fully capture the noise conditions in various areas of the city, which leads to limitations in monitoring results and is prone to missed reports and false alarms. Due to the limited number of sensors, it is impossible to achieve real-time and effective monitoring of noise events within the city, which means that noise problems cannot be discovered and solved in a timely manner, and residents' lives are continuously disturbed. In terms of tracing the source of noise, traditional methods often have difficulty tracing back to the specific source of noise and who generates the noise due to the lack of detailed acoustic feature information. This makes it difficult for urban management departments to take effective control measures because it is impossible to clearly identify which noise sources form noise pollution. In addition, traditional noise monitoring methods have delays in data processing and transmission, and cannot meet real-time requirements. They cannot meet the requirements of urban noise events that require immediate processing and response. Traditional methods for solving noise have serious delays and lags, which further aggravates the problem of urban noise pollution. In summary, traditional urban noise monitoring technology has obvious defects in comprehensiveness, real-time and noise source tracing, and cannot meet the needs of noise management in today's urbanization process. For this reason, the inventor proposed a noise monitoring device based on an autonomous network and a noise source tracing optimization method of an improved Seq2Seq algorithm, using an innovative technology to overcome these problems, improve the accuracy, real-time and operability of noise monitoring, and effectively solve the problem of noise pollution. Summary of the invention
[0003] To solve the above technical problems, the present invention proposes a noise monitoring device based on an autonomous network and a noise source tracing optimization method. By setting a device box and a back cover of the noise monitoring device of the autonomous network, a radio microphone and a communication antenna are installed on the upper part of the device box, a circuit board and a battery are installed in the device box, and a power interface is set on one side of the circuit board; the steps of the noise source tracing optimization method of the improved Seq2Seq algorithm based on the noise monitoring device of the autonomous network are as follows: step one, deployment of the noise monitoring device of the autonomous network; step two, noise data collection and processing; step three, construction of an acoustic voiceprint library of the autonomous network; step four, construction of an improved Seq2Seq algorithm model network; the noise monitoring device and the noise source tracing optimization method based on the autonomous network establish a cloud voiceprint library in the autonomous network, adopt a multi-model Seq2Seq noise monitoring network architecture, use cosine calculation and intermediate voiceprint priority matching principle to trace the noise source, quickly determine the source of noise, and effectively and quickly solve noise pollution.
[0004] To achieve the above object, the technical solution adopted by the present invention is:
[0005] The noise monitoring device based on autonomous network includes a radio microphone, a communication antenna, a device box, a circuit board, a battery, a power interface and a back cover. The noise monitoring device based on autonomous network is provided with a device box and a back cover, and a waterproof ring is provided between the device box and the back cover. A radio microphone and a communication antenna are installed on the upper part of the device box. A network management system is established by using the communication antenna to monitor the status and data transmission of the monitoring device. A circuit board and a battery are installed in the device box. The circuit board includes a microphone, a digital signal processor, a memory and a storage unit. A cloud storage system is established to receive and store noise data streams of autonomous network monitoring devices of various noise nodes in the city, and then generate noise decibel values, frequencies and amplitudes. A power interface is provided on one side of the circuit board where the battery is installed, and a power port is provided on the device box at a position corresponding to the power interface.
[0006] As a further improvement of the device of the present invention, the microphone adopts the Knowles SPU0410LR5H-QB microphone chip, the digital signal processor adopts the Cortex-A72 processor chip, the memory and storage unit include K4F8E3S4HM-MGCJ memory chip, NAND flash memory, Ethernet interface chip, NVIDIA image processing chip, DSPC5535 voiceprint processing and recognition chip and Texas Instruments power management chip.
[0007] The present invention provides a noise source tracing optimization method of an improved Seq2Seq algorithm of a noise monitoring device based on an autonomous network, which is characterized by comprising the following steps:
[0008] Step 1: Deployment of noise monitoring devices in the autonomous network: deploy the noise monitoring devices in the autonomous network to the noise nodes in each area, establish a network management system using communication antennas, and monitor the status of the monitoring devices and data transmission;
[0009] Step 2: Noise data collection and processing: Establish a cloud storage system to receive and store the noise data stream of the autonomous network monitoring device of each noise node in the city, and then generate the noise decibel value, frequency and amplitude. Use the YOLOV8 network to conduct deep learning on the waveform of the sound to form the waveform noise probability characteristic value, use the high amplitude fluctuation interval as one of the expressions to characterize the noise, and use the interval formula to calculate the high amplitude fluctuation interval characteristic value;
[0010] Step 3: Autonomous network acoustic voiceprint library construction: Acoustic feature vectors of noise nodes in each area are obtained through autonomous network monitoring devices, and then the acoustic voiceprint library of noise nodes in each area is constructed. The acoustic voiceprint library without autonomous network monitoring devices between nodes is constructed through linear interpolation and diffusion formulas, so as to realize the construction of noise voiceprint library for the whole city.
[0011] Step 4. Improve the construction of Seq2Seq algorithm model network. By improving the Seq2Seq algorithm model network, predict the acoustic vector generated by the autonomous network noise monitoring device in real time, use the voiceprint loss function to give the algorithm model loss, and match the vector with the acoustic voiceprint library to which it belongs. The specific judgment result is determined by the threshold formula, and finally the noise tracing work is completed.
[0012] Furthermore, the interval formula used in step 2 of the noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network is expressed as:
[0013] The interval formula is expressed as:
[0014]
[0015]
[0016] Among them, f i is the amplitude value at the current moment, f t is the amplitude value at the forward moment, F is the average amplitude of the forward 120 seconds, α is the abnormal amplitude amplification factor, and σ is the specific value of the high amplitude fluctuation interval.
[0017] Furthermore, the diffusion formula used in step 3 of the noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network is expressed as:
[0018] The diffusion formula is expressed as follows:
[0019]
[0020] Among them, χ i is the eigenvalue of the interpolation vector between the two autonomous network noise monitoring devices, k1 and k2 are the most significant features of the noise data monitored by the two autonomous networks in their respective areas, that is, if the autonomous network noise monitoring device is located in the main traffic road and industrial area, it is the decibel value feature, if it is in the commercial area, it is the amplitude, if it is in the residential area, it is the high amplitude fluctuation interval, and if it is in the school area, it is the frequency; d1 and d2 are the other original eigenvalues of the two autonomous network noise monitoring devices respectively.
[0021] Furthermore, the voiceprint loss function and threshold formula used in step 4 of the noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network are expressed as follows:
[0022] Among them, the voiceprint loss function is expressed as follows:
[0023]
[0024] Among them, Lo is the voiceprint loss value, γ i is the characteristic vector value of the acoustic data monitored by the autonomous network monitoring device in real time, that is, a vector composed of five data types: decibel value, high amplitude fluctuation interval, frequency, amplitude, and waveform noise probability. There are two types, namely noise and non-noise. n indicates the total amount of data of this type contained in the acoustic voiceprint library monitored by the autonomous network monitoring device at this location, and p i is the i-th data vector of this type in the acoustic voiceprint library monitored by the autonomous network monitoring device;
[0025] The threshold formula is as follows:
[0026]
[0027] Among them, y i Represents the vector value predicted by the network, Y i is the value closest to the cosine in the palmprint database, and ω is the judgment factor. This formula can be used to determine which acoustic voiceprint library the predicted vector belongs to.
[0028] The present invention provides a noise monitoring device based on an autonomous network and a noise source tracing optimization method. The device box and the back cover of the noise monitoring device of the autonomous network are provided, a radio microphone and a communication antenna are installed on the upper part of the device box, a circuit board and a storage battery are installed in the device box, and a power interface is provided on one side of the circuit board. The steps of the noise source tracing optimization method of the improved Seq2Seq algorithm based on the noise monitoring device of the autonomous network are as follows: step 1, deployment of the noise monitoring device of the autonomous network; step 2, noise data collection and processing; step 3, construction of an acoustic voiceprint library of the autonomous network; step 4, construction of an improved Seq2Seq algorithm model network; the noise monitoring device and the noise source tracing optimization method based on the autonomous network establish a cloud voiceprint library on the autonomous network, adopt a multi-model Seq2Seq noise monitoring network architecture, use cosine calculation and intermediate voiceprint priority matching principle to trace the noise source, and quickly determine the noise source; the benefits brought are:
[0029] 1. The noise monitoring device based on autonomous network and the noise source tracing optimization method can cover all areas of the city by deploying the noise monitoring device of autonomous network, and realize the ability of comprehensive monitoring of urban noise with the help of cloud storage and deep learning technology. This enables the city management department to obtain noise data across the city in a timely manner and respond to noise events in real time, so as to better protect the health and quality of life of residents;
[0030] 2. The noise monitoring device based on the autonomous network and the noise source tracing optimization method can trace the source of noise more accurately by constructing an acoustic voiceprint library and introducing a multi-model Seq2Seq noise monitoring network; this is crucial for taking targeted governance measures and helps urban management departments to quickly identify and solve noise pollution problems;
[0031] 3. The noise monitoring device and noise source tracing optimization method based on the autonomous network introduces the high amplitude fluctuation interval as one of the noise characteristics, and provides an interval formula to quantitatively describe this feature; different from the traditional noise characteristics which usually focus on parameters such as decibel value and frequency, the high amplitude fluctuation interval as a new quantitative feature to characterize noise can provide more information for noise monitoring and tracing, and help to more accurately distinguish different sources and types of noise;
[0032] 4. The noise monitoring device based on the autonomous network and the noise source tracing optimization method use an innovative noise monitoring network architecture, adopt the Seq2Seq model, and propose the intermediate voiceprint priority matching principle; this architecture allows the real-time collected acoustic data to be input into the algorithm network, and the source of the noise is determined by vector matching. The intermediate voiceprint priority matching principle further improves the matching accuracy and helps to quickly and reliably determine the noise source. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 It is a schematic diagram of the overall structure of the present invention;
[0034] Figure 2 It is a flow chart of the noise tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network of the present invention;
[0035] Figure 3 It is a schematic diagram of a sound waveform diagram of a noise source tracing optimization method of an improved Seq2Seq algorithm of a noise monitoring device based on an autonomous network of the present invention;
[0036] Figure 4 It is a schematic diagram of an acoustic voiceprint library of a noise source tracing optimization method of an improved Seq2Seq algorithm of a noise monitoring device based on an autonomous network of the present invention;
[0037] Figure 5 This is a schematic diagram of the algorithm network architecture of the noise tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network of the present invention.
[0038] The components marked in the figure are: 1. Radio microphone; 2. Communication antenna; 3. Equipment box; 4. Circuit board; 5. Battery; 6. Power interface; 7. Back cover. DETAILED DESCRIPTION
[0039] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:
[0040] like Figure 1 As shown, the noise monitoring device based on the autonomous network includes a radio microphone 1, a communication antenna 2, a device box 3, a circuit board 4, a battery 5, a power interface 6 and a back cover, and is characterized in that: the noise monitoring device based on the autonomous network is provided with a device box 3 and a back cover 7, a waterproof ring is provided between the device box 3 and the back cover 7, the device box 3 and the back cover 7 are installed with bolts, and have good waterproof performance after installation, and can be installed and used outdoors, a radio microphone 1 and a communication antenna 2 are installed on the upper part of the device box 3, the radio microphone 1 and the communication antenna 2 are waterproof, which can avoid the influence of water vapor on them when installed outdoors, and the communication antenna 2 adopts wireless communication to communicate with the cloud The terminal equipment is connected to transmit data, and a network management system is established using a communication antenna to monitor the status of the monitoring device and data transmission. A circuit board 4 and a battery 5 are installed in the device box 3 shown in the figure. The circuit board includes a microphone, a digital signal processor, a memory and a storage unit. A cloud storage system is established to receive and store the noise data stream of the autonomous network monitoring device of each noise node in the city, and then generate the noise decibel value, frequency and amplitude. A power interface 6 is set on one side of the circuit board 4 where the battery 5 is installed, and a power port is set at a position corresponding to the power interface 6 on the device box 3 shown in the figure. Power is provided to the battery 5 and the circuit board 4 through the power interface 6, and the battery 5 can also provide power to the circuit board alone.
[0041] The microphone of the present invention adopts the Knowles SPU0410LR5H-QB microphone chip, the digital signal processor adopts the Cortex-A72 processor chip, the memory and storage unit include K4F8E3S4HM-MGCJ memory chip, NAND flash memory, Ethernet interface chip, NVIDIA image processing chip, DSPC5535 voiceprint processing and recognition chip and Texas Instruments power management chip.
[0042] like Figure 2 As shown, a flow chart of a noise source tracing optimization method of an improved Seq2Seq algorithm of a noise monitoring device based on an autonomous network is shown, and the noise source tracing optimization method of an improved Seq2Seq algorithm of a noise monitoring device based on an autonomous network specifically includes the following steps:
[0043] Step 1: Deployment of noise monitoring devices in the autonomous network; deploy the noise monitoring devices in the autonomous network to the noise nodes in each area, establish a network management system using communication antennas, and monitor the status and data transmission of the monitoring devices; it is worth noting that the specific deployment location needs to be determined according to the environment; that is, first of all, it is necessary to ensure that the devices are deployed in key parts of the city, including traffic arteries, commercial areas, industrial areas, residential areas and school areas; the above-mentioned noise monitoring devices in the autonomous network need to be placed on the roofs of buildings, telephone poles or traffic signal poles to fully capture noise information; after ensuring that each monitoring point is stable and safe, configure the data communication module, and perform calibration and testing on the data quality control to ensure that the sensor generates accurate noise data; finally, establish a network management system to monitor the status and data transmission of the monitoring devices, and set up an alarm mechanism so that timely measures can be taken when problems occur in the noise monitoring devices of the autonomous network;
[0044] Step 2: Noise data collection and processing: Establish a cloud storage system to receive and store the noise data stream from the autonomous network monitoring device of each noise node in the city, and then generate noise decibel value, frequency and amplitude, and use the YOLOV8 network to conduct deep learning on the waveform of the sound to form the waveform noise probability characteristic value, use the high amplitude fluctuation interval as one of the expressions to characterize the noise, and use the interval formula to calculate the high amplitude fluctuation interval characteristic value; process the original noise data of each noise node in the cloud storage system to eliminate environmental interference and outliers; Figure 3The schematic diagram of the sound waveform of the noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network is shown. The waveform of the normal sound after optimization is different from the noise waveform. The method adopts a median filter to reduce high-frequency noise and outliers; in addition, for all missing values, in order to increase the response speed of the overall module, this application directly adopts the forward filling method, that is, directly uses the value of the previous time point to fill the missing value; in addition, each noise data source is labeled to facilitate the subsequent tracing to quickly locate; in this step, it is necessary to maintain the comparability and consistency of the noise data, so firstly, the stored data is converted into units, that is, all the sound pressure level data is converted into decibels, and the sampling frequency and the number of sampling bits are unified. In addition, all data are subjected to standard deviation scaling to ensure the consistency of the data in scale;
[0045] The characteristics of the overall acoustic data include decibel value, high amplitude fluctuation interval, frequency, amplitude, and waveform noise probability; among them, decibel value, frequency, and amplitude are directly used as characteristic values x1, x2, and x3; this application proposes for the first time that high amplitude fluctuation interval is used as one of the expression quantities to characterize noise, and the high amplitude fluctuation interval feature can be given by the interval formula;
[0046] The interval formula is expressed as:
[0047]
[0048]
[0049] Among them, f i is the amplitude value at the current moment, f t is the amplitude value at the forward moment, F is the average amplitude of the forward 120 seconds, α is the abnormal amplitude amplification factor, and σ is the specific value of the high amplitude fluctuation interval;
[0050] In the above formula, the average value of the amplitude within two minutes is first calculated by formula (1), and then the current amplitude value is judged by formula (2). If the current amplitude value is within the set range, the high amplitude fluctuation interval feature is given as 0, and if the current amplitude value exceeds the set range, the high amplitude fluctuation interval is set as the abnormal amplitude amplification factor, the current vibration amplitude and the relevant value of the average amplitude;
[0051] Perform deep learning on the waveform of the sound, i.e., image classification; use the YOLOV8 network architecture to perform image classification transfer learning, where the classification target is whether it is a noise waveform. If it is a noise waveform, it is judged as 1, otherwise it is judged as 0; in addition, the number of noise waveforms in this application is less than that of normal waveforms, so this application proposes for the first time to use noise waveform splicing technology to expand data; that is, by cutting and splicing limited noise waveforms, a new noise waveform is formed; when the network training is completed, the network confidence of the noise waveform of each waveform of the sound data gives the waveform noise probability characteristic value of the data;
[0052] Step 3: Autonomous network acoustic voiceprint library construction: Acoustic feature vectors of noise nodes in each area are obtained through autonomous network monitoring devices, and then the acoustic voiceprint library of noise nodes in each area is constructed. The acoustic voiceprint library without autonomous network monitoring devices between nodes is constructed through linear interpolation and diffusion formula, so as to realize the construction of noise voiceprint library for the whole city; Figure 4 As shown, it is a schematic diagram of the acoustic voiceprint library of the noise tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network of the present invention; as shown in step one and step two, the deployment of the autonomous network noise device and the feature extraction of the noise data are completed. Given, in this step, it is necessary to complete the construction of the acoustic voiceprint library of each location in each area of the city; it should be noted that in addition to constructing the acoustic voiceprint library for each location in each area of the city, the acoustic voiceprint library in this application also needs to be classified, that is, the feature data of the normal sound of each location are stored in the voiceprint library, including the classification The value of the sound, high amplitude fluctuation interval, frequency, amplitude and waveform noise probability; for noise data, it is also necessary to construct the noise soundprint library based on the autonomous network noise monitoring device at various locations in the city, that is, two acoustic soundprint libraries are constructed at this time. The data in these two acoustic soundprint libraries are the data of the autonomous network noise monitoring equipment of each node, but the first acoustic soundprint library only contains normal sounds, while the second acoustic soundprint library only contains the noise characteristic data of each node; in addition, this application proposes to use a diffusion formula to construct a city-wide soundprint library based on the scattered nodes and regional characteristics of the autonomous network noise monitoring device;
[0053] The diffusion formula is expressed as follows:
[0054]
[0055] Among them, χ iis the eigenvalue of the vector to be interpolated between the two autonomous network noise monitoring devices, k1 and k2 are the most significant features of the noise data monitored by the two autonomous networks in their respective areas, that is, if the autonomous network noise monitoring device is located in a traffic artery and an industrial area, it is the decibel value feature, if it is located in a commercial area, it is the amplitude, if it is located in a residential area, it is the high amplitude fluctuation interval, and if it is located in a school area, it is the frequency; d1 and d2 are the remaining original eigenvalues of the two autonomous network noise monitoring devices respectively;
[0056] Formula (3) can be used to complete the construction of the city-wide voiceprint library of all autonomous network noise monitoring devices' scattered nodes and regional characteristics. For example, when one autonomous network noise monitoring device is in the school area and another autonomous network noise monitoring device is in the industrial area, the nodes in the middle area of the two autonomous networks are interpolated and filled. Among the five feature data, decibel value, high amplitude fluctuation interval, frequency, amplitude and waveform noise probability, the decibel value is directly filled with the decibel value of the industrial park, and the frequency is directly filled with the frequency of the school. The remaining feature values, high amplitude fluctuation interval, amplitude, and waveform noise probability, are linearly interpolated between the two areas.
[0057] Step 4: Improve the Seq2Seq algorithm model network construction; by improving the Seq2Seq algorithm model network, predict the acoustic vector generated by the autonomous network noise monitoring device in real time, use the voiceprint loss function to give the algorithm model loss, and match the vector to its affiliated acoustic voiceprint library. The specific judgment result is determined by the threshold formula, and finally the noise tracing work is completed; Figure 5 As shown, a schematic diagram of the algorithm network architecture of the noise tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network of the present invention is shown; the improved Seq2Seq algorithm model network is constructed to judge the acoustic data monitored in real time, and the characteristic item of the network input is the acoustic data monitored by the autonomous network noise monitoring device at 10 consecutive moments, and its output item is also a vector value, which is used to match the acoustic voiceprint library in step three to find the vector most similar to it, and the loss value is the cosine distance between the predicted vector and the closest vector, and is given by the voiceprint loss function;
[0058] Among them, the voiceprint loss function is expressed as follows:
[0059]
[0060] Among them, Lo is the voiceprint loss value, γi is the characteristic vector value of the acoustic data monitored by the autonomous network monitoring device in real time, that is, a vector composed of five types of data: decibel value, high amplitude fluctuation interval, frequency, amplitude, and waveform noise probability. There are two types, namely noise and non-noise. n indicates the total amount of data of this type contained in the acoustic voiceprint library monitored by the autonomous network monitoring device at this location, and p iis the i-th data vector of this type in the acoustic voiceprint library monitored by the autonomous network monitoring device;
[0061] The initial values of the device monitoring are shown in the following table:
[0062]
[0063] A multi-model Seq2Seq noise monitoring network architecture is proposed for the acoustic voiceprint library. First, the acoustic data collected by the autonomous network noise monitoring device in real time is input into the algorithm network architecture, and then its output vector is matched. The cosine calculation is used for the matching principle between the vector output by the network and the vector in the voiceprint library, and a priority matching principle for the intermediate voiceprint is proposed. The specific process is to match the vector predicted by the algorithm network with the point of the middle difference of the autonomous network noise device in different areas. When the threshold formula is met, it is determined that the noise comes from the vicinity of the two autonomous network noise monitoring devices, and then a detailed match is performed with the acoustic voiceprint library of the two autonomous network noise monitoring devices to check which autonomous network noise monitoring device it belongs to. Finally, the source of the noise is determined according to the autonomous network noise monitoring device. When the intermediate interpolation data cannot match the predicted vector, that is, it cannot meet the threshold formula, the vector value is strongly matched with the noise vector of the autonomous network acoustic voiceprint library at each location to check whether there is an acoustic voiceprint library vector that meets the threshold formula; when none of them exists, it is determined to be a normal sound.
[0064] The threshold formula is as follows:
[0065]
[0066] Among them, y i Represents the vector value predicted by the network, Y i is the value closest to the cosine in the palmprint database, and ω is the judgment factor. This formula can be used to determine which acoustic voiceprint database the predicted vector belongs to.
[0067] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.
Claims
1. A noise source tracing optimization method for an improved Seq2Seq algorithm of a noise monitoring device based on an autonomous network, wherein the noise monitoring device based on the autonomous network comprises a radio microphone (1), a communication antenna (2), a device box (3), a circuit board (4), a storage battery (5), a power interface (6) and a back cover, wherein the noise monitoring device based on the autonomous network is provided with a device box (3) and a back cover (7), a waterproof ring is provided between the device box (3) and the back cover (7), a radio microphone (1) and a communication antenna (2) are installed on the upper part of the device box (3), and a network is established by using the communication antenna (2) A management system is provided to monitor the status and data transmission of the monitoring device. A circuit board (4) and a battery (5) are installed in the device box (3). The circuit board (4) includes a microphone, a digital signal processor, a memory and a storage unit. A cloud storage system is established to receive and store the noise data stream of the autonomous network monitoring device of each noise node in the city, and then generate the noise decibel value, frequency and amplitude. A power interface (6) is arranged on the side of the circuit board (4) where the battery (5) is installed. A power port is arranged on the device box (3) at a position corresponding to the power interface (6). The microphone adopts a Knowles SPU0410LR5H-QB microphone chip, the digital signal processor adopts a Cortex-A72 processor chip, the memory and storage unit include a K4F8E3S4HM-MGCJ memory chip, a NAND flash memory, an Ethernet interface chip, an NVIDIA image processing chip, a DSPC5535 voiceprint processing and recognition chip and a Texas Instruments power management chip. The characteristics are: The following steps are involved: Step 1: Deployment of noise monitoring devices in the autonomous network: deploy the noise monitoring devices in the autonomous network to the noise nodes in each area, establish a network management system using communication antennas, and monitor the status of the monitoring devices and data transmission; Step 2: Noise data collection and processing: Establish a cloud storage system to receive and store the noise data stream of the autonomous network monitoring device of each noise node in the city, and then generate the noise decibel value, frequency and amplitude. Use the YOLOV8 network to conduct deep learning on the waveform of the sound to form the waveform noise probability characteristic value, use the high amplitude fluctuation interval as one of the expressions to characterize the noise, and use the interval formula to calculate the high amplitude fluctuation interval characteristic value; The interval formula used in step 2 is expressed as: The interval formula is expressed as: Among them, f i is the amplitude value at the current moment, f t is the amplitude value at the forward moment, F is the average amplitude of the forward 120 seconds, α is the abnormal amplitude amplification factor, and σ is the specific value of the high amplitude fluctuation interval; Step 3: Autonomous network acoustic voiceprint library construction: Acoustic feature vectors of noise nodes in each area are obtained through autonomous network monitoring devices, and then the acoustic voiceprint library of noise nodes in each area is constructed. The acoustic voiceprint library without autonomous network monitoring devices between nodes is constructed through linear interpolation and diffusion formulas, so as to realize the construction of noise voiceprint library for the whole city. Step 4. Improve the construction of Seq2Seq algorithm model network. By improving the Seq2Seq algorithm model network, predict the acoustic vector generated by the autonomous network noise monitoring device in real time, use the voiceprint loss function to give the algorithm model loss, and match the vector with the acoustic voiceprint library to which it belongs. The specific judgment result is determined by the threshold formula, and finally the noise tracing work is completed.
2. The noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network according to claim 1 is characterized by: The diffusion formula used in step 3 of the noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network is expressed as: The diffusion formula is expressed as follows: Among them, χ i is the eigenvalue of the interpolation vector between the two autonomous network noise monitoring devices, k1 and k2 are the most significant features of the noise data monitored by the two autonomous networks in their respective areas, that is, if the autonomous network noise monitoring device is located in the main traffic road and industrial area, it is the decibel value feature, if it is in the commercial area, it is the amplitude, if it is in the residential area, it is the high amplitude fluctuation interval, and if it is in the school area, it is the frequency; d1 and d2 are the other original eigenvalues of the two autonomous network noise monitoring devices respectively.
3. The noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network according to claim 1 is characterized by: The voiceprint loss function and threshold formula used in step 4 of the noise source tracing optimization method of the improved Seq2Seq algorithm of the noise monitoring device based on the autonomous network are expressed as follows: Among them, the voiceprint loss function is expressed as follows: Among them, Lo is the voiceprint loss value, ri is the characteristic vector value of the acoustic data monitored by the autonomous network monitoring device in real time, that is, a vector composed of five data types: decibel value, high amplitude fluctuation interval, frequency, amplitude, and waveform noise probability. There are two types, namely noise and non-noise. n indicates the total amount of data of this type contained in the acoustic voiceprint library monitored by the autonomous network monitoring device at this location, and p i is the i-th data vector of this type in the acoustic voiceprint library monitored by the autonomous network monitoring device; The threshold formula is as follows: Among them, y i Represents the vector value predicted by the network, Y i is the value closest to the cosine in the palmprint database, and ω is the judgment factor. This formula can be used to determine which acoustic voiceprint database the predicted vector belongs to.
