Method for intelligently identifying urban noise source based on neural network

Abstract

The invention discloses a method for intelligently identifying an urban noise source based on a neural network, relates to the technical field of noise identification, and aims to solve the problems that the current urban noise source identification method is difficult to realize real-time detection of the noise source, so that the accuracy of noise source category identification is reduced, the difficulty of noise source classification is relatively high, and the accuracy of noise source classification is low. The method comprises the following steps: establishing an urban noise source database, establishing an urban noise basic characteristic parameter database, carrying out analysis on the basis of the urban noise source database to establish a noise source characteristic parameter standard data set, determining noise source characteristic parameters, and determining a target area. The urban noise monitoring system is constructed in the center of the target area, the feature parameter vector of the noise source of the target area is measured, the intelligent noise source recognition model is automatically constructed by applying the neural network algorithm, discrimination of different noise sources is assisted, and the recognition rate is high.

Description

Technical Field

[0001] The present invention relates to the technical field of noise identification, and more specifically to a method for intelligently identifying urban noise sources based on a neural network. Background Art

[0002] Noise sources are the source of noise, and intelligent identification of noise sources is the premise for noise source analysis and the construction of urban sound environment monitoring station network. However, relevant technical specifications and technological development obviously lag behind the speed of urban development. Intelligent identification of noise sources is the premise for noise source monitoring and control, which directly affects the management and service level of urban sound environment. However, in the existing intelligent noise source identification methods, facing the complex and changeable noise source categories, conventional identification methods have been difficult to meet the needs of intelligent noise source identification. Due to the lack of noise source monitoring equipment, it is impossible to meet the needs of fast, comprehensive and real-time detection. Therefore, the current noise source identification mainly relies on human ear recognition and identification of sound spectrum features, but the accuracy of human ear recognition alone is low. At the same time, the sound spectrum features of noise sources have little difference, and the classification of noise sources is difficult. Conventional detection methods are not easy to achieve the identification and classification of noise sources. In view of this, we propose a method for intelligently identifying urban noise sources based on neural networks. Summary of the invention

[0003] The purpose of the present invention is to provide a method for intelligently identifying urban noise sources based on a neural network, so as to solve the technical problem that the current urban noise source identification method is difficult to achieve real-time detection of noise sources, resulting in a decrease in the accuracy of noise source category identification, making noise source classification more difficult and unable to meet the technical problem of intelligent noise source identification requirements.

[0004] To solve the above technical problems, the present invention provides the following technical solutions: a method for intelligently identifying urban noise sources based on a neural network, a method for intelligently identifying urban noise sources based on a neural network, comprising the following steps: S1: Establish urban noise source database; S2: Establish a database of basic characteristic parameters of urban noise; S3: Analyze based on the urban noise source database to establish a standard data set of noise source characteristic parameters and determine the noise source characteristic parameters; S4: Determine the target area, build an urban noise monitoring system in the center of the target area, and measure the characteristic parameter vector of the noise source in the target area; S5: Calculate the radiation value of the noise source in the target area. The formula is: ; In the formula, is the sampling time, Locate the sensor at the location of the sound source The time when the maximum noise is detected, Locate the sensor at the location of the sound source The maximum noise detected at is the weight coefficient; S6: Calculate the target area feature vector , , where The geographical coordinates of the sensor locating the sound source, is the radiation value of the noise source in the target area, is the type of noise source in the target area; S7: Mark the noise sources in the target area; S8: Establish a neural network based on the marked noise source parameters of the target area and establish a noise source intelligent identification model; S9: In the target area, the noise source parameters of the target area are input into the noise source intelligent identification model to identify the type of the target noise source.

[0005] The present invention analyzes and identifies six types of noise sources in the city, establishes a noise source database and a noise source basic characteristic parameter database, and determines a standard data set of noise source characteristic parameters. By calculating the noise source radiation value and characteristic vector, a neural network is established, and the best neural network model is extracted according to the accuracy, so as to establish a noise source intelligent recognition model, which can assist in the identification of different noise sources, realize the extraction of characteristic parameters of different noise sources, quantitatively calculate the noise radiation values ​​of different noise sources, and use the neural network algorithm to automatically build a noise source intelligent recognition model. In actual detection, the characteristic vector of the noise source is obtained by calculation, and the characteristic vector is input into the noise source intelligent recognition model, so that the noise source type can be intelligently identified, and a higher recognition rate can be obtained.

[0006] Preferably, the step S1 further comprises the following steps: S101: Collect urban noise source data; The noise source data includes the name of the sound source, the category of the sound source, the attribute of the sound source, the location of the sound source, and the shape of the sound source. According to the attribute of the sound source category, the noise source data is divided into six types: traffic noise source, industrial noise source, construction noise source, sanitation machinery noise source, entertainment noise source, and human voice noise source; S102: Classify the sound source data into noise sources and non-noise sources according to sound source attributes; S103: Determine the relative position of each noise source in the city according to the position of the sound source; The sound source location includes the latitude and longitude of the sound source's geographical location and its specific location in the city. S104: Determine the area of ​​the noise source according to the shape of the sound source.

[0007] Preferably, the step S3 further comprises the following steps: S301: Determine a standard data set of noise source characteristic parameters corresponding to radiation values ​​of different types of noise sources; S302: Determine the distance between the noise source and the sensor, establish a standard data set of radiation values ​​between the sensor and the noise source to be measured, and obtain the distance range between the noise source and the sensor in the standard data set of radiation values ​​between the sensor and the noise source to be measured.

[0008] Preferably, the noise source characteristic parameter standard data set includes: Traffic noise source radiation value: , where is the radiation value of the traffic noise source, indicating the sound power level radiated by the source point. Indicates the sound source power corresponding to the maximum value of the noise source, represents the vehicle speed corresponding to the maximum value of the noise source, Indicates the temperature and humidity corresponding to the maximum value of the noise source, They represent the temperature and humidity constants corresponding to the traffic noise sources, Indicates that the noise source is spatially Axis and The axis corresponds to the height; Radiation values ​​of industrial noise sources: , where is the radiation value of the industrial noise source, Indicates the mechanical power or electrical power corresponding to the maximum value of the noise source; Radiation value of construction noise source: , where is the radiation value of the construction noise source, They represent the temperature and humidity constants corresponding to the construction noise sources respectively; Radiation value of noise source of sanitation machinery: , where They represent the temperature and humidity constants corresponding to the noise sources of sanitation machinery respectively; Entertainment noise source radiation value: , where is the radiation value of the entertainment noise source, Indicates the number of people corresponding to the maximum value of this type of noise source; Radiation value of human voice noise source: , where The radiation value of the human voice noise source.

[0009] Preferably, the urban noise monitoring system consists of The noise monitoring network is constructed by using a sound source localization sensor.

[0010] Preferably, the step S4 further comprises the following steps: S401: First, determine the target area, use sound source positioning sensors to build a noise monitoring network, determine the number and location of sensors, and measure each sound source positioning sensor to obtain characteristic parameters of the noise source in the target area; The characteristic parameters of the noise source in the target area include noise frequency and noise level; S402: Input the obtained characteristic parameters into the noise source basic characteristic parameter database, and calculate and obtain the characteristic parameter vector of the noise source in the target area; S403: Obtaining the actual position of the noise source in the target area using a sound source positioning sensor; S404: Establishing a target area noise source characteristic vector according to the characteristic parameter vector of the target area noise source.

[0011] Preferably, the target area noise source is marked by: marking the target area feature vector Input into the urban noise source characteristic parameter database, match the characteristic vector of the target area in the urban noise source characteristic parameter database The most similar urban noise source parameters can be marked by marking the corresponding urban noise source parameters to label the noise source type in the target area.

[0012] Preferably, the step S8 further includes the following steps: S801: Using the marked noise source parameters of the target area as training samples of the neural network; S802: Input the sample parameters into the neural network to determine the weights between the nodes of each layer of the neural network Threshold , and use the test sample set to test the neural network and determine the error function of the neural network ; S803: Output actual value according to neural network , calculate the error function , and adjust the threshold and weights , the formula is: ; ; In the formula, is the number of iterations, is the training learning rate; when hour, ; hour, ; In an embodiment of the present invention, when the output error is equal to the predefined error ,Right now , the training is completed and the iteration is exited; otherwise, when the output error is less than the preset error, the sample data is input into the neural network for training, and when the output error is greater than the preset error, the process proceeds to step S802; S804: After the training is completed, an intelligent recognition model of noise sources in the target area is obtained.

[0013] Preferably, the step S802 further includes the following steps: S802a: For each training sample feature parameter vector and output weights The formula for the product is: + ; In the formula, For each training sample feature parameter vector and output weights The product of is the feature vector for each target area; S802b: According to , get the actual value of the neural network output ; If adopted Function as activation function, then , where Expressed as a natural constant; Among them, when the actual value and at the moment Time corresponds to The predicted value of samples satisfy ; Actual value and at the moment The predicted value at satisfy When , set the iteration termination condition, otherwise, go to step S803.

[0014] Preferably, when establishing the noise source intelligent identification model, an intelligent classification system is established, and the intelligent classification system consists of a sensor layer, a collection and processing layer, and an application layer. The intelligent classification system inputs the target area noise source parameters into the application layer, and the application layer compares the target area noise source parameters with the established target area noise source intelligent identification model to identify the target noise source type. .

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention analyzes and identifies six types of noise sources in the city, establishes a noise source database and a noise source basic characteristic parameter database, and determines a standard data set of noise source characteristic parameters. By calculating the noise source radiation value and characteristic vector, a neural network is established, and the best neural network model is extracted according to the accuracy, thereby establishing a noise source intelligent recognition model, which can assist in the identification of different noise sources, realize the extraction of characteristic parameters of different noise sources, quantitatively calculate the noise radiation values ​​of different noise sources, and use the neural network algorithm to automatically construct a noise source intelligent recognition model. In actual detection, the characteristic vector of the noise source is obtained by calculation, and the characteristic vector is input into the noise source intelligent recognition model, so that the noise source type can be intelligently identified, and a higher recognition rate can be obtained.

[0016] 2. The present invention quantifies the characteristics of different sound sources within the city, extracts the characteristics of urban noise sources, establishes a noise source basic characteristic parameter database based on different noise sources, and realizes noise source identification by establishing an intelligent noise source identification model, thereby reducing the difficulty of noise source identification. This method has practical significance for urban noise source identification.

[0017] 3. Based on the collection of sound source data, the present invention constructs a noise source database and a basic characteristic parameter database of urban noise, and performs data processing to obtain data sets of different types of noise sources. The data of different noise sources are analyzed and a neural network is constructed. This method solves the problem of difficulty in identifying and classifying noise sources in urban noise monitoring, and has practical significance for the classification monitoring of urban noise sources. DETAILED DESCRIPTION

[0018] The present invention relates to a method for intelligently identifying urban noise sources based on a neural network, comprising the following steps: S1: Establish urban noise source database; S101: Collect urban noise source data; In an embodiment of the present invention, the noise source data includes the name of the sound source, the category of the sound source, the attribute of the sound source, the location of the sound source, and the shape of the sound source; In an embodiment of the present invention, the noise source data is divided into six types according to the sound source category attribute: traffic noise source, industrial noise source, construction noise source, sanitation machinery noise source, entertainment noise source, and human voice noise source; S102: Classify the sound source data into noise sources and non-noise sources according to sound source attributes; S103: Determine the relative position of each noise source in the city according to the position of the sound source; In an embodiment of the present invention, the sound source location includes the longitude and latitude of the sound source's geographical location and its specific location in the city; S104: Determine the area of ​​the noise source according to the shape of the sound source; S2: Establish a database of basic characteristic parameters of urban noise; In an embodiment of the present invention, the basic characteristic parameter database of urban noise includes wind speed, air temperature, temperature and humidity, noise source height, vehicle displacement, mechanical noise source power, vehicle speed and crowd size; In the embodiment of the present invention, the noise source height refers to the distance between the height of the mechanical noise source and the ground. Taking the construction noise source as an example, it specifically refers to the distance between the height of the mechanical noise source and the ground corresponding to the maximum noise level generated by the noise source point; the temperature and humidity refer to the temperature and humidity values ​​of the environment around the noise source. Taking the construction noise source as an example, it specifically refers to the temperature and humidity values ​​of the environment corresponding to the maximum noise level generated by the mechanical noise source point; the speed of the motor vehicle is obtained through the traffic monitoring equipment; Based on the collection of sound source data, the present invention constructs a noise source database and a basic characteristic parameter database of urban noise, performs data processing, obtains data sets of different types of noise sources, analyzes the data of different noise sources, and constructs a neural network. This method solves the problem of difficulty in identifying and classifying noise sources in urban noise monitoring, and has practical significance for classified monitoring of urban noise sources.

[0019] S3: Establish a standard data set of noise source characteristic parameters; Based on the above urban noise source database, the collected urban noise source data are analyzed to determine the characteristic parameters of the noise sources; S301: Determine a standard data set of noise source characteristic parameters corresponding to radiation values ​​of different types of noise sources: Traffic noise source radiation value: , where is the radiation value of the traffic noise source, indicating the sound power level radiated by the source point. Indicates the sound source power corresponding to the maximum value of the noise source, represents the vehicle speed corresponding to the maximum value of the noise source, Indicates the temperature and humidity corresponding to the maximum value of the noise source, They represent the temperature and humidity constants corresponding to the traffic noise sources, Indicates that the noise source is spatially Axis and The axis corresponds to the height; Radiation values ​​of industrial noise sources: , where is the radiation value of the industrial noise source, Indicates the mechanical power or electrical power corresponding to the maximum value of the noise source; Radiation value of construction noise source: , where is the radiation value of the construction noise source, They represent the temperature and humidity constants corresponding to the construction noise sources respectively; Radiation value of noise source of sanitation machinery: , where They represent the temperature and humidity constants corresponding to the noise sources of sanitation machinery respectively; Entertainment noise source radiation value: , where is the radiation value of the entertainment noise source, Indicates the number of people corresponding to the maximum value of this type of noise source; Radiation value of human voice noise source: , where Is the radiation value of the human voice noise source; S302: Determine the distance between the noise source and the sensor, establish a standard data set of radiation values ​​between the sensor and the noise source to be measured, and obtain the distance range between the noise source and the sensor in the standard data set of radiation values ​​between the sensor and the noise source to be measured; S4: Determine the target area and build an urban noise monitoring system in the center of the target area. A noise monitoring network is constructed by using a sound source localization sensor, and the characteristic parameter vector of the noise source in the target area is measured by the sound source localization sensor; S401: First, determine the target area, use sound source positioning sensors to build a noise monitoring network, determine the number and location of sensors, and measure each sound source positioning sensor to obtain characteristic parameters of the noise source in the target area; In the embodiment of the present invention, the characteristic parameters of the noise source in the target area include the noise frequency and the noise level. When measuring the noise level, the sampling time needs to be selected. , during which the maximum sound level in the area is monitored and recorded as ; S402: Input the obtained characteristic parameters into the noise source basic characteristic parameter database, and calculate and obtain the characteristic parameter vector of the noise source in the target area; S403: Obtaining the actual position of the noise source in the target area using a sound source positioning sensor; S404: establishing a target area noise source characteristic vector according to the characteristic parameter vector of the target area noise source; The present invention quantifies the characteristics of different sound sources in the city, extracts the characteristics of urban noise sources, establishes a noise source basic characteristic parameter database based on different noise sources, and realizes noise source identification by establishing an intelligent noise source identification model, thereby reducing the difficulty of noise source identification. This method has practical significance for urban noise source identification.

[0020] S5: Calculate the radiation value of the noise source in the target area. The formula is: ; In the formula, is the sampling time, Locate the sensor at the location of the sound source The time when the maximum noise is detected, Locate the sensor at the location of the sound source The maximum noise detected at is the weight coefficient; S6: Calculate the target area feature vector , , where The geographical coordinates of the sensor locating the sound source, is the radiation value of the noise source in the target area, is the type of noise source in the target area; S7: Mark the noise sources in the target area; The target area feature vector Input into the urban noise source characteristic parameter database, match the characteristic vector of the target area in the urban noise source characteristic parameter database The most similar urban noise source parameters, marking the corresponding urban noise source parameters, can mark the noise source type of the target area; S8: Establish a neural network based on the marked noise source parameters of the target area and establish a noise source intelligent identification model; S801: Using the marked noise source parameters of the target area as training samples of the neural network; S802: Input the sample parameters into the neural network to determine the weights between the nodes of each layer of the neural network Threshold , and use the test sample set to test the neural network and determine the error function of the neural network ; S802a: For each training sample feature parameter vector and output weights The formula for the product is: + ; In the formula, For each training sample feature parameter vector and output weights The product of is the feature vector for each target area; S802b: According to , get the actual value of the neural network output ; If adopted Function as activation function, then , where Expressed as a natural constant; When the actual value and at the moment Time corresponds to The predicted value of samples satisfy ; Actual value and at the moment The predicted value at satisfy When , set the iteration termination condition, otherwise, go to step S803; S803: Output actual value according to neural network , calculate the error function , and adjust the threshold and weights , the formula is: ; ; In the formula, is the number of iterations, is the training learning rate; when hour, ; hour, ; In an embodiment of the present invention, when the output error is equal to the predefined error ,Right now , the training is completed and the iteration is exited; otherwise, when the output error is less than the preset error, the sample data is input into the neural network for training, and when the output error is greater than the preset error, the process proceeds to step S802; S804: After the training is completed, an intelligent recognition model of noise sources in the target area is obtained; S9: In the target area, inputting the noise source parameters of the target area into the noise source intelligent identification model; In an embodiment of the present invention, when establishing the intelligent recognition model of the noise source in the target area, an intelligent classification system is established, and the intelligent classification system is composed of a sensor layer, a collection and processing layer, and an application layer.

[0021] In an embodiment of the present invention, the intelligent classification system inputs the target area noise source parameters into the application layer, and the application layer compares the target area noise source parameters with the established target area noise source intelligent identification model to identify the target noise source type. .

[0022] The embodiments of the present invention disclose preferred embodiments, but are not limited thereto. A person skilled in the art can easily understand the spirit of the present invention based on the above embodiments and make different extensions and changes. However, as long as they do not deviate from the spirit of the present invention, they are all within the protection scope of the present invention.

Claims

1. A method for intelligently identifying urban noise sources based on neural networks, characterized in that: The following steps are involved: S1: Establish urban noise source database; S2: Establish a database of basic characteristic parameters of urban noise; S3: Analyze based on the urban noise source database to establish a standard data set of noise source characteristic parameters and determine the noise source characteristic parameters; S4: Determine the target area, build an urban noise monitoring system in the center of the target area, and measure the characteristic parameter vector of the noise source in the target area; S5: Calculate the radiation value of the noise source in the target area. The formula is: ； In the formula, is the sampling time, Locate the sensor at the location of the sound source The time when the maximum noise is detected, Locate the sensor at the location of the sound source The maximum noise detected at is the weight coefficient; S6: Calculate the target area feature vector , , where The geographical coordinates of the sensor locating the sound source, is the radiation value of the noise source in the target area, is the type of noise source in the target area; S7: Mark the noise sources in the target area; S8: Establish a neural network based on the marked noise source parameters of the target area and establish a noise source intelligent identification model; S9: In the target area, the noise source parameters of the target area are input into the noise source intelligent identification model to identify the type of the target noise source.

2. The method for intelligently identifying urban noise sources based on a neural network according to claim 1, characterized in that: The step S1 also includes the following steps: S101: Collect urban noise source data; The noise source data includes the name of the sound source, the category of the sound source, the attribute of the sound source, the location of the sound source, and the shape of the sound source. According to the attribute of the sound source category, the noise source data is divided into six types: traffic noise source, industrial noise source, construction noise source, sanitation machinery noise source, entertainment noise source, and human voice noise source; S102: Classify the sound source data into noise sources and non-noise sources according to sound source attributes; S103: Determine the relative position of each noise source in the city according to the position of the sound source; The sound source location includes the latitude and longitude of the sound source's geographical location and its specific location in the city. S104: Determine the area of ​​the noise source according to the shape of the sound source.

3. The method for intelligently identifying urban noise sources based on a neural network according to claim 2 is characterized in that: The step S3 also includes the following steps: S301: Determine a standard data set of noise source characteristic parameters corresponding to radiation values ​​of different types of noise sources; S302: Determine the distance between the noise source and the sensor, establish a standard data set of radiation values ​​between the sensor and the noise source to be measured, and obtain the distance range between the noise source and the sensor in the standard data set of radiation values ​​between the sensor and the noise source to be measured.

4. The method for intelligently identifying urban noise sources based on a neural network according to claim 3 is characterized in that: The noise source characteristic parameter standard data set includes: Traffic noise source radiation value: , where is the radiation value of the traffic noise source, indicating the sound power level radiated by the source point. Indicates the sound source power corresponding to the maximum value of the noise source, represents the vehicle speed corresponding to the maximum value of the noise source, Indicates the temperature and humidity corresponding to the maximum value of the noise source, They represent the temperature and humidity constants corresponding to the traffic noise sources, Indicates that the noise source is spatially Axis and The axis corresponds to the height; Radiation values ​​of industrial noise sources: , where is the radiation value of the industrial noise source, Indicates the mechanical power or electrical power corresponding to the maximum value of the noise source; Radiation value of construction noise source: , where is the radiation value of the construction noise source, They represent the temperature and humidity constants corresponding to the construction noise sources respectively; Radiation value of noise source of sanitation machinery: , where They represent the temperature and humidity constants corresponding to the noise sources of sanitation machinery respectively; Entertainment noise source radiation value: , where is the radiation value of the entertainment noise source, Indicates the number of people corresponding to the maximum value of this type of noise source; Radiation value of human voice noise source: , where The radiation value of the human voice noise source.

5. The method for intelligently identifying urban noise sources based on a neural network according to claim 4 is characterized in that: The urban noise monitoring system consists of The noise monitoring network is constructed by using a sound source localization sensor.

6. The method for intelligently identifying urban noise sources based on a neural network according to claim 5, characterized in that: The step S4 also includes the following steps: S401: First, determine the target area, use sound source positioning sensors to build a noise monitoring network, determine the number and location of sensors, and measure each sound source positioning sensor to obtain characteristic parameters of the noise source in the target area; The characteristic parameters of the noise source in the target area include noise frequency and noise level; S402: Input the obtained characteristic parameters into the noise source basic characteristic parameter database, and calculate and obtain the characteristic parameter vector of the noise source in the target area; S403: Obtaining the actual position of the noise source in the target area using a sound source positioning sensor; S404: Establishing a target area noise source characteristic vector according to the characteristic parameter vector of the target area noise source.

7. The method for intelligently identifying urban noise sources based on a neural network according to claim 6, characterized in that: The noise source in the target area is marked in the following way: the feature vector of the target area Input into the urban noise source characteristic parameter database, match the characteristic vector of the target area in the urban noise source characteristic parameter database The most similar urban noise source parameters can be marked by marking the corresponding urban noise source parameters to label the noise source type in the target area.

8. The method for intelligently identifying urban noise sources based on a neural network according to claim 7, characterized in that: The step S8 also includes the following steps: S801: Using the marked noise source parameters of the target area as training samples of the neural network; S802: Input the sample parameters into the neural network to determine the weights between the nodes of each layer of the neural network Threshold , and use the test sample set to test the neural network and determine the error function of the neural network ; S803: Output actual value according to neural network , calculate the error function , and adjust the threshold and weights , the formula is: ； ； In the formula, is the number of iterations, is the training learning rate; when hour, ; hour, ; In an embodiment of the present invention, when the output error is equal to the predefined error ,Right now , training is completed and the iteration is exited; Otherwise, when the output error is less than the preset error, the sample data is input into the neural network for training; when the output error is greater than the preset error, the process proceeds to step S802; S804: After the training is completed, an intelligent recognition model of noise sources in the target area is obtained.

9. The method for intelligently identifying urban noise sources based on a neural network according to claim 8, characterized in that: The step S802 further includes the following steps: S802a: For each training sample feature parameter vector and output weights The formula for the product is: + ； In the formula, For each training sample feature parameter vector and output weights The product of is the feature vector for each target area; S802b: According to , get the actual value of the neural network output ; If adopted Function as activation function, then , where Expressed as a natural constant; Among them, when the actual value and at the moment Time corresponds to The predicted value of samples satisfy ; Actual value and at the moment The predicted value at satisfy When , set the iteration termination condition, otherwise, go to step S803.

10. The method for intelligently identifying urban noise sources based on neural networks according to claim 9, characterized in that: When establishing the noise source intelligent identification model, an intelligent classification system is established. The intelligent classification system consists of a sensor layer, an acquisition processing layer, and an application layer. The intelligent classification system inputs the target area noise source parameters into the application layer. The application layer compares the target area noise source parameters with the established target area noise source intelligent identification model to identify the target noise source type. .

Images