Background noise deduction method for electromagnetic radiation monitoring

Abstract

The invention discloses a background noise deduction method for electromagnetic radiation monitoring, and relates to the technical field of noise processing. Comprising the steps of collecting electromagnetic radiation data from a monitoring area as a training sample; constructing a generative adversarial network which comprises a generator and a discriminator; training a generative adversarial network model by using the target-free electromagnetic radiation data; inputting the monitored data into the generative adversarial network model, generating corresponding background noise estimation by a generator, and deducting a background noise part through combination with the monitored data to obtain a purified electromagnetic radiation signal; and analyzing the electromagnetic radiation data from which the background noise is deducted, evaluating the electromagnetic radiation level of the environment, and generating a detailed monitoring report. The system can adapt to a complex and changeable electromagnetic environment, and provides a more accurate and reliable monitoring result; by accurately deducting background noise, the accuracy and reliability of electromagnetic radiation monitoring data are greatly improved.

Description

Technical Field

[0001] The invention relates to the technical field of noise processing, in particular to a background noise subtraction method for electromagnetic radiation monitoring. Background Art

[0002] Electromagnetic radiation monitoring is easily affected by environmental background noise, which reduces the accuracy and reliability of monitoring data. Existing background noise processing methods, such as Fourier transform and wavelet analysis, usually rely on fixed noise models, which are difficult to adapt to complex and changing electromagnetic environments. Most of them require offline processing and cannot meet the needs of real-time monitoring. Summary of the invention

[0003] In view of the problems existing in the existing scene noise processing method, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is how to use the generative adversarial network model to model and estimate the background noise in the electromagnetic environment in real time, and improve the accuracy and reliability of the monitoring data by deducting the estimated background noise.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In the first aspect, an embodiment of the present invention provides a background noise subtraction method for electromagnetic radiation monitoring, which includes collecting electromagnetic radiation data from a monitoring area as training samples; constructing a generative adversarial network, including a generator and a discriminator; using data without target electromagnetic radiation to train the generative adversarial network model; inputting the monitored data into the generative adversarial network model, the generator generates a corresponding background noise estimate, and by combining it with the monitoring data, subtracting the background noise part to obtain a purified electromagnetic radiation signal; analyzing the electromagnetic radiation data after subtracting the background noise, evaluating the environmental electromagnetic radiation level, and generating a detailed monitoring report.

[0007] As a preferred solution of the background noise subtraction method for electromagnetic radiation monitoring of the present invention, the generator G is constructed including: the input layer receives a random noise vector z, whose dimension is d, that is, z∈R d ; The second hidden layer contains h 2 neurons, using the ReLU activation function; the output layer has the same dimension as the background noise data, and uses the tanh activation function to generate the virtual background noise G(z).

[0008] As a preferred solution of the background noise subtraction method for electromagnetic radiation monitoring of the present invention, the discriminator D is constructed including: the input layer receives the real background noise data x or the generated virtual noise data G(z); the first hidden layer is implemented by a fully connected layer, including h′ 1neurons, using the Leaky ReLU activation function for processing; the second hidden layer structure contains h′ 2 neurons, using the Leaky ReLU activation function; the output layer is a fully connected layer, which outputs a probability value D(x) and uses the sigmoid activation function to determine the authenticity of the data.

[0009] As a preferred solution of the background noise subtraction method for electromagnetic radiation monitoring of the present invention, the training of the generative adversarial network model using data without target electromagnetic radiation includes the following steps: training the generative adversarial network model using monitoring data containing only background noise, and the training process is achieved by optimizing the following objective function:

[0010]

[0011] in, is the real noise data distribution, is the prior distribution of random noise.

[0012] As a preferred solution of the background noise subtraction method for electromagnetic radiation monitoring of the present invention, the training of the generative adversarial network model includes: using monitoring data x containing only background noise as a training set, assuming that the background noise data follows the distribution P r (x), that is, x~P r (x); from the real data distribution P r (x) randomly extracts a part of the background noise data x, and at the same time extracts a part of the background noise data x from the random noise distribution P z (z), extract the corresponding number of noise vectors z, input these noise vectors into the generator G, and generate the corresponding virtual noise data G(z); the discriminator D receives a part of the real background noise data x and a part of the generated virtual noise data G(z), and calculates the probability that each input data is the real data respectively; define the loss function L of the discriminator D , by minimizing L D , improve the discriminator's ability to distinguish real noise from generated noise; when training the generator G, the parameters of the discriminator D remain unchanged, and the generator accepts the noise distribution P z (z) and generates the corresponding virtual noise G(z); during the adversarial training process, the discriminator and the generator are trained alternately.

[0013] As a preferred solution of the background noise removal method for electromagnetic radiation monitoring of the present invention, the operation steps of obtaining the purified electromagnetic radiation signal are as follows: t Input to the trained generator G, the generator outputs the background noise estimate G(x t ); From the original monitoring data xt Deduct the generated background noise G(x t ), and obtain the purified electromagnetic radiation signal s t ; For the purified signal s t Perform amplitude correction to restore the true amplitude of the signal.

[0014] As a preferred solution of the background noise subtraction method for electromagnetic radiation monitoring described in the present invention, the analysis of the electromagnetic radiation data after background noise subtraction and the assessment of the environmental electromagnetic radiation level include: calculating the root mean square value of the signal to reflect the overall radiation intensity; performing spectrum analysis on the signal to obtain a power spectrum density curve to identify the frequency components of the main radiation source; statistically analyzing the energy distribution ratio of the signal in different frequency bands; analyzing the time-frequency characteristics of the signal to obtain a time-frequency spectrum to characterize the dynamic change law of radiation.

[0015] In the second aspect, an embodiment of the present invention provides a background noise subtraction system for electromagnetic radiation monitoring, which includes: an electromagnetic radiation data acquisition module, which is used to collect electromagnetic radiation data from a monitoring area as a training sample; a generative adversarial network construction module, which is used to construct a generative adversarial network, including a generator and a discriminator; a non-target electromagnetic radiation data training module, which is used to train a generative adversarial network model using non-target electromagnetic radiation data; a background noise subtraction module, which is used to input the monitored data into the generative adversarial network model, and the generator generates a corresponding background noise estimate, and by combining it with the monitoring data, subtracts the background noise part to obtain a purified electromagnetic radiation signal; a data analysis and report generation module, which is used to analyze the electromagnetic radiation data after deducting the background noise, evaluate the environmental electromagnetic radiation level, and generate a detailed monitoring report.

[0016] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the background noise subtraction method for electromagnetic radiation monitoring as described in the first aspect of the present invention are implemented.

[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the background noise subtraction method for electromagnetic radiation monitoring as described in the first aspect of the present invention are implemented.

[0018] The beneficial effects of the present invention are as follows: the present invention can model and estimate the background noise in the electromagnetic environment in real time, thereby realizing instant purification of electromagnetic radiation monitoring data. Compared with the traditional method that requires offline processing, the present invention can adapt to complex and changeable electromagnetic environments and provide more accurate and reliable monitoring results; by accurately deducting background noise, this patent greatly improves the accuracy and reliability of electromagnetic radiation monitoring data; the present invention can significantly reduce the interference of environmental noise on electromagnetic radiation monitoring and optimize the monitoring process. This not only improves the efficiency of data processing, but also provides a reliable basis for further data analysis and environmental assessment; the present invention is not limited to specific types of electromagnetic radiation or specific environments, and its powerful adaptive ability enables it to be widely used in various electromagnetic radiation monitoring scenarios, including but not limited to industry, medical care, environmental protection and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 Flowchart of the background noise subtraction method for electromagnetic radiation monitoring based on generative adversarial networks

[0021] Figure 2 Schematic diagram of the generative adversarial network model structure.

[0022] Figure 3 Schematic diagram of the comparison of denoising effects of generative adversarial networks on electromagnetic radiation data. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example 1

[0027] Reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, and provides a background noise subtraction method for electromagnetic radiation monitoring, such as Figure 1 As shown, including,

[0028] S1: Collect electromagnetic radiation data from the monitoring area as training samples, including data with target electromagnetic radiation and data without target electromagnetic radiation (only background noise).

[0029] Electromagnetic radiation data is collected in the monitoring area, including data with and without target radiation, and the collected data is divided into two categories: data containing target radiation and data containing only background noise; the data is preprocessed by denoising, normalization, etc., and key parameters reflecting signal and noise characteristics are extracted.

[0030] S2: Construct a generative adversarial network, in which the generator is responsible for generating data similar to the real background noise, and the discriminator is responsible for distinguishing whether the input data comes from the real background noise or the data generated by the generator.

[0031] Preferably, construct Figure 2 The generative adversarial network model shown in the figure includes a generator G and a discriminator D. The generator G receives random noise z as input and generates virtual background noise G(z) through a multi-layer fully connected network. The discriminator D receives real or generated noise data and discriminates the authenticity probability D(x) of the data through a multi-layer fully connected network, as follows:

[0032] The specific details of constructing the generator G and the discriminator D are as follows:

[0033] The architecture design of the generator G first considers the input layer, which receives a random noise vector z with dimension d. Here, z represents the input of the generator, which is usually a random noise vector sampled from a standard normal distribution N(0,1), and d is the dimension of the noise vector, that is, z∈R d .

[0034] After the input layer, the first hidden layer consists of a fully connected layer containing h 1 neurons and use the ReLU activation function for nonlinear transformation. Here, h 1 Represents the number of neurons in the first hidden layer of the generator. Its calculation formula is:

[0035] h 1 =ReLU(W 1 z+b 1 )

[0036] Among them, W 1is the weight matrix connecting the input layer and the first hidden layer, b 1 is the bias vector for the first hidden layer.

[0037] Then, the second hidden layer is also a fully connected layer, containing h 2 neurons, and the ReLU activation function is also used. The calculation formula is:

[0038] h 2 =ReLU(W 2 h 1 +b 2 )

[0039] Among them, W 2 is the weight matrix connecting the first hidden layer and the second hidden layer, b 2 is the bias vector for the second hidden layer.

[0040] Finally, the output layer has the same dimension as the background noise data, and the tanh activation function is used to generate the virtual background noise G(z), which is calculated as:

[0041] G(z)=tanh(W 2 h 2 +b 3 )

[0042] Here, W 2 is the weight matrix connecting the second hidden layer and the output layer, b 3 is the bias vector of the output layer, and G(z) is the output of the generator, which represents the generated virtual background noise data.

[0043] The architecture design of the discriminator D also starts from the input layer, which receives the real background noise data x or the generated virtual noise data G(z). Here, x represents the input of the discriminator, which can be the real background noise data or the noise data generated by the generator.

[0044] The first hidden layer is implemented by a fully connected layer, including h′ 1 neurons, using the Leaky ReLU activation function for processing, the calculation formula is:

[0045] h′ 1 =Leaky ReLU(W′ 1 x+b′ 1 )

[0046] Among them, W′ 1 is the weight matrix connecting the input layer and the first hidden layer, b′ 1 is the bias vector for the first hidden layer.

[0047] The second hidden layer structure is similar to the first hidden layer, including h′ 2 neurons, using the Leaky ReLU activation function, the calculation formula is:

[0048] h′ 2 =Leaky ReLU(W′ 2 h′ 1 +b′ 2 )

[0049] Among them, W′ 2 is the weight matrix connecting the first hidden layer and the second hidden layer, b′ 2 is the bias vector for the second hidden layer.

[0050] The final output layer is a fully connected layer that outputs a probability value D(x) and uses the sigmoid activation function to determine the authenticity of the data. The calculation formula is:

[0051] D(x)=sigmoid(W′ 3 h′ 2 +b′ 3 )

[0052] Among them, W′ 3 is the weight matrix connecting the second hidden layer and the output layer, b′ 3 is the bias vector of the output layer, and D(x) represents the output of the discriminator, that is, the probability that the input data is the real background noise.

[0053] Training steps for the discriminator and generator:

[0054] During the training process, the discriminator D is first trained. To this end, the real data distribution P r A batch of background noise data x is randomly extracted from (x), and a batch of random noise z is extracted from the standard normal distribution N(0,1), and the corresponding virtual background noise G(z) is generated through the generator. The real background noise data x and the generated virtual noise G(z) are input into the discriminator D respectively, and then the loss function L of the discriminator is calculated D , and use the gradient descent method to update the parameters of the discriminator. The loss function L D The calculation formula is as follows:

[0055]

[0056] Among them, m is the batch size, which means the number of data samples processed by the generator and discriminator respectively in each training.

[0057] Next, fix the parameters of the discriminator and train the generator G. To this end, new random noise z is extracted from the standard normal distribution N(0,1) again, and the virtual background noise G(z) is generated through the generator. Calculate the loss function L of the generator G , and use the gradient descent method to update the parameters of the generator. The loss function L of the generator G The calculation formula is as follows:

[0058]

[0059] During the entire training process, the discriminator and generator are trained alternately until the generator can generate virtual noise that is indistinguishable from real background noise. Through this adversarial training method, the generator and discriminator eventually achieve optimal performance in their respective tasks.

[0060] Notes in implementation: The number of neurons and layers in the network architecture of the generator and discriminator are determined according to the complexity and requirements of the data. The specific values ​​are the optimal structures verified by actual experiments. In order to ensure the stability of training, the present invention uses batch normalization and label smoothing technology in implementation to improve the convergence of the model.

[0061] S3: Use data without target electromagnetic radiation to train a generative adversarial network model, so that the generator can highly simulate the real background noise characteristics.

[0062] Specifically, the generative adversarial network model is trained using monitoring data containing only background noise, so that the generator and the discriminator can conduct adversarial learning, and finally generate background noise data that is indistinguishable from the real thing. The training process is achieved by optimizing the following objective function:

[0063]

[0064] in, is the real noise data distribution, is the prior distribution of random noise.

[0065] In this step, the model training process of the Generative Adversarial Network (GAN) will be introduced in detail, including data preparation, discriminator training, generator training, and specific methods of adversarial training.

[0066] First, in the data preparation stage, monitoring data x containing only background noise is used as the training set. Assume that these real background noise data follow the distribution P r (x), that is, x~P r (x). The input of the generator is from a random noise distribution P z The noise vector z sampled in (z), i.e. z~P z (z).

[0067] In the training process, we first need to train the discriminator D. From the real data distribution P r (x) randomly extracts a batch of background noise data x, and at the same time extracts a batch of background noise data x from the random noise distribution P z (z). These noise vectors are input into the generator G to generate the corresponding virtual noise data G(z). Next, the discriminator D receives a part of the real background noise data x and a part of the generated virtual noise data G(z), and calculates the probability that each input data is the real data. The loss function L of the discriminator is D Defined as:

[0068] L D = -[logD(x)+log(1-D(G(z)))]

[0069] In this formula, D(x) represents the output of the discriminator for the real background noise data x, that is, the probability that it is judged to be real noise; D(G(z)) represents the output of the discriminator for the generated noise G(z), that is, the probability that it is judged to be real noise.

[0070] By minimizing L D , which can improve the discriminator's ability to distinguish between real noise and generated noise.

[0071] To update the discriminator parameters θ D , using gradient descent, the specific update formula is:

[0072]

[0073] Among them, α is the learning rate, is the gradient of the loss function with respect to the discriminator parameters.

[0074] When training the generator G, the parameters of the discriminator D remain unchanged. The generator receives z (z) and generates the corresponding virtual noise G(z).

[0075] The goal of the generator is to deceive the discriminator into thinking that the generated noise comes from real data. The loss function of the generator is L G Defined as:

[0076] L G = -log(D(G(z)))

[0077] In this formula, D(G(z)) is the probability of the discriminator to generate noise; by minimizing L G , the generator can generate data that is closer to the real background noise. The parameters of the generator θ G Updated by the following formula:

[0078]

[0079] in, is the gradient of the loss function with respect to the generator parameters.

[0080] During the entire adversarial training process, the discriminator and the generator are trained alternately. The discriminator tries to maximize its loss function L G , thereby improving its ability to distinguish between real data and generated data; the generator attempts to minimize its loss function L G , to generate more realistic background noise data. The training process continues until the generator can generate virtual noise data that is indistinguishable from real background noise, that is, the GAN model reaches a convergence state.

[0081] S4: During the electromagnetic radiation monitoring process, the monitored data is input into the generative adversarial network model, and the generator generates a corresponding background noise estimate. By combining it with the monitoring data, the background noise part is deducted to obtain the purified electromagnetic radiation signal.

[0082] Preferably, the trained generative adversarial network model is used to monitor the real-time data x t Perform background noise subtraction. Input it into the generator G to obtain the background noise estimate G(x t ), then deduct it from the original data to obtain the purified electromagnetic radiation signal, and finally perform amplitude correction on the signal after noise deduction to restore the true amplitude of the signal, specifically including:

[0083] Background noise estimation: The real-time monitoring data x t Input to the trained generator G, the generator outputs the background noise estimate G(x t ).

[0084] Noise subtraction: From the original monitoring data x t Deduct the generated background noise G(x t ), and obtain the purified electromagnetic radiation signal s t :

[0085] s t =x t -G(x t )

[0086] Amplitude correction: In order to ensure the physical authenticity of the signal, the purified signal s t For amplitude correction, the following linear correction methods can usually be used:

[0087]

[0088] Among them, β and γ are correction coefficients obtained by fitting experimental data.

[0089] S5: Analyze the electromagnetic radiation data after deducting the background noise, evaluate the environmental electromagnetic radiation level, and generate a detailed monitoring report.

[0090] Preferably, the electromagnetic radiation signal after background noise is deducted is analyzed to evaluate the radiation level in the monitoring area.

[0091] The main analysis contents include: calculating the root mean square value of the signal to reflect the overall radiation intensity; performing spectrum analysis on the signal to obtain the power spectrum density curve and identify the frequency components of the main radiation source; statistically analyzing the energy distribution ratio of the signal in different frequency bands; analyzing the time-frequency characteristics of the signal to obtain the time-frequency spectrum and characterize the dynamic change law of radiation. The specific corresponding formulas are as follows:

[0092] The RMS value is calculated to reflect the overall radiation intensity:

[0093]

[0094] Where N is the number of sampling points, s t It is the purified electromagnetic radiation signal.

[0095] Power spectral density (PSD) analysis to identify the frequency content of the main radiating sources:

[0096]

[0097] in, is the Fourier transform result, and f is the frequency.

[0098] Energy distribution ratio, statistical signal energy distribution in different frequency bands:

[0099]

[0100] Among them, E(f 1 ,f 2 ) indicates that in the frequency band [f 1 ,f 2 ]’s total energy.

[0101] Time-frequency analysis: analyze the time-frequency characteristics of the signal and obtain the time-frequency spectrum:

[0102] STFT t}(t,f)=∫s(τ)w(t-τ)e -j2πfτ dτ

[0103] Where w(t) is the window function, t is time, and f is frequency.

[0104] According to relevant electromagnetic radiation standards, evaluate the impact of current radiation levels on human health and give a safety level. Based on the above analysis, generate an electromagnetic radiation monitoring report, use charts and other visual methods to intuitively present the analysis results, and give improvement suggestions.

[0105] This embodiment also provides a computer device, which is suitable for the background noise subtraction method of electromagnetic radiation monitoring, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the background noise subtraction method of electromagnetic radiation monitoring proposed in the above embodiment.

[0106] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device 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 and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0107] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the background noise subtraction method for electromagnetic radiation monitoring proposed in the above embodiment is implemented.

[0108] In summary, the present invention can model and estimate the background noise in the electromagnetic environment in real time, thereby realizing instant purification of electromagnetic radiation monitoring data. Compared with the traditional method that requires offline processing, the present invention can adapt to complex and changeable electromagnetic environments and provide more accurate and reliable monitoring results; by accurately deducting background noise, this patent greatly improves the accuracy and reliability of electromagnetic radiation monitoring data; the present invention can significantly reduce the interference of environmental noise on electromagnetic radiation monitoring and optimize the monitoring process. This not only improves the efficiency of data processing, but also provides a reliable basis for further data analysis and environmental assessment; the present invention is not limited to specific types of electromagnetic radiation or specific environments, and its powerful adaptive ability enables it to be widely used in various electromagnetic radiation monitoring scenarios, including but not limited to industry, medical care, environmental protection and other fields.

[0109] Example 2

[0110] The present embodiment provides a background noise subtraction system for electromagnetic radiation monitoring, including an electromagnetic radiation data acquisition module, which is used to collect electromagnetic radiation data from a monitoring area as a training sample; a generative adversarial network construction module, which is used to construct a generative adversarial network, including a generator and a discriminator; a non-target electromagnetic radiation data training module, which is used to train a generative adversarial network model using non-target electromagnetic radiation data; a background noise subtraction module, which is used to input the monitored data into the generative adversarial network model, and the generator generates a corresponding background noise estimate, and by combining it with the monitoring data, subtracts the background noise part to obtain a purified electromagnetic radiation signal; a data analysis and report generation module, which is used to analyze the electromagnetic radiation data after the background noise is subtracted, evaluate the environmental electromagnetic radiation level, and generate a detailed monitoring report.

[0111] Example 3

[0112] Reference Figure 3 , which is the third embodiment of the present invention, and this embodiment provides a background noise subtraction method for electromagnetic radiation monitoring. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through comparative experiments.

[0113] like Figure 3 As shown, the grayscale curve in the figure is shown in the legend to distinguish three types of curves. Noisy Signal refers to the original data with background noise collected in the monitoring area, Denoised Signal is the data after the background noise is deducted by the method of the present invention, and True Signal refers to the theoretical value of the monitoring data in this area.

[0114] pass Figure 3 It can be found that the Noisy Signal is the original data with background noise, and the data is messy and difficult to analyze. The Denoised Signal data obtained by this method after background noise subtraction is close to the theoretical value data, which better restores the real data obtained by measurement.

[0115] It can be seen that compared with the traditional method that requires offline processing, the present invention can adapt to complex and changeable electromagnetic environments and provide more accurate and reliable monitoring results.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A background noise subtraction method for electromagnetic radiation monitoring, characterized in that: include, Collect electromagnetic radiation data from the monitoring area as training samples; Build a generative adversarial network, including a generator and a discriminator; Use data from non-target electromagnetic radiation to train a generative adversarial network model; The monitored data is input into the generative adversarial network model, and the generator generates a corresponding background noise estimate. By combining it with the monitored data and deducting the background noise part, a purified electromagnetic radiation signal is obtained; Analyze the electromagnetic radiation data after deducting background noise, evaluate the environmental electromagnetic radiation level, and generate a detailed monitoring report.

2. The background noise subtraction method for electromagnetic radiation monitoring according to claim 1, characterized in that: Constructing the generator G includes: The input layer receives a random noise vector z with dimension d, i.e. z∈R d ; The second hidden layer contains h2 neurons and uses the ReLU activation function; The output layer has the same dimension as the background noise data, and the tanh activation function is used to generate virtual background noise G(z).

3. The background noise subtraction method for electromagnetic radiation monitoring according to claim 2, characterized in that: Constructing the discriminator D includes: The input layer receives the real background noise data x or the generated virtual noise data G(z); The first hidden layer is implemented by a fully connected layer, containing h′1 neurons, and processed using the Leaky ReLU activation function; The second hidden layer structure contains h′2 neurons and uses the Leaky ReLU activation function; The output layer is a fully connected layer that outputs a probability value D(x) and uses the sigmoid activation function to determine the authenticity of the data.

4. The background noise subtraction method for electromagnetic radiation monitoring according to claim 3, characterized in that: The method of using the data of non-target electromagnetic radiation to train the generative adversarial network model comprises the following steps: The generative adversarial network model is trained using monitoring data containing only background noise. The training process is achieved by optimizing the following objective function: in, is the real noise data distribution, is the prior distribution of random noise.

5. The background noise subtraction method for electromagnetic radiation monitoring according to claim 4, characterized in that: The training of the generative adversarial network model includes: Use monitoring data x containing only background noise as the training set, assuming that the background noise data follows the distribution P r (x), that is, x~P r (x); From the real data distribution P r (x) randomly extracts a part of the background noise data x, and at the same time extracts a part of the background noise data x from the random noise distribution P z (z), extract the corresponding number of noise vectors z, input these noise vectors into the generator G, and generate the corresponding virtual noise data G(z); The discriminator D receives a part of the real background noise data x and a part of the generated virtual noise data G(z), and calculates the probability that each input data is the real data respectively; Define the loss function L of the discriminator D , by minimizing L D , improve the discriminator's ability to distinguish between real noise and generated noise; When training the generator G, the parameters of the discriminator D remain unchanged, and the generator accepts the noise distribution P z (z) and generates the corresponding virtual noise G(z); During adversarial training, the discriminator and generator are trained alternately.

6. The background noise subtraction method for electromagnetic radiation monitoring according to claim 5, characterized in that: The operation steps of obtaining the purified electromagnetic radiation signal are as follows: Real-time monitoring data x t Input to the trained generator G, the generator outputs the background noise estimate G(x t ); From the original monitoring data x t Deduct the generated background noise G(x t ), and obtain the purified electromagnetic radiation signal s t ; After purification, the signal t Perform amplitude correction to restore the true amplitude of the signal.

7. The background noise subtraction method for electromagnetic radiation monitoring according to claim 6, characterized in that: The analyzing of electromagnetic radiation data after background noise deduction and evaluation of the environmental electromagnetic radiation level comprises: Calculate the RMS value of the signal to reflect the overall radiation intensity; Perform spectrum analysis on the signal to obtain the power spectrum density curve and identify the frequency of the main radiation source; Statistical signal energy distribution ratio in different frequency bands; Analyze the time-frequency characteristics of the signal, obtain the time-frequency spectrum, and characterize the dynamic change law of radiation.

8. A background noise subtraction system for electromagnetic radiation monitoring, based on the background noise subtraction method for electromagnetic radiation monitoring according to any one of claims 1 to 7, characterized in that: Also includes, An electromagnetic radiation data acquisition module is used to collect electromagnetic radiation data from the monitoring area as training samples; Generative adversarial network building module, used to build a generative adversarial network, including a generator and a discriminator; A non-target electromagnetic radiation data training module, used to train a generative adversarial network model using non-target electromagnetic radiation data; A background noise subtraction module is used to input the monitored data into the generative adversarial network model, and the generator generates a corresponding background noise estimate, and by combining it with the monitored data, the background noise part is subtracted to obtain a purified electromagnetic radiation signal; The data analysis and report generation module is used to analyze the electromagnetic radiation data after deducting the background noise, evaluate the environmental electromagnetic radiation level, and generate a detailed monitoring report.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the background noise subtraction method for electromagnetic radiation monitoring described in 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 background noise subtraction method for electromagnetic radiation monitoring described in any one of claims 1 to 7 are implemented.

Images