Interference component processing method and device, equipment, storage medium and product

Abstract

The invention discloses a signal interference component processing method and device, equipment, a storage medium and a product, and relates to the technical field of 5G signals, and the method comprises the steps: obtaining an initial 5G signal, and preprocessing the initial 5G signal to obtain a real-time 5G signal; extracting a time domain feature, a frequency domain feature and a topological feature from the real-time 5G signal feature through a multi-stage generative adversarial network model to obtain a discrimination result; marking an interference signal in the real-time 5G signal according to a judgment result to obtain an interference component; performing adaptive filtering on the interference component to obtain a target 5G signal; according to the method, the difference between a normal 5G signal and an interference signal is identified layer by layer by comparing a simulation interference signal with actual 5G signal data through a multi-stage generative adversarial network model discriminator, and a generator and the discriminator are enabled to play a game on the basis of adversarial training, so that the accuracy and efficiency of interference detection are improved; time domain, frequency domain and topological features are combined, so that the discriminator analyzes and identifies interference signals more comprehensively, and the interference detection precision is improved.

Description

Technical Field

[0001] The present application relates to the field of 5G signal technology, and in particular to interference component processing methods, devices, equipment, storage media and products. Background Art

[0002] In the development of 5G communication technology, due to its high frequency band and wide bandwidth characteristics, 5G signals are more susceptible to various interferences, including adjacent channel interference, co-channel interference and environmental noise. These interferences seriously affect the quality and system performance of 5G communications. Traditional interference detection and elimination methods mainly rely on fixed algorithms and rules, which are difficult to flexibly cope with complex and changing interference environments, thus limiting the stability and reliability of 5G networks.

[0003] In the prior art, interference detection usually adopts methods such as spectrum analysis, power spectrum density estimation, and time-frequency analysis. These methods have certain effects when dealing with simple interference, but they are often powerless when faced with complex and diverse interference signals. For example, the spectrum analysis method detects interference by observing abnormal components in the signal spectrum, but in a crowded spectrum environment, it is difficult to accurately distinguish normal signals from interference signals. The power spectrum density estimation method identifies interference by calculating the changes in the power spectrum density of the signal, but this method has low sensitivity to interference signals with a high dynamic range. The time-frequency analysis method combines time domain and frequency domain features for interference detection, but the computational complexity is high and it is difficult to meet the requirements of real-time processing.

[0004] These defects severely limit the performance and user experience of 5G networks. Therefore, there is an urgent need for a 5G signal processing method that can efficiently, accurately and in real time detect and eliminate complex interference to improve the stability and reliability of 5G networks. Summary of the invention

[0005] The main purpose of this application is to provide an interference component processing method, device, equipment, storage medium and product, aiming to solve the technical problem that it is difficult to accurately distinguish normal signals from interference signals when facing diversified interference signals, and it is difficult to accurately identify various interference signals, resulting in unsatisfactory noise reduction effect and affecting the stability and reliability of 5G networks.

[0006] To achieve the above objectives, the present application proposes a method for processing signal interference components, the method comprising:

[0007] Acquire an initial 5G signal, and preprocess the initial 5G signal to obtain a real-time 5G signal;

[0008] Extracting features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features, and topological features, and obtaining recognition results based on the time domain features, the frequency domain features, and the topological features;

[0009] Marking the interference signal in the real-time 5G signal according to the discrimination result to obtain the interference component;

[0010] The interference component is adaptively filtered to obtain a target 5G signal.

[0011] In one embodiment, the multi-stage generative adversarial network model includes a first-stage discriminator and a second-stage discriminator;

[0012] The feature extraction of the real-time 5G signal by the multi-level generative adversarial network model obtains time domain features, frequency domain features and topological features, and the discrimination result is obtained according to the time domain features, the frequency domain features and the topological features, including:

[0013] Extracting the time domain features and the frequency domain features of the real-time 5G signal through the first-level discriminator, and obtaining a first discriminant output according to the time domain features and the frequency domain features;

[0014] Extracting the topological features of the real-time 5G signal through the second-level discriminator, and obtaining a second discriminant output according to the topological features;

[0015] The first discrimination output and the second discrimination output are processed step by step to obtain a recognition result.

[0016] In one embodiment, the feature extraction of the real-time 5G signal by the multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtaining the discrimination result according to the time domain features, the frequency domain features and the topological features, further includes:

[0017] Build an initial generative adversarial network model;

[0018] Performing adversarial training on the initial generative adversarial network model based on the generator loss function and the discriminator loss function of the initial generative adversarial network model to obtain a reference generative adversarial network model;

[0019] The model parameters of the reference generative adversarial network model are meta-gradient updated by presetting a meta-learning data set to obtain a multi-level generative adversarial network model.

[0020] In one embodiment, the initial generative adversarial network model is subjected to adversarial training based on the generator loss function and the discriminator loss function of the initial generative adversarial network model to obtain a reference generative adversarial network model, including:

[0021] Acquire a 5G base station signal, and preprocess the 5G base station signal to obtain a training signal;

[0022] Generate a training simulation interference signal according to the training signal by the generator of the initial generative adversarial network model;

[0023] Obtaining a training discriminant output according to the simulated interference signal by a discriminator of the initial generative adversarial network model;

[0024] Constructing a generator loss function according to the training simulation interference signal, and constructing a discriminator loss function according to the training discriminant output;

[0025] Minimize the generator loss function and maximize the discriminator loss function to obtain training model parameters;

[0026] Back propagation is performed based on the training model parameters to obtain a reference generative adversarial network model.

[0027] In one embodiment, the step of performing meta-gradient updating on the model parameters of the reference generative adversarial network model by using a preset meta-learning data set to obtain a multi-level generative adversarial network model includes:

[0028] Performing adversarial training on the reference generative adversarial network model using a preset meta-learning data set to obtain training parameters, and constructing a meta-learning loss function according to the training parameters;

[0029] Performing meta-gradient parameter update on the training parameters based on the meta-learning loss function to obtain target parameters;

[0030] The model parameters of the reference generative adversarial network model are updated according to the target parameters to obtain a multi-level generative adversarial network model.

[0031] In one embodiment, the obtaining of the initial 5G signal and preprocessing the initial 5G signal to obtain the real-time 5G signal includes:

[0032] Acquire an initial 5G signal, and perform wavelet transform on the initial 5G signal to obtain a multi-scale component signal;

[0033] Denoising the multi-scale component signal according to a preset soft threshold to obtain a denoised signal, and normalizing and mapping the denoised signal to obtain a reference 5G signal;

[0034] The reference 5G signal is converted into a frequency domain to obtain a frequency domain 5G signal, and the frequency domain 5G signal is filtered and normalized to obtain a real-time 5G signal.

[0035] In addition, to achieve the above-mentioned purpose, the present application also proposes a signal interference component processing device, the signal interference component processing device comprising:

[0036] The signal acquisition module 10 is used to acquire an initial 5G signal and pre-process the initial 5G signal to obtain a real-time 5G signal;

[0037] The adversarial training module 20 is used to extract features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtain recognition results according to the time domain features, the frequency domain features and the topological features;

[0038] An interference signal filtering module 30 is used to mark the interference signal in the real-time 5G signal according to the discrimination result to obtain an interference component;

[0039] The interference signal filtering module 30 is also used to adaptively filter the interference components to obtain the target 5G signal.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a signal interference component processing device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the signal interference component processing method as described above.

[0041] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the signal interference component processing method described above are implemented.

[0042] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the signal interference component processing method described above are implemented.

[0043] One or more technical solutions proposed in this application have at least the following technical effects: by learning the preprocessed 5G signal data, a simulated interference signal is generated; the multi-level generative adversarial network model discriminator identifies the difference between the normal 5G signal and the interference signal layer by layer by comparing the simulated interference signal and the actual 5G signal data; based on adversarial training, the generators and discriminators at each level compete with each other, thereby improving the accuracy and efficiency of interference detection; and combining the time domain, frequency domain and topological features to enable the discriminator to more comprehensively analyze and identify the interference signal, thereby improving the accuracy of interference detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 A schematic diagram of a flow chart provided for the first embodiment of the method for processing signal interference components of the present application;

[0047] Figure 2 A structural diagram of a multi-level generative adversarial network model provided in Example 1 of the signal interference component processing method of the present application;

[0048] Figure 3 A specific processing flow chart of interference signals based on multi-level generative adversarial network model training is provided in Example 1 of the signal interference component processing method of the present application;

[0049] Figure 4 A flow chart of the second embodiment of the method for processing signal interference components of the present application;

[0050] Figure 5 This is a schematic diagram of the module structure of the signal interference component processing device according to an embodiment of the present application;

[0051] Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the signal interference component processing method in the embodiment of the present application.

[0052] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0054] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0055] The main solution of the embodiment of the present application is: obtain an initial 5G signal, pre-process the initial 5G signal to obtain a real-time 5G signal; extract features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtain an identification result based on the time domain features, the frequency domain features and the topological features; mark the interference signal in the real-time 5G signal according to the discrimination result to obtain an interference component; and adaptively filter the interference component to obtain a target 5G signal.

[0056] In terms of interference elimination, the existing technology mainly adopts methods such as adaptive filtering and noise reduction algorithms. These methods can effectively eliminate interference in certain specific scenarios, but when faced with complex interference environments, it is often necessary to frequently adjust the filtering parameters, making it difficult to achieve automated processing. For example, the adaptive filtering algorithm eliminates interference by adjusting the filter weights, but when multiple types of interference coexist, it is difficult for the filter to adapt to multiple interferences at the same time, resulting in poor filtering effects. The noise reduction algorithm relies on the accuracy of the noise model, but in a complex environment, the noise model is difficult to accurately describe various interference signals, resulting in unsatisfactory noise reduction effects.

[0057] The present application provides a solution to obtain an initial 5G signal, pre-process the initial 5G signal to obtain a real-time 5G signal; extract time domain features, frequency domain features and topological features of the real-time 5G signal features through a multi-level generative adversarial network model to obtain a discrimination result; mark the interference signal in the real-time 5G signal according to the discrimination result to obtain the interference component; adaptively filter the interference component to obtain the target 5G signal; this method uses a multi-level generative adversarial network model discriminator to compare the simulated interference signal and the actual 5G signal data, and identifies the difference between the normal 5G signal and the interference signal layer by layer, based on adversarial training, the generator and the discriminator are made to compete with each other, thereby improving the accuracy and efficiency of interference detection, and combining the time domain, frequency domain and topological features to enable the discriminator to more comprehensively analyze and identify the interference signal, thereby improving the accuracy of interference detection.

[0058] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a signal interference component processing method, etc. The following takes the signal interference component processing method as an example to illustrate this embodiment and the following embodiments.

[0059] Based on this, the present application embodiment provides a method for processing signal interference components, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the signal interference component processing method of the present application.

[0060] In this embodiment, the signal interference component processing method includes steps S10 to S40:

[0061] Step S10: Acquire an initial 5G signal, and preprocess the initial 5G signal to obtain a real-time 5G signal.

[0062] It is understandable that the initial 5G signal may be original signal data collected from a 5G base station, and the acquisition of the initial 5G signal may be actively collected.

[0063] It should be noted that the real-time 5G signal may be a signal obtained by preprocessing the initial 5G signal, and may be a signal obtained after denoising, wavelet transform, decomposition, normalization, etc. are performed on the initial 5G signal.

[0064] In a feasible implementation, step S10 may include steps A11 to A13:

[0065] Step A11, obtaining an initial 5G signal, and performing a wavelet transform on the initial 5G signal to obtain a multi-scale component signal.

[0066] It should be noted that before performing wavelet transform on the initial 5G signal, the initial 5G signal may be denoised first and then subjected to wavelet transform.

[0067] It should be understood that wavelet transform, also known as wavelet transform, is a signal processing and data analysis technique that can provide information in both the time domain and the frequency domain. Unlike Fourier transform, wavelet transform uses a scalable short-time window function (called wavelet) that can focus on different parts of the signal to analyze different frequency components.

[0068] It should be further explained that the multi-scale component signal can be obtained by performing multi-scale analysis on the 5G signal according to the discrete wavelet transform; specifically, a wavelet basis function suitable for communication signal processing can be selected, and the number of layers to be decomposed can be determined according to the required resolution and signal characteristics. Each layer of decomposition will produce a series of low-frequency approximation (A) and high-frequency detail (D) coefficients. DWT is applied to the original 5G signal, and the signal is decomposed into components at multiple scales through a filter bank. After each layer of decomposition, the signal is divided into two parts: an approximate signal (A) containing lower frequency components, and a detail signal (D) containing higher frequency components. Analyzing the components at different scales obtained from the DWT can help identify specific patterns, anomalies or noise in the signal. Detail coefficients at different levels may correspond to different physical phenomena, such as multipath effects, channel fading or other interference factors.

[0069] It can be understood that the multi-scale component signal may be a component signal at multiple scales after wavelet decomposition.

[0070] Step A12, denoising the multi-scale component signal according to a preset soft threshold to obtain a denoised signal, and normalizing and mapping the denoised signal to obtain a reference 5G signal. (How much is it, provide a specific example)

[0071] It can be understood that denoising the multi-scale component signal according to a preset soft threshold to obtain a denoised signal may be removing noise components from the multi-scale component signal by a soft threshold method to obtain the denoised signal.

[0072] It should be understood that the soft thresholding method is a denoising method widely used in signal processing and data analysis, and is particularly suitable for signal denoising based on wavelet transform. The basic idea is to perform thresholding on coefficients in the wavelet domain to remove or reduce noise components while trying to maintain the important features of the original signal. The soft threshold in this embodiment can be c·σlog(N), where c is less than (2) 1 / 2 The value of , N represents the number of data points of the denoised signal, and σ represents the standard deviation of the noise.

[0073] It should be noted that the selection of threshold is the key to the soft threshold method. It determines which wavelet coefficients are considered as noise and are modified. Common threshold selection methods include universal threshold, heuristic threshold, minimum description length (MDL) criterion, etc. The threshold is usually related to the standard deviation of the noise. For each wavelet coefficient w, if its absolute value is greater than the selected threshold T, the coefficient will be adjusted to sign(w)·(|w|-T), that is, it will decrease towards zero; if the absolute value is less than or equal to the threshold, the coefficient will be set to zero.

[0074] It should be noted that the normalized mapping of the denoised signal to obtain the reference 5G signal can be performed by normalizing the denoised 5G signal data using a maximum and minimum normalization method to map the signal amplitude range to [0, 1].

[0075] In the specific implementation, S11, collect the original signal data from the 5G base station to form a 5G base station signal data set; S12, denoise the 5G base station signal data set, use the wavelet transform method to decompose the original 5G signal data x(t) into multi-scale components, and remove the noise component through the soft threshold method to obtain denoised 5G signal data; S13, normalize the denoised 5G signal data, use the maximum and minimum normalization method to map the signal amplitude range to [0,1]; S14, store the normalized 5G signal data in the preprocessed 5G base station signal data set.

[0076] Step A13: Perform frequency domain conversion on the reference 5G signal to obtain a frequency domain 5G signal, and filter and normalize the frequency domain 5G signal to obtain a real-time 5G signal.

[0077] It is understandable that performing frequency domain conversion on the reference 5G signal to obtain a frequency domain 5G signal may be performing frequency domain conversion on the reference 5G signal through Fourier transform to obtain a frequency domain 5G signal.

[0078] It should be understood that the pre-processed time domain signal is converted into a frequency domain signal, and then interference detection and elimination are performed to further optimize the recognition and processing capabilities of the interference signal.

[0079] It should be noted that performing frequency domain conversion on the reference 5G signal to obtain a frequency domain 5G signal can be to obtain preprocessed time domain signal data and apply fast Fourier transform to convert the time domain signal into a frequency domain signal X(f): performing frequency domain preprocessing on the signal X(f), including denoising and normalization, to obtain a preprocessed frequency domain signal.

[0080] In this implementation, by preprocessing the acquired 5G signal through wavelet transform, denoising, filtering, normalization, and frequency domain conversion, the data features of the input multi-level generative adversarial network model are made clearer and more specific, thereby more accurately analyzing the interference signal in the 5G signal.

[0081] The above are only feasible implementations of step S10 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S10.

[0082] Step S20: extract features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtain recognition results based on the time domain features, the frequency domain features and the topological features.

[0083] It should be noted that the multi-level generative adversarial network model includes a first-level discriminator and a second-level discriminator. The structure diagram of the multi-level generative adversarial network model can be referred to Figure 2 , the figure includes the discriminator, the generator, and the adversarial training diagram of the generator and the discriminator.

[0084] It should be noted that the time domain features, frequency domain features and topological features of the real-time 5G signal are extracted through discriminators at each level, and the comprehensive features are obtained by integrating each feature.

[0085] In a feasible implementation, step S20 may include steps A21 to A23:

[0086] Step A21, extracting the time domain features and frequency domain features of the real-time 5G signal through the first-level discriminator, and obtaining a first discrimination output based on the time domain features and the frequency domain features.

[0087] It is understandable that the first discriminant output may be an evaluation result of the first-level discriminator on the input signal in a multi-level generative adversarial network (Multi-level GAN) model.

[0088] It should be understood that the first-level discriminator focuses on extracting time domain features (such as mean, variance, peak, etc.) and frequency domain features (such as power spectral density, frequency components, etc.) from real-time signal data to construct a comprehensive description of the input signal.

[0089] It should be noted that the first discriminant output is a feature vector, which contains detailed information about the input signal, which can be used to distinguish true from false, and can also be used for subsequent feature extraction, anomaly marking, etc.

[0090] Step A22: extract the topological features of the real-time 5G signal through the second-level discriminator, and obtain a second discrimination output based on the topological features.

[0091] It should be noted that the second-level discriminator uses the graph neural network structure and combines the topological features of the real-time signal to calculate the second discriminant output.

[0092] It should be noted that the second-level discriminator is usually evaluated based on deeper or higher-level features to further improve the accuracy of identifying signal authenticity and abnormal situations. The discriminant output of the second-level discriminator is its judgment on the authenticity of the input signal based on the extracted time domain features and frequency domain features.

[0093] It is understandable that by capturing deeper features of the signal, the second-level discriminator helps to improve the recognition accuracy of the entire system, especially when facing complex or well-disguised interference signals.

[0094] Step A23, processing the first discrimination output and the second discrimination output step by step to obtain a recognition result.

[0095] It is understandable that the process of processing the discriminant outputs of the first-level and second-level discriminators step by step to obtain the discrimination results at each level is crucial. This process not only helps to improve the recognition accuracy, but also enhances the robustness and generalization ability of the system.

[0096] It should be noted that the discrimination results at each level are the probability of a discriminator determining whether various signals in the real-time 5G signal are interference signals.

[0097] It should be noted that the first discrimination output and the second discrimination output are processed step by step to obtain discrimination results at each level, and the recognition result is obtained according to the weight coefficients of the discriminators at each level and the discrimination results at each level.

[0098] It should be noted that a weight is assigned to each level of discriminant output to reflect its importance in the final decision. The weight can be dynamically adjusted based on the performance of the discriminators at each layer (such as accuracy, recall, etc.).

[0099] It should be emphasized that the discrimination results of each level of discriminators can refer to the following formulas for detailed monitoring of interference components in real-time 5G signal data:

[0100]

[0101] Among them, ω iis the weight coefficient of each level of discriminator, which is set according to its contribution to the identification of interference components. total Represents the recognition result, D i (x i ,x f ,x g ) represents the discrimination results of each level of discriminator, and x of the first discriminator g is 0.

[0102] It should be emphasized that this embodiment optimizes the discriminators at all levels and the final classifier at the same time, thereby achieving more efficient feature utilization and higher classification accuracy.

[0103] In the specific implementation, the first-level and second-level discriminators respectively gave their own evaluation results, and the comprehensive score obtained by the weighted average method was 0.3 (less than the preset threshold 0.5), so it was determined to be an interference signal.

[0104] In this implementation, a multi-level generative adversarial network model is constructed, including multiple generators and discriminators. The generator generates a simulated interference signal by learning the preprocessed 5G signal data, and the discriminator identifies the difference between the normal 5G signal and the interference signal layer by layer by comparing the simulated interference signal and the actual 5G signal data. Through adversarial training, the generators and discriminators at all levels compete with each other, which improves the accuracy and efficiency of interference detection. The graph neural network structure is introduced into the discriminator, and the time domain, frequency domain and topological features are combined to enable the discriminator to analyze and identify interference signals more comprehensively, thereby improving the accuracy of interference detection.

[0105] The above are only feasible implementations of step S20 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S20.

[0106] Step S30: Mark the interference signal in the real-time 5G signal according to the judgment result to obtain the interference component.

[0107] It is understandable that marking the interference signal in the real-time 5G signal can be to add an "interference" tag and record a specific timestamp.

[0108] It should be noted that after real-time 5G is identified, it can be determined whether it is an interference signal through the judgment result. If it is an interference signal, the band signal of the real-time 5G signal is filtered as an interference component.

[0109] Step S40: Adaptively filter the interference component to obtain the target 5G signal.

[0110] It is understandable that the target 5G signal is the signal after the interference signal is filtered out.

[0111] It can be understood that by using frequency domain separation technology, the interference components are identified to be concentrated in a specific frequency band, and the frequency band is removed by a filter to restore the quality of the original signal. Furthermore, the extracted interference components are saved in a database for further analysis in order to understand the characteristics and possible sources of the interference.

[0112] It should be emphasized that during the training process, the effects of interference detection and elimination are monitored in real time, dynamic optimization is performed based on feedback data, and the multi-level generative adversarial network model and interference elimination algorithm are continuously adjusted to optimize the system's adaptability and processing efficiency in various interference environments.

[0113] It should be further explained that during the training process, the appropriate adaptive filtering algorithm is selected according to the interference component identified, and the interference signal component D is selected according to the interference component. int The characteristics of the filter are used to set the initial filter parameters; the input signal of the adaptive filter is defined as x real,1 (t), the filter weight is ω(t), and the output signal is y(t); combined with the characteristics of the generative adversarial network, an adaptive filtering method based on variational inference is adopted to optimize the filter weight. The following formula can be referred to:

[0114]

[0115] Where η is the learning rate, ▽ w is the gradient of the filter weight; p(z) is the noise distribution, and the error e(t) between the filter output signal y(t) and the expected signal d(t) is calculated; the error is fed back to the filter weight update process, and the filter weight is further optimized through the variational inference method:

[0116]

[0117] Among them, μ is the learning rate and λ is the regularization parameter, which controls the influence of the generative adversarial network on the weight update; until the error e(t) converges or reaches the preset number of iterations, the interference elimination is completed.

[0118] It should be noted that the detected interference signal component D int (f) Perform adaptive filtering to remove interference signal components and restore the pure frequency domain signal X clean (f); The restored pure frequency domain signal X clean (f) Convert back to the time domain signal X through inverse fast Fourier transform clean (t); for the restored pure time domain signal X clean (t) Perform post-processing; store the processed pure 5G signal data in the output data set D clean middle.

[0119] It should be emphasized that the specific processing flow of interference signals based on the training of the multi-level generative adversarial network model can be referred to Figure 3 , Figure 3 First, data collection and preprocessing are carried out, and then the multi-level generative adversarial network model is constructed and trained. The meta-learning algorithm is introduced to optimize the trained model. Interference signal detection and interference signal elimination are performed on the optimized model. System feedback and optimization are performed based on the detection results. According to the optimization results, it is determined whether to perform interference signal detection again. After the detection is completed, frequency domain transformation is performed to obtain the 5G signal after the interference signal is cleared.

[0120] This embodiment provides a method for processing signal interference components. By learning the preprocessed 5G signal data, a simulated interference signal is generated. The multi-level generative adversarial network model discriminator compares the simulated interference signal with the actual 5G signal data, and identifies the difference between the normal 5G signal and the interference signal layer by layer. Based on adversarial training, the generators and discriminators at each level compete with each other, which improves the accuracy and efficiency of interference detection. The combination of time domain, frequency domain and topological features enables the discriminator to more comprehensively analyze and identify interference signals, thereby improving the accuracy of interference detection.

[0121] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 4 Before step S20, the signal interference component processing method further includes steps S201 to S203:

[0122] Step S201: construct an initial generative adversarial network model.

[0123] It should be noted that a multi-level generative adversarial network model is constructed. The generator generates a simulated interference signal by learning the preprocessed 5G signal data. The discriminator identifies the difference between the normal 5G signal and the interference signal layer by layer through the simulated interference signal and the actual 5G signal data. Through adversarial training, the generators and discriminators at all levels compete with each other for training.

[0124] In the specific implementation, the multi-level generative adversarial network model is constructed to include multiple generators G i and the discriminator D i , where i=1, 2, 3,..., and n represents different levels.

[0125] It is understandable that the structure of the initial generative adversarial network model is the same as that of the multi-level generative adversarial network model, but the model parameters of the initial generative adversarial network model can be randomly generated model parameters, and the parameters are optimized through subsequent training to obtain the final multi-level generative adversarial network model.

[0126] Step S202: performing adversarial training on the initial generative adversarial network model based on the generator loss function and the discriminator loss function of the initial generative adversarial network model to obtain a reference generative adversarial network model.

[0127] It is understandable that the generator loss function and the discriminator loss function can be that semi-supervised learning and self-supervised learning methods are introduced in the adversarial training process, semi-supervised training is performed using partially labeled data, and self-supervised training is performed using unlabeled data. In adversarial training, the generator loss function is formed by the training results of the discriminator recognizing the interference signal generated by the generator; the discriminator loss function is generated according to the training results of the discriminator successfully recognizing the signal of the interferer generated by the generator.

[0128] It should be noted that the generator loss function is minimized and the discriminator loss function is maximized to obtain the training model parameters, and the model parameters are adjusted through back propagation to obtain the reference generative adversarial network model.

[0129] In a feasible implementation, step S202 may include steps A2021 to A2026:

[0130] Step A2021, obtain the 5G base station signal, preprocess the 5G base station signal, and obtain a training signal.

[0131] It should be noted that the 5G base station signal may be a 5G signal including an interference signal emitted by a collected base station.

[0132] It is understandable that the preprocessing method for preprocessing the 5G base station signal is the same as the preprocessing method for the initial 5G signal.

[0133] Step A2022: Generate a training simulation interference signal based on the training signal through the generator of the initial generative adversarial network model.

[0134] It should be noted that the generator learns the preprocessed 5G signal dataset D pre , generate the simulated interference signal G i (z|c), where z is a random noise vector and c is a conditioning variable containing metadata about the signal.

[0135] Step A2023, obtaining a training discriminant output according to the simulated interference signal through the discriminator of the initial generative adversarial network model.

[0136] It should be noted that the discriminator D i Combined with the graph neural network structure, the time domain, frequency domain and topological characteristics of the signal are used to receive the generator G i The generated simulated interference signal G i(z|c) and actual 5G signal data x norm (t) Get the training discriminant output, which can be specifically referred to the following formula:

[0137] D i (x|c)=σ(W i，t ·x t +W i，f ·x f +W i，g ·x g +b i )

[0138] Among them, σ represents the activation function, W i,t , W i,f and W i,g They are respectively the discriminator D i The time domain, frequency domain and graph structure weight matrix, x t 、x f and x g are the time domain, frequency domain and graph structure characteristics of the input signal, respectively, i is the bias vector;

[0139] In a specific implementation, the generator G i Generate a simulated interference signal G i (z|c), where z is a random noise vector and c is a conditional variable; the discriminator D i Receive the generated simulated interference signal G i (z|c) and the preprocessed frequency domain signal D i (G i (z|c),X pre (f)) obtain the training discriminant output; through adversarial training, the parameters of the generator and discriminator are optimized to improve the accuracy of interference detection and elimination.

[0140] Step A2024, constructing a generator loss function based on the training simulated interference signal, and constructing a discriminator loss function based on the training discriminant output.

[0141] It is understandable that this embodiment trains the model using semi-supervised training and self-supervised training, wherein the semi-supervised training data contains labeled signals, and the loss function of the generator and the loss result of the discriminator can be constructed based on the labeled results and the discriminant results.

[0142] It should be understood that self-supervised training can enable the model's generator to learn more signal features, thereby generating more realistic simulated interference signals.

[0143] It should be noted that in order to identify the difference between normal 5G signals and interference signals, adversarial training is performed on the generator and the discriminator. By minimizing the loss function of the generator and maximizing the loss function of the discriminator, the generators and discriminators at all levels compete with each other.

[0144] Among them, the loss function of the generator is:

[0145]

[0146] The loss function of the discriminator is:

[0147]

[0148] Among them, p z (z) is the noise distribution, p c (c) is the conditional variable distribution, p data (x) is the actual signal data distribution, x t 、x f and x g are time domain, frequency domain and graph structure features respectively, G i (z t |c) G i (z f |c) G i (z g |c) The time domain, frequency domain and graph structure signals output by the generator;

[0149] Finally, through the adversarial training of the multi-level generative adversarial network model, the generators G at each level are i and the discriminator D i Play against each other.

[0150] Step A2025, minimize the generator loss function and maximize the discriminator loss function to obtain training model parameters.

[0151] It should be noted that the generator G of the multi-level generative adversarial network model is initialized i and the discriminator D i Parameters, including the weight matrix W i,t , W i,f and W i,g and the bias vector b i ; Randomly extract noise vector z and conditional variable c from noise distribution and conditional variable distribution; Generator G i Generate a simulated interference signal G i (z|c); Discriminator D i Combined with the graph neural network structure, calculate the discriminant output D i (G i (z|c)|c); the discriminator D i The received simulated interference signal Gi (z|c) and actual 5G signal data x norm (t) Combine the signal’s time domain, frequency domain, and topological features to calculate its discriminant output D i (x t , x f , x g |c), where x t 、x f and x g They are the time domain, frequency domain and graph structure characteristics of the actual signal respectively;

[0152] Furthermore, semi-supervised learning and self-supervised learning methods are introduced in the adversarial training process. Partially labeled data are used for semi-supervised training, and unlabeled data are used for self-supervised training to optimize the robustness and generalization ability of the generator and discriminator. The specific optimization can refer to the following formula:

[0153]

[0154] Among them, λ is a trade-off term that controls the balance between adversarial loss and gradient penalty term. It is a gradient penalty term for the discriminator, ensuring that the discriminator is smooth and robust in the input space. The discriminator optimization can refer to the following formula:

[0155]

[0156] Among them, α is a trade-off term that controls the balance between the generation loss and the generator gradient penalty term; It is a gradient penalty term for the generator, ensuring that the generator is smooth and stable in the input space.

[0157] Step A2026, back propagation is performed based on the training model parameters to obtain a reference generative adversarial network model.

[0158] It should be noted that the generator G is updated by the back propagation algorithm and the gradient descent method i and the discriminator D i Parameters to minimize the generator's loss function L Gi And maximize the discriminator loss function L Di .

[0159] It should be emphasized that the training of the multi-level generative adversarial network model is completed until the adversarial training reaches the preset number of iterations or the loss function converges.

[0160] In this embodiment, semi-supervised learning and self-supervised learning methods are introduced in the adversarial training process, and semi-supervised training is performed using partially labeled data and self-supervised training is performed using unlabeled data. This method optimizes the robustness and generalization ability of the generator and discriminator, improves the stability and reliability of the model in the face of complex interference environments, and refines the monitoring of interference components in real-time 5G signal data through the comprehensive output of multi-level discriminators and the discrimination results of discriminators at all levels. By calculating the discrimination results of the comprehensive output, accurate identification and processing of interference signals are achieved.

[0161] The above are only feasible implementations of step S202 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S202.

[0162] Step S203: performing meta-gradient update on the model parameters of the reference generative adversarial network model using a preset meta-learning data set to obtain a multi-level generative adversarial network model.

[0163] Understandably, Meta-Learning, also known as "learning how to learn", is a machine learning method that aims to enable models to quickly adapt to new tasks or environments without extensive retraining. It extracts common knowledge from multiple related tasks and uses this knowledge to accelerate and improve the learning process for new tasks.

[0164] It should be noted that, when constructing a meta-learning algorithm framework, a meta-task set T is defined, each meta-task includes a training set and a validation set, each training set is used as a preset meta-learning training set, and each validation set is used as a preset meta-learning validation set.

[0165] The preset meta-learning training set can contain multiple different tasks or environments, and each task corresponds to a set of input-output pairs (such as different types of 5G signals and their labels); the preset meta-learning validation set is used to evaluate the generalization ability of the model on unseen tasks. It also consists of multiple tasks, but is different from the tasks in the training set.

[0166] It should be noted that the meta-gradient update can be to adjust the parameters of the generator and the discriminator according to the meta-gradient update formula.

[0167] In a feasible implementation, step S203 may include steps A2031 to A2033:

[0168] Step A2031, performing adversarial training on the reference generative adversarial network model using a preset meta-learning data set to obtain training parameters, and constructing a meta-learning loss function based on the training parameters.

[0169] It should be noted that the generator is updated with one or several gradients using the preset meta-learning training set data to obtain a specific generator. Similarly, the discriminator is fine-tuned to obtain a specific discriminator, and the generator and discriminator losses on the task are calculated based on the specific generator and discriminator.

[0170] In the specific implementation, for each meta-task, the 5G signal data and noise vector z in the training set are used to train the generator G i Perform training, generate simulated interference signals, and use the discriminator D i Conduct adversarial training on the generated simulated interference signal and actual 5G signal data to update the parameters of the generator and discriminator;

[0171] The loss function is generated based on the updated generator and discriminator. The loss function is as follows:

[0172]

[0173] Where x represents the actual 5G signal data in the validation set.

[0174] Step A2032: performing meta-gradient parameter update on the training parameters based on the meta-learning loss function to obtain target parameters.

[0175] It should be noted that the meta-gradient update is performed based on the meta-learning loss function, and the initial parameters θ of the generator and discriminator are updated using the gradient descent method. G and θ D , the specific update formula is as follows:

[0176]

[0177] Among them, β is the meta-learning rate, which is used to control the update step size;

[0178] Step A2033, updating the model parameters of the reference generative adversarial network model according to the target parameters to obtain a multi-level generative adversarial network model.

[0179] It should be noted that all meta-tasks are trained until the model converges to obtain generators and discriminators that adapt to various interference types and environmental changes.

[0180] In this implementation, by introducing a meta-learning algorithm, constructing a meta-task set, and using a training set and a validation set to perform multi-task training on the generator and discriminator, the model can quickly adapt to new interference types and environmental changes. The introduction of the meta-learning algorithm significantly improves the model's adaptability and generalization capabilities, ensuring stable performance under different interference environments.

[0181] The above are only feasible implementations of step S203 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S203.

[0182] This embodiment provides a method for processing signal interference components. On the basis of a multi-level generative adversarial network model, a hierarchical meta-learning operator is further introduced to enhance the model's adaptability to different interference types and environmental changes. By defining a meta-task set T, each meta-task contains a training set and a validation set, and training the training set of each meta-task, the generator and the discriminator continuously optimize their parameters in generating simulated interference signals and adversarial training. The validation set is used to evaluate the generalization ability of the model, and meta-gradient updates are performed based on the validation loss. Through the introduction of hierarchical meta-learning operators, the multi-level generative 5G adversarial network model in this scheme can quickly adapt to new interference types and environmental changes, and further, this innovative meta-learning application in the field of 5G signal processing is proposed for the first time in this proposal, and is intended to solve the problem of difficult generalization of models in the prior art.

[0183] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the signal interference component processing method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0184] This application also provides a signal interference component processing device, please refer to Figure 5 , the signal interference component processing device comprises:

[0185] The signal acquisition module 10 is used to acquire an initial 5G signal and pre-process the initial 5G signal to obtain a real-time 5G signal;

[0186] The adversarial training module 20 is used to extract features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtain recognition results according to the time domain features, the frequency domain features and the topological features;

[0187] An interference signal filtering module 30 is used to mark the interference signal in the real-time 5G signal according to the discrimination result to obtain an interference component;

[0188] The interference signal filtering module 30 is also used to adaptively filter the interference components to obtain the target 5G signal.

[0189] The signal interference component processing device provided by the present application adopts the signal interference component processing method in the above embodiment, which can solve the technical problem that it is difficult to accurately distinguish normal signals from interference signals when facing diversified interference signals, and it is difficult to accurately identify various interference signals, resulting in unsatisfactory noise reduction effects and affecting the stability and reliability of 5G networks. Compared with the prior art, the beneficial effects of the signal interference component processing device provided by the present application are the same as the beneficial effects of the signal interference component processing method provided by the above embodiment, and the other technical features in the signal interference component processing device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0190] In one embodiment, the adversarial training module 20 is further used to extract the time domain features and the frequency domain features of the real-time 5G signal through the first-level discriminator, and obtain a first discriminant output according to the time domain features and the frequency domain features;

[0191] Extracting the topological features of the real-time 5G signal through the second-level discriminator, and obtaining a second discriminant output according to the topological features;

[0192] The first discrimination output and the second discrimination output are processed step by step to obtain a recognition result.

[0193] In one embodiment, the adversarial training module 20 is also used to construct an initial generative adversarial network model;

[0194] Performing adversarial training on the initial generative adversarial network model based on the generator loss function and the discriminator loss function of the initial generative adversarial network model to obtain a reference generative adversarial network model;

[0195] The model parameters of the reference generative adversarial network model are meta-gradient updated by presetting a meta-learning data set to obtain a multi-level generative adversarial network model.

[0196] In one embodiment, the adversarial training module 20 is further used to obtain a 5G base station signal, pre-process the 5G base station signal, and obtain a training signal;

[0197] Generate a training simulation interference signal according to the training signal by the generator of the initial generative adversarial network model;

[0198] Obtaining a training discriminant output according to the simulated interference signal by a discriminator of the initial generative adversarial network model;

[0199] Constructing a generator loss function according to the training simulation interference signal, and constructing a discriminator loss function according to the training discriminant output;

[0200] Minimize the generator loss function and maximize the discriminator loss function to obtain training model parameters;

[0201] Back propagation is performed based on the training model parameters to obtain a reference generative adversarial network model.

[0202] In one embodiment, the adversarial training module 20 is further used to perform adversarial training on the reference generative adversarial network model through a preset meta-learning data set to obtain training parameters, and construct a meta-learning loss function according to the training parameters;

[0203] Performing meta-gradient parameter update on the training parameters based on the meta-learning loss function to obtain target parameters;

[0204] The model parameters of the reference generative adversarial network model are updated according to the target parameters to obtain a multi-level generative adversarial network model.

[0205] In one embodiment, the signal acquisition module 10 is further used to acquire an initial 5G signal, and perform a wavelet transform on the initial 5G signal to obtain a multi-scale component signal;

[0206] Denoising the multi-scale component signal according to a preset soft threshold to obtain a denoised signal, and normalizing and mapping the denoised signal to obtain a reference 5G signal;

[0207] The reference 5G signal is converted into a frequency domain to obtain a frequency domain 5G signal, and the frequency domain 5G signal is filtered and normalized to obtain a real-time 5G signal.

[0208] The present application provides a signal interference component processing device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the signal interference component processing method in the above-mentioned embodiment one.

[0209] Reference below Figure 6 , which shows a schematic diagram of the structure of a signal interference component processing device suitable for implementing an embodiment of the present application. The signal interference component processing device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6The signal interference component processing device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0210] like Figure 6 As shown, the signal interference component processing device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the signal interference component processing device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the signal interference component processing device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a signal interference component processing device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0211] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0212] The signal interference component processing device provided by the present application adopts the signal interference component processing method in the above embodiment, which can solve the technical problem that it is difficult to accurately distinguish normal signals from interference signals when facing diversified interference signals, and it is difficult to accurately identify various interference signals, resulting in unsatisfactory noise reduction effects and affecting the stability and reliability of 5G networks. Compared with the prior art, the beneficial effects of the signal interference component processing device provided by the present application are the same as the beneficial effects of the signal interference component processing method provided by the above embodiment, and the other technical features in the signal interference component processing device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0213] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0214] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0215] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the signal interference component processing method in the above-mentioned embodiment.

[0216] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0217] The above-mentioned computer-readable storage medium may be included in the signal interference component processing device; or it may exist independently without being assembled into the signal interference component processing device.

[0218] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the signal interference component processing device, the signal interference component processing device: obtains an initial 5G signal, preprocesses the initial 5G signal to obtain a real-time 5G signal; extracts features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtains identification results based on the time domain features, the frequency domain features and the topological features; marks the interference signal in the real-time 5G signal according to the discrimination result to obtain an interference component; and adaptively filters the interference component to obtain a target 5G signal.

[0219] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0220] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0221] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0222] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned signal interference component processing method, which can solve the technical problems that it is difficult to accurately distinguish normal signals from interference signals when facing diversified interference signals, and it is difficult to accurately identify various interference signals, resulting in unsatisfactory noise reduction effects and affecting the stability and reliability of 5G networks. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the signal interference component processing method provided by the above-mentioned embodiment, and will not be repeated here.

[0223] The present application also provides a computer program product, including a computer program, which implements the steps of the signal interference component processing method as described above when the computer program is executed by a processor.

[0224] The computer program product provided by the present application can solve the technical problem that it is difficult to accurately distinguish normal signals from interference signals when facing diversified interference signals, and it is difficult to accurately identify various interference signals, resulting in unsatisfactory noise reduction effects and affecting the stability and reliability of 5G networks. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the signal interference component processing method provided by the above embodiment, and will not be repeated here.

[0225] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for processing signal interference components, characterized in that: The interference signal identification method comprises: Acquire an initial 5G signal, and preprocess the initial 5G signal to obtain a real-time 5G signal; Extracting features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features, and topological features, and obtaining recognition results based on the time domain features, the frequency domain features, and the topological features; Marking the interference signal in the real-time 5G signal according to the discrimination result to obtain the interference component; The interference component is adaptively filtered to obtain a target 5G signal.

2. The method according to claim 1, characterized in that The multi-level generative adversarial network model includes a first-level discriminator and a second-level discriminator; The feature extraction of the real-time 5G signal by the multi-level generative adversarial network model obtains time domain features, frequency domain features and topological features, and the discrimination result is obtained according to the time domain features, the frequency domain features and the topological features, including: Extracting the time domain features and the frequency domain features of the real-time 5G signal through the first-level discriminator, and obtaining a first discriminant output according to the time domain features and the frequency domain features; Extracting the topological features of the real-time 5G signal through the second-level discriminator, and obtaining a second discriminant output according to the topological features; The first discrimination output and the second discrimination output are processed step by step to obtain a recognition result.

3. The method according to claim 1, characterized in that The feature extraction of the real-time 5G signal by the multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtaining the discrimination result according to the time domain features, the frequency domain features and the topological features, further includes: Build an initial generative adversarial network model; Performing adversarial training on the initial generative adversarial network model based on the generator loss function and the discriminator loss function of the initial generative adversarial network model to obtain a reference generative adversarial network model; The model parameters of the reference generative adversarial network model are meta-gradient updated by presetting a meta-learning data set to obtain a multi-level generative adversarial network model.

4. The method according to claim 3, characterized in that The generator loss function and the discriminator loss function of the initial generative adversarial network model are used to perform adversarial training on the initial generative adversarial network model to obtain a reference generative adversarial network model, including: Acquire a 5G base station signal, and preprocess the 5G base station signal to obtain a training signal; Generate a training simulation interference signal according to the training signal by the generator of the initial generative adversarial network model; Obtaining a training discriminant output according to the simulated interference signal by a discriminator of the initial generative adversarial network model; Constructing a generator loss function according to the training simulation interference signal, and constructing a discriminator loss function according to the training discriminant output; Minimize the generator loss function and maximize the discriminator loss function to obtain training model parameters; Back propagation is performed based on the training model parameters to obtain a reference generative adversarial network model.

5. The method according to claim 3, characterized in that The method of performing meta-gradient updating on the model parameters of the reference generative adversarial network model by presetting the meta-learning data set to obtain a multi-level generative adversarial network model includes: Performing adversarial training on the reference generative adversarial network model using a preset meta-learning data set to obtain training parameters, and constructing a meta-learning loss function according to the training parameters; Performing meta-gradient parameter update on the training parameters based on the meta-learning loss function to obtain target parameters; The model parameters of the reference generative adversarial network model are updated according to the target parameters to obtain a multi-level generative adversarial network model.

6. The method according to claim 1, characterized in that The obtaining of the initial 5G signal and preprocessing the initial 5G signal to obtain a real-time 5G signal includes: Acquire an initial 5G signal, and perform wavelet transform on the initial 5G signal to obtain a multi-scale component signal; Denoising the multi-scale component signal according to a preset soft threshold to obtain a denoised signal, and normalizing and mapping the denoised signal to obtain a reference 5G signal; The reference 5G signal is converted into a frequency domain to obtain a frequency domain 5G signal, and the frequency domain 5G signal is filtered and normalized to obtain a real-time 5G signal.

7. A signal interference component processing device, characterized in that: The signal interference component processing device comprises: A signal acquisition module, used to acquire an initial 5G signal and pre-process the initial 5G signal to obtain a real-time 5G signal; An adversarial training module, used to extract features of the real-time 5G signal through a multi-level generative adversarial network model to obtain time domain features, frequency domain features and topological features, and obtain recognition results based on the time domain features, the frequency domain features and the topological features; An interference signal filtering module, used to mark the interference signal in the real-time 5G signal according to the discrimination result to obtain the interference component; The interference signal filtering module is also used to adaptively filter the interference components to obtain the target 5G signal.

8. A signal interference component processing device, characterized in that: The signal interference component processing device includes: a memory, a processor, and a signal interference component processing program stored in the memory and executable on the processor, wherein the signal interference component processing program is configured to implement the signal interference component processing method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a signal interference component processing program, and when the signal interference component processing program is executed by the processor, the signal interference component processing method according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product comprises a signal interference component processing program, and when the signal interference component processing program is executed by a processor, the steps of the signal interference component processing method according to any one of claims 1 to 6 are implemented.

Images