Wireless interference source detection method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of wireless communication, in particular to a wireless interference source detection method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a training data set which comprises channel state data and interference source type tags under different wireless interference sources; an interference source type identification model is established based on a Transform model, and the interference source type identification model comprises a block division layer, a feature extraction layer, a block merging layer and an output layer; and training an interference source type identification model by using the training data set, stopping training until the identification performance of the interference source type identification model reaches a preset standard, and identifying the type of the wireless interference source in the target wireless interference environment by using the interference source type identification model. Therefore, the problem that the interference source cannot be identified in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a method, device, electronic device and storage medium for detecting a wireless interference source. Background Art

[0002] With the popularity of IoT devices in home environments, the stability and security of home wireless communication systems have become particularly important. Wireless communication systems in home environments are responsible for data transmission between various smart home devices, such as security monitoring, smart lighting, and smart home appliance control. However, there are usually various wireless devices in the home environment, and the wireless signals between these devices interfere with each other, which may cause a decrease in communication quality and affect the normal use of home devices. Especially when multiple devices are used at the same time, the interference of wireless signals will be more serious.

[0003] In recent years, deep learning has been widely used in the fields of signal processing and image processing. Due to its unique architectural design and self-attention mechanism, Transformer can capture global dependencies when processing sequence data and has strong modeling capabilities in the spatiotemporal dimension. In the field of signal recognition, Transformer has shown significant advantages, especially when processing complex signals with time-dynamic changes, it can more accurately capture potential patterns and features. Applying Transformer to wireless interference signal detection in home environments can not only improve the accuracy of recognition, but also improve processing efficiency through its parallel computing characteristics, thereby achieving more real-time interference signal recognition and response, providing more reliable protection for smart home environments.

[0004] At present, there are many types of wireless interference sources in the home environment, including interference from smart home devices, Bluetooth devices, microwave ovens and other household appliances. The interference signals generated by these interference sources seriously affect the stability and reliability of wireless communications. At present, the interference detection method is mainly based on data such as channel utilization, SNR and RSSI, and judges the interference intensity in the environment according to the threshold, or uses WiFi channel scanning to obtain the WiFi signal distribution in the 2.4GHz and 5GHz channels to judge the interference size. These methods can only preliminarily judge the interference intensity in the environment, but cannot identify the source of interference in the environment. Summary of the invention

[0005] The present application provides a wireless interference source detection method, device, electronic device and storage medium to solve the problem that the interference source cannot be identified in the related art.

[0006] The first aspect of the present application provides a wireless interference source detection method, comprising the following steps: obtaining a training data set, wherein the training data set includes channel state data and interference source type labels under different wireless interference sources; establishing an interference source type recognition model based on a Transformer model, wherein the interference source type recognition model includes a block partitioning layer, a feature extraction layer, a block merging layer and an output layer, the block partitioning layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit, the feature extraction layer is used to extract features from multiple block units and perform feature learning, the block merging layer is used to merge adjacent image blocks, and the output layer is used to output the type of wireless interference source; using the training data set to train the interference source type recognition model until the recognition performance of the interference source type recognition model reaches a preset standard and the training is stopped, and using the interference source type recognition model to identify the type of wireless interference source in the target wireless interference environment.

[0007] Optionally, the feature extraction layer includes multiple Swin Transformer Block modules, each Swin Transformer Block includes a window multi-head self-attention mechanism and a window moving multi-head self-attention mechanism, which are used to divide the multiple block units into multiple non-intersecting windows, wherein:

[0008] Windowed multi-head self-attention mechanism, which is used to calculate attention within each window through normalization layer, windowed multi-head self-attention mechanism, multi-layer perceptron and skip connection;

[0009] The window moving multi-head self-attention mechanism is used to splice different windows and calculate attention through normalization layers, window moving multi-head self-attention mechanism units, multi-layer perceptrons and jump connections.

[0010] Optionally, the output layer includes: a normalization layer, a global pooling layer and a fully connected layer, wherein the normalization layer is used to maintain the stability of the interference source type identification model training; the global pooling layer is used to extract global features from the channel state time-frequency diagram; the fully connected layer is used to integrate all feature information in the block partition layer, the feature extraction layer and the block merging layer, and use the loss function to determine the type of wireless interference source.

[0011] Optionally, the loss function is:

[0012]

[0013] Among them, K represents the number of interference sources, y q Indicates a label, a q It represents the result of classification output, and q represents the type of interference source.

[0014] Optionally, obtaining a training data set includes: obtaining channel state data under different wireless interference environments; filtering, normalizing, preprocessing, and Fourier transforming the channel state data to obtain a corresponding channel state time-frequency diagram, and marking the channel state time-frequency diagram with corresponding interference source type labels to obtain a training data set of channel state data in different interference sources.

[0015] The second aspect of the present application provides a wireless interference source detection method, including: obtaining channel state data in a target wireless interference environment; filtering and Fourier transforming the channel state data to obtain a channel state time-frequency diagram; inputting the channel state time-frequency diagram into an interference source type recognition model, and the interference source type recognition model outputs a wireless interference source detection result in the target wireless interference environment, wherein the interference source type recognition model is established based on a Transformer model.

[0016] The third aspect of the present application provides a wireless interference source detection device, including: a first acquisition module, used to acquire a training data set, wherein the training data set includes channel state data and interference source type labels under different wireless interference sources; an establishment module, used to establish an interference source type recognition model based on a Transformer model, wherein the interference source type recognition model includes a block division layer, a feature extraction layer, a block merging layer and an output layer, the block division layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit, the feature extraction layer is used to extract features from multiple block units and perform feature learning, the block merging layer is used to merge adjacent image blocks, and the output layer is used to output the type of wireless interference source; a training module, used to train the interference source type recognition model using the training data set, until the recognition performance of the interference source type recognition model reaches a preset standard to stop training, and use the interference source type recognition model to identify the type of wireless interference source in the target wireless interference environment.

[0017] Optionally, the feature extraction layer includes multiple Swin Transformer Block modules, each Swin Transformer Block includes a window multi-head self-attention mechanism and a window moving multi-head self-attention mechanism, which are used to divide the multiple block units into multiple non-intersecting windows, wherein:

[0018] Windowed multi-head self-attention mechanism, which is used to calculate attention within each window through normalization layer, windowed multi-head self-attention mechanism, multi-layer perceptron and skip connection;

[0019] The window moving multi-head self-attention mechanism is used to splice different windows and calculate attention through normalization layers, window moving multi-head self-attention mechanism units, multi-layer perceptrons and jump connections.

[0020] Optionally, the loss function is:

[0021]

[0022] Among them, K represents the number of interference sources, y q Indicates a label, a q It represents the result of classification output, and q represents the type of interference source.

[0023] Optionally, the output layer includes: a normalization layer, a global pooling layer and a fully connected layer, wherein the normalization layer is used to maintain the stability of the interference source type identification model training; the global pooling layer is used to extract global features from the channel state time-frequency diagram; the fully connected layer is used to integrate all feature information in the block partition layer, the feature extraction layer and the block merging layer, and use the loss function to determine the type of wireless interference source.

[0024] Optionally, the first acquisition module is further used to: obtain channel state data under different wireless interference environments; filter, normalize, preprocess and Fourier transform the channel state data to obtain a corresponding channel state time-frequency diagram, and mark the channel state time-frequency diagram with corresponding interference source type labels to obtain a training data set of channel state data in different interference sources.

[0025] The fourth aspect of the present application provides a wireless interference source detection device, including: a second acquisition module, used to obtain channel state data in a target wireless interference environment; a processing module, used to filter and Fourier transform the channel state data to obtain a channel state time-frequency diagram; an input module, used to input the channel state time-frequency diagram into an interference source type identification model, the interference source type identification model outputs the wireless interference source detection result in the target wireless interference environment, wherein the interference source type identification model is established based on a Transformer model.

[0026] The fifth aspect of the present application provides a vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform a wireless interference source detection method as described in the above embodiment.

[0027] The sixth aspect of the present application provides a computer-readable storage medium on which a computer program or instruction is stored. The computer program or instruction is executed by a processor to perform a wireless interference source detection method as described in the above embodiment.

[0028] Therefore, this application has at least the following beneficial effects:

[0029] The embodiment of the present application can use the channel state data and interference source type labels under different wireless interference sources to train the interference source type recognition model until the recognition performance of the model reaches the preset standard and stops training, thereby ensuring the robustness and generalization ability of the model, so that the interference source type recognition model is fully trained and verified, so as to have high reliability and accuracy, and the interference source type recognition model includes block division, feature extraction, block merging and output layer, which realizes the effective detection of the type of wireless interference source and provides reliable anti-interference protection for the home wireless communication environment. Therefore, the technical problems such as the inability to identify the interference source in the related technology are solved.

[0030] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0032] Figure 1 A flowchart of a wireless interference source detection method provided according to an embodiment of the present application;

[0033] Figure 2 A schematic diagram of the structure of an interference source type identification model provided according to an embodiment of the present application;

[0034] Figure 3 It is a schematic diagram of the structure of two consecutive Swin Transformer Block modules provided according to an embodiment of the present application;

[0035] Figure 4 A flowchart of a wireless interference source detection method provided according to another embodiment of the present application;

[0036] Figure 5 A specific flow chart of a wireless interference source detection method provided according to a specific embodiment of the present application;

[0037] Figure 6 A schematic diagram of a wireless interference source detection device provided according to an embodiment of the present application;

[0038] Figure 7 A schematic diagram of a wireless interference source detection device provided according to another embodiment of the present application;

[0039] Figure 8 It is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0040] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0041] The wireless interference source detection method, device, electronic device and storage medium of the embodiment of the present application are described below with reference to the accompanying drawings. The current interference detection method mentioned in the above background technology is mainly based on data such as channel utilization, SNR and RSSI, and the size of the interference intensity in the environment is judged according to the threshold, or the WiFi channel scanning is used to obtain the WiFi signal distribution in the 2.4GHz and 5GHz channels to judge the interference size, but it can only preliminarily judge the interference intensity in the environment, and it is impossible to achieve the problem of identifying the source of interference in the environment. The present application provides a wireless interference source detection method, in which the channel state data and interference source type labels under different wireless interference sources can be used to train the interference source type recognition model, until the recognition performance of the model reaches the preset standard to stop training, the robustness and generalization ability of the model are guaranteed, the interference source type recognition model is fully trained and verified, so as to have high reliability and accuracy, and the interference source type recognition model includes block division, feature extraction, block merging and output layer, which realizes the effective detection of the type of wireless interference source. Thus, the problems such as the inability to identify the source of interference in the related art are solved.

[0042] Specifically, Figure 1 A flow chart of a wireless interference source detection method provided in an embodiment of the present application is mainly described from an offline perspective, focusing on the model training process.

[0043] like Figure 1 As shown, the wireless interference source detection includes the following steps:

[0044] In step S101, a training data set is obtained, wherein the training data set includes channel state data under different wireless interference sources and interference source type labels.

[0045] It is understandable that the embodiment of the present application can obtain a training data set to facilitate the subsequent training of a wireless interference source type identification model.

[0046] In an embodiment of the present application, a training data set is obtained, including: obtaining channel state data under different wireless interference environments; filtering, normalizing, preprocessing, and Fourier transforming the channel state data to obtain a corresponding channel state time-frequency diagram, and marking the channel state time-frequency diagram with corresponding interference source type labels to obtain a training data set of channel state data in different interference sources.

[0047] It can be understood that the embodiments of the present application can obtain channel state data under different wireless interference environments, and filter, normalize, preprocess and Fourier transform the channel state data to obtain the corresponding channel state time-frequency diagram to improve the accuracy of subsequent classification, and annotate the channel state time-frequency diagram with corresponding interference source type labels to obtain a training data set of channel state data in different interference sources.

[0048] Among them, the embodiments of the present application can generate various wireless interferences in a home environment, so as to collect channel state data CSI in different wireless interference environments.

[0049] Specifically, the embodiments of the present application can filter and normalize the collected CSI and perform short-time Fourier transform to obtain the corresponding CSI time-frequency diagram, and after marking the corresponding interference source type labels, a training data set of CSI information in different interference sources is obtained.

[0050] In order to accurately reflect the dynamic changes of CSI signals in the time-frequency domain, the short-time Fourier transform (STFT) is used to process the signal in segments. The window function is used, and the length and overlap rate of the window function are determined according to the characteristics of the CSI signal. The spectrum information of the signal is extracted within the window of each window function to generate local spectrum analysis results.

[0051] The short-time Fourier transform is performed on the segmented CSI signal, and its expression is:

[0052]

[0053] Where f and t represent the frequency index and time domain index respectively, N represents the number of sampling points, and w * The conjugate of the window function is represented. Fourier transform is performed on the signal in each time window to obtain the frequency distribution of the signal at each time point. The time domain signal is converted into the time-frequency domain to show the change pattern of the CSI signal in time and frequency.

[0054] The spectrum data after STFT processing is arranged and visualized to generate a time-frequency diagram of the CSI signal. The horizontal axis of the time-frequency diagram represents time, the vertical axis represents frequency, and the color of the image represents the magnitude of the spectrum amplitude.

[0055] In step S102, an interference source type recognition model is established based on the Transformer model, wherein the interference source type recognition model includes a block division layer, a feature extraction layer, a block merging layer and an output layer, the block division layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit, the feature extraction layer is used to extract features from multiple block units and perform feature learning, the block merging layer is used to merge adjacent image blocks to reduce the number of block units in the feature extraction process, and the output layer is used to output the type of wireless interference source.

[0056] It can be understood that the embodiment of the present application can establish an interference source type identification model based on the Transformer model. The interference source type identification model is the ST-Transformer model. The overall structure is as follows: Figure 2 As shown, specifically, the interference source type identification model includes a block division layer, a feature extraction layer, a block merging layer and an output layer, wherein:

[0057] The block partitioning layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit;

[0058] The feature extraction layer (ST-transformer layer) is used to extract features from multiple block units and perform feature learning;

[0059] The block merging layer is used to merge adjacent image blocks to reduce the number of block units in the feature extraction process;

[0060] The output layer is used to output the type of wireless interference source.

[0061] It should be noted that the embodiment of the present application establishes an interference source type identification model based on the Transformer model, and slightly modifies the Tansformer model to obtain the ST-Transformer model, so that it can be better used for interference source type identification. The ST-Transformer model is based on the Transformer model architecture.

[0062] In step S103, the interference source type recognition model is trained using the training data set until the recognition performance of the interference source type recognition model reaches a preset standard and the training is stopped. The interference source type recognition model is used to identify the type of wireless interference source in the target wireless interference environment.

[0063] Among them, the preset standard can be set according to the specific situation, for example, the recognition accuracy can reach a certain value.

[0064] It can be understood that the embodiments of the present application can use the training data set to train the interference source type identification model until the recognition performance of the interference source type identification model reaches a preset standard and stop training, and use the interference source type identification model to identify the type of wireless interference source in the target wireless interference environment to achieve identification of the interference source type.

[0065] It should be noted that the embodiment of the present application can divide the training data set into a data set and a test set, use the training set to train the interference source type identification model, and use the test set to test the interference source type identification model, until the recognition performance of the interference source type identification model reaches a preset standard and the training is stopped.

[0066] In an embodiment of the present application, the feature extraction layer includes multiple Swin Transformer Block modules, each of which includes a window multi-head self-attention mechanism and a window moving multi-head self-attention mechanism, which are used to divide multiple block units into multiple non-intersecting windows, wherein:

[0067] Windowed multi-head self-attention mechanism, which is used to calculate attention within each window through normalization layer, windowed multi-head self-attention mechanism, multi-layer perceptron and skip connection;

[0068] The window moving multi-head self-attention mechanism is used to splice different windows and calculate attention through normalization layers, window moving multi-head self-attention mechanism units, multi-layer perceptrons and jump connections.

[0069] In the embodiment of the present application, the attention calculation formula of the window multi-head self-attention mechanism is:

[0070]

[0071] Among them, Q w , K w 、V w is the query, key, and value matrix within each window, Softmax represents the normalized exponential function, T represents the transpose operation, and dk is the smoothing factor;

[0072] The attention calculation formula for each Swin Transformer Block is:

[0073]

[0074] Among them, LN represents the normalization operation, z l-1 represents the output feature of the l-1th time, represents the output feature of the W-MSA unit for the lth time, z l represents the output features of the lth multi-layer perceptron, represents the output characteristics of the l+1th time through the SW-MSA unit, z l+1 Represents the output features of the multilayer perceptron after the l+1th time.

[0075] It can be understood that the feature extraction layer ST-transformer module of the embodiment of the present application is used for convolutional feature extraction and learning.

[0076] Furthermore, the ST-transformer module mainly includes four groups of Swin Transformer Blocks modules, which split the input H×W×3 image into non-overlapping N×(P 2 ×3) patch, each patch is regarded as a patch token, and a total of N are split; these patch tokens are fed into several Swin Transformer Blocks with improved self-attention. As the network deepens, the number of tokens gradually decreases through the Patch Merging layer; each Stage changes the dimension of the tensor to form a hierarchical representation.

[0077] Furthermore, Swin Transformer Blocks uses the window multi-head self-attention mechanism (W-MSA) and the window moving multi-head self-attention mechanism (SW-MSA), whose attention is expressed as:

[0078]

[0079] Among them, Q w , K w 、V w is the query, key, and value matrix within each window, Softmax represents the normalized exponential function, T represents the transpose operation, and dk is the smoothing factor;

[0080] The input feature map is divided into multiple non-intersecting windows. The W-WSA module only calculates attention in each window through the normalization layer, W-MSA unit, multi-layer perceptron and skip connection. In order to transfer information between different windows, the SW-WSA module will splice different windows and calculate attention through the normalization layer, SW-MSA unit, multi-layer perceptron and skip connection. The structure of two consecutive SwinTransformer Blocks is as follows: Figure 3 As shown, the calculation can be expressed as:

[0081]

[0082] Among them, LN represents the normalization operation, z l-1 represents the output feature of the l-1th time, represents the output feature of the W-MSA unit for the lth time, z l represents the output features of the lth multi-layer perceptron, represents the output characteristics of the l+1th time through the SW-MSA unit, z l+1 Represents the output features of the multilayer perceptron after the l+1th time.

[0083] In an embodiment of the present application, the output layer includes: a normalization layer, a global pooling layer and a fully connected layer, wherein the normalization layer is used to maintain the stability of the interference source type identification model training; the global pooling layer is used to extract global features from the channel state time-frequency diagram; the fully connected layer is used to integrate all feature information in the block partition layer, the feature extraction layer and the block merging layer, and use the loss function to determine the type of wireless interference source.

[0084] It can be understood that the output layer of the embodiment of the present application includes a normalization layer, a global pooling layer and a fully connected layer. Specifically, the normalization layer is used to maintain the stability of the interference source type identification model training, the global pooling layer is responsible for extracting global features from the channel state time-frequency diagram without adding additional parameters, and the fully connected layer is responsible for integrating the previous feature information to make the final classification decision, that is, using the loss function to determine the type of wireless interference source.

[0085] Among them, the loss function can be multi-classification cross entropy

[0086] Among them, K represents the number of interference sources, y q Indicates a label, a q It represents the result of classification output, and q represents the type of interference source.

[0087] In summary, the ST-Transformer model in the embodiment of the present application is based on the Transformer model architecture and realizes hierarchical feature extraction from local to global, allowing the model to gradually acquire a larger receptive field while retaining local information, which helps to capture multi-scale features in interference signals; ST-Transformer limits the self-attention calculation to the local window through the window attention mechanism, and migrates the window between different levels, effectively reducing the computational cost while maintaining the global connectivity of information, which makes ST-Transformer more efficient when processing high-dimensional CSI data.

[0088] According to the wireless interference source detection method proposed in the embodiment of the present application, the channel state data and interference source type labels under different wireless interference sources can be used to train the interference source type identification model until the recognition performance of the model reaches a preset standard and the training is stopped, thereby ensuring the robustness and generalization ability of the model. The interference source type identification model includes block division, feature extraction, block merging and output layer, thereby realizing effective detection of the type of wireless interference source.

[0089] The following embodiments focus on describing the wireless interference source detection method from an online perspective, focusing on implementation.

[0090] Figure 4 The present invention is a flowchart of a wireless interference source detection method provided according to another embodiment of the present application.

[0091] like Figure 4 As shown, the wireless interference source detection method includes the following steps:

[0092] In step S201, channel state data in a target wireless interference environment is obtained.

[0093] The target wireless interference environment may be a wireless interference environment to be measured.

[0094] In step S202, the channel state data is filtered and Fourier transformed to obtain a channel state time-frequency diagram.

[0095] It can be understood that the embodiment of the present application can obtain a channel state time-frequency diagram after filtering and Fourier transforming the channel state data. The processing method has been described in the above embodiment and will not be repeated here.

[0096] In step S203, the channel state time-frequency diagram is input into an interference source type identification model, and the interference source type identification model outputs a wireless interference source detection result in a target wireless interference environment, wherein the interference source type identification model is established based on a Transformer model.

[0097] The method for constructing the interference source type identification model has been described in the above embodiments and will not be repeated here.

[0098] It can be understood that the embodiment of the present application can input the channel state time-frequency diagram into the interference source type identification model, and the interference source type identification model outputs the wireless interference source detection result in the target wireless interference environment to realize the identification of the wireless interference source type.

[0099] According to the wireless interference source detection method proposed in the embodiment of the present application, by pre-processing the channel state data of the target wireless interference environment to be measured and inputting a pre-trained interference source type identification model, the type of wireless interference source can be detected quickly and accurately, which is suitable for real-time monitoring needs in practical applications. The interference source type identification model has been fully trained and verified, and has high reliability and accuracy, thereby improving the accuracy of model identification.

[0100] The following describes the wireless interference source detection method of the present application embodiment from the perspective of model training and implementation through a specific embodiment. The specific process is as follows: Figure 5 As shown, including:

[0101] S1: In a home environment, various wireless interferences are created, and the collectors collect the corresponding channel state information CSI under different wireless interference environments.

[0102] By artificially creating different interference source environments in a home environment, measurement personnel use WiFi routers and computer equipment to collect and save CSI data in different types of interference source environments.

[0103] S2: The collected CSI is filtered, normalized, and short-time Fourier transformed to obtain the corresponding CSI time-frequency diagram. After marking the corresponding interference source type labels, a data set of CSI information in different interference sources is obtained, and the data set is divided into a training set and a test set.

[0104] The CSI data obtained in step S1 is processed using a Butterworth low-pass filter to remove the influence of noise on the data during the acquisition process, and the amplitude and phase of the CSI data are normalized.

[0105] The signal is processed in segments by using short-time Fourier transform (STFT). The window function uses a Hamming window, and the overlap rate of adjacent windows is 50%. The short-time Fourier transform is performed on the segmented CSI signal, and its expression is:

[0106]

[0107] Among them, f and t represent the frequency index and time domain index respectively, N represents the number of sampling points, and w* represents the conjugate of the window function. Fourier transform is performed on the signal in each time window to obtain the frequency distribution of the signal at each time point. The time domain signal is converted into time-frequency domain representation to show the change pattern of the CSI signal in time and frequency.

[0108] The spectrum data after STFT processing is arranged and visualized to generate a time-frequency diagram of the CSI signal. The horizontal axis of the time-frequency diagram represents time, the vertical axis represents frequency, and the color of the image represents the magnitude of the spectrum amplitude. According to the interference source environment in step S1, the corresponding interference source type label is marked to create a data set, and the data set is divided into a training set and a test set in a ratio of 7:3.

[0109] S3: Use the sample data and interference source category labels of the training set to train the network parameters of the ST-Transformer model constructed in step S2 to obtain the trained ST-Transformer model, and verify the recognition performance of the interference source type recognition model through the test set.

[0110] The ST-transformer (feature extraction layer) module is used for convolutional feature extraction and learning. The PatchPartition (block partitioning layer) module splits the input H×W×3 image into non-overlapping N×(4 2 ×3) patch (image block), each patch is regarded as a patch token (block unit), and a total of N patches are split;

[0111] like Figure 2 As shown, the ST-transformer module has a total of 4 stages (levels), and the number of Swin Transformer Blocks contained in each stage is 2, 2, 6, and 2 respectively;

[0112] After passing through the Patch Partition module, the feature dimension of a token becomes [H / 4, W / 4, 48]. In Stagel, the linear embedding layer projects the token tensor to an arbitrary dimension C, making the feature dimension of the token become [H / 4, H / 4, C], and passes through the first group of Swin Transformer Blocks modules;

[0113] When the token reaches the first Patch Merging layer, each group of 2×2 adjacent patches is spliced, the number of tokens becomes 1 / 4 of the original, and the feature dimension of the token becomes [H / 8, W / 8, 2C]. It passes through the second group of Swin Transformer modules and 4 stages in sequence. The number of tokens is gradually reduced through the Patch Merging layer (PatchMeraging); each stage changes the dimension of the tensor, thus forming a hierarchical representation.

[0114] The structure of the Swin Transformer Block network is as follows Figure 3As shown in the figure, Swin Transformer Blocks uses the window multi-head self-attention mechanism (W-MSA) and the window moving multi-head self-attention mechanism (SW-MSA), and its attention is expressed as:

[0115]

[0116] Among them, Q w , K w 、V w is the query, key, and value matrix within each window, Softmax represents the normalized exponential function, T represents the transpose operation, and dk is the smoothing factor;

[0117] The input feature map is divided into multiple non-intersecting windows. The W-WSA module only calculates attention in each window through the normalization layer, W-MSA unit, multi-layer perceptron and skip connection. In order to transfer information between different windows, the SW-WSA module will splice different windows and calculate attention through the normalization layer, SW-MSA unit, multi-layer perceptron and skip connection. The calculation of a SwinTransformer Blocks can be expressed as:

[0118]

[0119] Among them, LN represents the normalization operation, z l-1 represents the output feature of the l-1th time, represents the output feature of the W-MSA unit for the lth time, z l represents the output features of the lth multi-layer perceptron, represents the output characteristics of the l+1th time through the SW-MSA unit, z l+1 Represents the output features of the multi-layer perceptron after l+1 times.

[0120] The output layer in the ST-transformer model network includes a normalization layer, a global pooling layer, and a fully connected layer; the normalization layer is used to maintain the stability of the model training, the global pooling layer is responsible for extracting global features without adding additional parameters, and the fully connected layer is responsible for integrating the previous feature information to make the final classification decision. The loss function is multi-classification cross entropy Among them, K represents the number of interference sources, y q Indicates a label, a q It represents the result of classification output, and q represents the type of interference source.

[0121] The trained ST-Transformer model is obtained, and the recognition performance of the interference source type recognition model is verified through the test set.

[0122] S4: After filtering and short-time Fourier transform processing, the CSI data of the interference source type to be tested generated in real time is input into the verified interference source type identification model to obtain the result of interference source type identification, that is, the main types of wireless interference sources existing in the output environment.

[0123] The trained interference source type identification model is deployed, and the real-time CSI data of the interference source type to be tested is filtered and short-time Fourier transform processed, and then input into the verified interference source type identification model to obtain the result of interference source type identification, that is, the main types of wireless interference sources existing in the output environment.

[0124] Next, the wireless interference source detection device proposed in accordance with the embodiment of the present application is described with reference to the accompanying drawings.

[0125] Figure 6 It is a block diagram of a wireless interference source detection device according to an embodiment of the present application.

[0126] like Figure 6 As shown, the wireless interference source detection device 10 includes: a first acquisition module 101 , a building module 102 and a training module 103 .

[0127] Among them, the first acquisition module 101 is used to obtain a training data set, wherein the training data set includes channel state data and interference source type labels under different wireless interference sources; the establishment module 102 is used to establish an interference source type recognition model based on the Transformer model, wherein the interference source type recognition model includes a block division layer, a feature extraction layer, a block merging layer and an output layer, the block division layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit, the feature extraction layer is used to extract features from multiple block units and perform feature learning, the block merging layer is used to merge adjacent image blocks, and the output layer is used to output the type of wireless interference source; the training module 103 is used to train the interference source type recognition model using the training data set until the recognition performance of the interference source type recognition model reaches a preset standard to stop training, and use the interference source type recognition model to identify the type of wireless interference source in the target wireless interference environment.

[0128] In an embodiment of the present application, the feature extraction layer includes multiple Swin Transformer Block modules, each of which includes a window multi-head self-attention mechanism and a window moving multi-head self-attention mechanism, which are used to divide multiple block units into multiple non-intersecting windows, wherein:

[0129] Windowed multi-head self-attention mechanism, which is used to calculate attention within each window through normalization layer, windowed multi-head self-attention mechanism, multi-layer perceptron and skip connection;

[0130] The window moving multi-head self-attention mechanism is used to splice different windows and calculate attention through normalization layers, window moving multi-head self-attention mechanism units, multi-layer perceptrons and jump connections.

[0131] In an embodiment of the present application, the output layer includes: a normalization layer, a global pooling layer and a fully connected layer, wherein the normalization layer is used to maintain the stability of the interference source type identification model training; the global pooling layer is used to extract global features from the channel state time-frequency diagram; the fully connected layer is used to integrate all feature information in the block partition layer, the feature extraction layer and the block merging layer, and use the loss function to determine the type of wireless interference source.

[0132] In the embodiment of the present application, the loss function is:

[0133]

[0134] Among them, K represents the number of interference sources, y q Indicates a label, a q It represents the result of classification output, and q represents the type of interference source.

[0135] In an embodiment of the present application, the first acquisition module 101 is further used to: obtain channel state data under different wireless interference environments; filter, normalize, preprocess and Fourier transform the channel state data to obtain a corresponding channel state time-frequency diagram, and mark the channel state time-frequency diagram with corresponding interference source type labels to obtain a training data set of channel state data in different interference sources.

[0136] It should be noted that the aforementioned explanation of the embodiment of the wireless interference source detection method is also applicable to the wireless interference source detection device of this embodiment, and will not be repeated here.

[0137] According to the wireless interference source detection device proposed in the embodiment of the present application, the channel state data and interference source type labels under different wireless interference sources can be used to train the interference source type identification model until the recognition performance of the model reaches a preset standard and the training is stopped, thereby ensuring the robustness and generalization ability of the model. The interference source type identification model includes block division, feature extraction, block merging and output layer, thereby realizing effective detection of wireless interference source types.

[0138] Figure 7 It is a block diagram of a wireless interference source detection device according to another embodiment of the present application.

[0139] like Figure 7 As shown, the wireless interference source detection device 20 includes: a second acquisition module 201 , a processing module 202 and an input module 203 .

[0140] Among them, the second acquisition module 201 is used to obtain channel state data in the target wireless interference environment; the processing module 202 is used to filter and Fourier transform the channel state data to obtain a channel state time-frequency diagram; the input module 203 is used to input the channel state time-frequency diagram into the interference source type identification model, and the interference source type identification model outputs the wireless interference source detection result in the target wireless interference environment, wherein the interference source type identification model is established based on the Transformer model.

[0141] It should be noted that the aforementioned explanation of the embodiment of the wireless interference source detection method is also applicable to the wireless interference source detection device of this embodiment, and will not be repeated here.

[0142] According to the wireless interference source detection device proposed in the embodiment of the present application, by pre-processing the channel state data in the target wireless interference environment and inputting a pre-trained interference source type identification model, the type of wireless interference source can be quickly and accurately detected, which is suitable for real-time monitoring needs in practical applications. The interference source type identification model has been fully trained and verified, and has high reliability and accuracy, thereby improving the accuracy of model identification.

[0143] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0144] A memory 801 , a processor 802 , and a computer program stored in the memory 801 and executable on the processor 802 .

[0145] When the processor 802 executes the program, the wireless interference source detection method provided in the above embodiment is implemented.

[0146] Furthermore, the electronic device further comprises:

[0147] The communication interface 803 is used for communication between the memory 801 and the processor 802 .

[0148] The memory 801 is used to store computer programs that can be executed on the processor 802 .

[0149] The memory 801 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0150] If the memory 801, the processor 802 and the communication interface 803 are implemented independently, the communication interface 803, the memory 801 and the processor 802 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0151] Optionally, in a specific implementation, if the memory 801, the processor 802 and the communication interface 803 are integrated on a chip, the memory 801, the processor 802 and the communication interface 803 can communicate with each other through an internal interface.

[0152] The processor 802 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0153] An embodiment of the present application also provides a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the above-mentioned wireless interference source detection method is implemented.

[0154] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0155] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0156] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0157] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one or a combination of multiple of the following technologies known in the art: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0158] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

Claims

1. A method for detecting a wireless interference source, characterized in that: The following steps are involved: Acquire a training data set, wherein the training data set includes channel state data and interference source type labels under different wireless interference sources; An interference source type recognition model is established based on the Transformer model, wherein the interference source type recognition model includes a block division layer, a feature extraction layer, a block merging layer and an output layer, wherein the block division layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit, the feature extraction layer is used to extract features from multiple block units and perform feature learning, the block merging layer is used to merge adjacent image blocks, and the output layer is used to output the type of wireless interference source; The interference source type identification model is trained using the training data set until the identification performance of the interference source type identification model reaches a preset standard and the training is stopped, and the interference source type identification model is used to identify the type of wireless interference source in the target wireless interference environment.

2. The method for detecting a wireless interference source according to claim 1, characterized in that: The feature extraction layer includes multiple Swin Transformer Block modules, each of which includes a window multi-head self-attention mechanism and a window moving multi-head self-attention mechanism, which are used to divide the multiple block units into multiple non-intersecting windows, wherein: The window multi-head self-attention mechanism is used to calculate attention in each window through a normalization layer, a window multi-head self-attention mechanism, a multi-layer perceptron and a skip connection; The window moving multi-head self-attention mechanism is used to splice different windows and calculate attention through a normalization layer, a window moving multi-head self-attention mechanism unit, a multi-layer perceptron and a jump connection.

3. The method for detecting a wireless interference source according to claim 1, characterized in that: The output layer includes: a normalization layer, a global pooling layer and a fully connected layer, wherein: The normalization layer is used to maintain the stability of the interference source type identification model training; The global pooling layer is used to extract global features from the channel state time-frequency diagram; The fully connected layer is used to integrate all feature information in the block division layer, the feature extraction layer and the block merging layer, and determine the type of the wireless interference source by using a loss function.

4. The method for detecting a wireless interference source according to claim 3, characterized in that: The loss function is: Among them, K represents the number of interference sources, y q Indicates a label, a q It represents the result of classification output, and q represents the type of interference source.

5. The method for detecting a wireless interference source according to claim 1, characterized in that: The step of obtaining a training data set includes: Obtain channel status data under different wireless interference environments; The channel state data is filtered, normalized, preprocessed, and Fourier transformed to obtain a corresponding channel state time-frequency diagram, and the channel state time-frequency diagram is labeled with a corresponding interference source type label to obtain a training data set of channel state data in different interference sources.

6. A method for detecting a wireless interference source, characterized in that: The following steps are involved: Obtain channel status data in the target wireless interference environment; The channel state data is filtered and Fourier transformed to obtain a channel state time-frequency diagram; The channel state time-frequency diagram is input into an interference source type identification model, and the interference source type identification model outputs a wireless interference source detection result under the target wireless interference environment, wherein the interference source type identification model is established based on a Transformer model.

7. A wireless interference source detection device, characterized in that: include: A first acquisition module is used to acquire a training data set, wherein the training data set includes channel state data and interference source type labels under different wireless interference sources; Establish a module, used to establish an interference source type identification model based on the Transformer model, wherein the interference source type identification model includes a block division layer, a feature extraction layer, a block merging layer and an output layer, the block division layer is used to divide the channel state time-frequency diagram into multiple image blocks of the same size, and each image block is used as a block unit, the feature extraction layer is used to extract features from multiple block units and perform feature learning, the block merging layer is used to merge adjacent image blocks, and the output layer is used to output the type of wireless interference source; A training module is used to train the interference source type identification model using the training data set until the recognition performance of the interference source type identification model reaches a preset standard and then stop training, and use the interference source type identification model to identify the type of wireless interference source in the target wireless interference environment.

8. A wireless interference source detection device, characterized in that: include: A second acquisition module is used to acquire channel state data in a target wireless interference environment; A processing module, used for filtering and Fourier transforming the channel state data to obtain a channel state time-frequency diagram; An input module is used to input the channel state time-frequency diagram into an interference source type identification model, and the interference source type identification model outputs the wireless interference source detection result under the target wireless interference environment, wherein the interference source type identification model is established based on the Transformer model.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the wireless interference source detection method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: The computer program or instruction is executed by a processor to implement the wireless interference source detection method according to any one of claims 1 to 6.

Images