WIFI signal indoor imaging method based on improved VAE-MoE architecture

Abstract

The invention discloses a WIFI signal indoor imaging method based on an improved VAE-MoE architecture. The WIFI signal indoor imaging method comprises the steps of path planning and data acquisition, data preprocessing, analog signal propagation, linearization physical model construction, deep learning model training and image generation. According to the method, two groups of WIFI signal devices and indoor cameras are arranged, RSSI information data are collected through path planning and WIFI signals, staged optimization is performed in combination with physical modeling and a deep learning technology, the data imaging accuracy is remarkably improved, the indoor imaging problem in a limited data environment is solved, and the indoor imaging quality is improved. And the method has relatively high technical innovation and practical application value.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a method for indoor imaging of WIFI signals based on an improved VAE-MoE architecture. Background Art

[0002] Real-time monitoring and analysis of indoor environments are of great significance in applications such as smart homes, security monitoring, indoor navigation, and the Internet of Things. Traditional indoor detection methods, such as LiDAR, infrared detection, and ultrasonic detection, are usually limited by expensive equipment and are limited by equipment installation and environmental adaptation requirements in practical applications. In contrast, indoor detection based on wireless communication signals (such as WIFI signals) has broad application prospects due to its wide applicability, low cost, and high signal flexibility.

[0003] As a mature wireless communication technology, WIFI signals are now widely used in indoor positioning and other fields. The rapid development of deep learning technology in recent years has led to the attention of WIFI signal imaging methods based on deep learning, and has achieved remarkable results in the field of image recognition and processing. Through the feature extraction capability of deep learning, the processing effect of WIFI signal data can be significantly improved, thereby enhancing the accuracy and clarity of indoor imaging. However, the sparsity and complexity of WIFI signals in practical applications and how to effectively preprocess and optimize data to remove noise and interference are challenges. In addition, the propagation characteristics of WIFI signals are closely related to the physical environment. How to combine physical models with deep learning methods to simulate the signal propagation process and improve the high precision and accuracy of imaging results is still a hot topic in current research. In the existing WIFI signal indoor imaging technology, the propagation of WIFI signals is affected by many factors, such as multipath effects, wall occlusion and signal attenuation. The traditional WIFI signal received signal strength (RSSI) analysis method is easily interfered in complex environments, and the precision and accuracy of imaging results are limited. In addition, the penetration and reflection characteristics of WIFI signals make indoor imaging more complicated, and traditional image processing methods are difficult to fully cope with these challenges. Therefore, developing a method for indoor imaging of WIFI signals based on the improved VAE-MoE architecture is the key to solving the above problems. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a WIFI signal indoor imaging method based on an improved VAE-MoE architecture, introduce VAE module, MoE module and Self-Attention, and successfully improve the quality of generated images based on WIFI signal RSSI data and RGB image data. Self-Attention effectively captures the long-range dependency information in the image, making the generated image more accurate.

[0005] The object of the present invention is achieved by comprising the following steps: S100, path planning and data acquisition: The WIFI transmitter is installed on a mobile device, and the WIFI receiver is installed on another mobile device. The two mobile devices are located on opposite sides of the building to be tested, and move around the periphery of the building to be tested in a clockwise or counterclockwise direction. During the movement, the WIFI signal emitted by the WIFI transmitter is received by the WIFI receiver, and the time-synchronized WIFI signal RSSI data and the location data of the mobile device are obtained; the indoor RGB image is collected using an RGB camera; S200, data preprocessing: preprocess the original WIFI signal data, including distance attenuation compensation and sparse signal processing, to ensure data integrity and provide support for precision imaging; it should be made clear that the WIFI signal strength usually weakens as the signal propagation distance increases, especially in indoor environments, where the presence of objects such as walls and furniture can cause signal attenuation, and during the WIFI signal processing, the received signal may be interfered by noise or other non-ideal factors, resulting in reduced signal quality. To ensure that the received WIFI signal data is suitable for subsequent physical modeling and high-precision imaging tasks, distance attenuation compensation and sparse signal processing are effective ways to preprocess the received RSSI data; S300, simulating signal propagation: constructing a volume integral wave equation for the preprocessed data, and establishing the relationship between signal propagation and object distribution; this step introduces the volume integral wave equation, which takes into account the influence of all scatterers in the target area, and provides the necessary physical basis for accurate modeling of signal propagation from the transmitting end to the receiving end and accurate imaging of WIFI signals in the present invention; S400, constructing a linearized physical model: based on the volume integral wave equation, Rytov approximation linearization is introduced and physical modeling is performed by discretizing the volume integral wave equation and the RSSI simplified model; based on the preliminary processing based on the volume integral wave equation, Rytov approximation is further introduced for linear processing; the corresponding physical modeling process is completed by discretizing the volume integral wave equation and combining it with the RSSI simplified model; The present invention can effectively extract useful features from signals and remove noise and interference by introducing Rytov linear approximation and sparse signal processing technology, thereby achieving more accurate indoor area image reconstruction; S500, deep learning model training and image generation: extract potential features from WIFI signal RSSI data and RGB image data through MLP, convert the WIFI signal RSSI data and RGB image data into high-dimensional feature vectors of fixed dimensions, and then use the VAE encoder module to encode the multimodal input data to generate the mean and variance of the latent space, and then use the MoE module to perform weighted combination on the latent space, and then generate RGB images through the generator module, and then use the discriminator module to judge the authenticity of the generated image, and optimize the generator through training, and finally output a high-precision image; effectively process complex multimodal input data (WIFI signal and image data), ensure full use of data and improve indoor imaging accuracy in deep learning, especially in terms of detail retention and global structural consistency; Among them, in step S100, the mobile device can use a self-programmed unmanned vehicle, which is equipped with an antenna and precise positioning, and efficiently plans the route for the unmanned vehicle to collect WIFI signal RSSI data, collects multi-perspective data, improves data collection efficiency, and ensures data integrity; this self-programmed unmanned vehicle can use a commercially available unmanned vehicle, equipped with a positioning antenna, a PC interface, and draw the edge shape and perimeter of the unknown house to be measured in the form of lines on the PC interface, pre-set the planned route, and when the unmanned vehicle travels along the planned route, the positioning antenna receives the driving path information of the two unmanned vehicles, ensuring that the positioning antenna is accurately aligned, and performing precise positioning on the PC interface to improve the accurate recording of WIFI signals and the high-quality generation of images.

[0006] The moving path of the mobile device can be that one mobile device is located at the south of the building to be tested, and the other is located at the north of the building to be tested. The two mobile devices move clockwise at the same time, covering different parts of the path to improve the collection efficiency.

[0007] Preferably, in step S100, the mobile device moves and collects data along multiple layers of moving paths, specifically: the initial layer moving path of the mobile device is 0.5 meters away from the outer wall of the building to be tested, and the initial layer movement is completed by circling the building to be tested, and then a layer of moving path is set every 1 meter in the direction away from the initial layer, and each layer of moving path circles the building to be tested; after each layer of moving path is completed, each mobile device scans the target area again at the 0° starting point, 45° point, 90° point and 135° point of the layer path, and collects wall-penetrating signal characteristics at different angles.

[0008] Preferably, in step S100, the mobile device collects RSSI data and corresponding location data every time it moves 0.1-0.3 meters to ensure data integrity and consistency.

[0009] Preferably, the distance attenuation compensation in step S200 is specifically: by measuring the original RSSI measurement value of the WIFI signal under barrier-free conditions and without any obstruction between the two unmanned vehicles, a distance-related attenuation model of the signal in an open space is obtained; the formula is: in, : Received signal strength (RSSI) raw measurement value; :Derive the distance-dependent power attenuation value based on the obstacle-free path signal measurement; : Remove the signal after distance attenuation for subsequent modeling and imaging.

[0010] The distance attenuation formula is derived from the free space path loss formula: Where d is the distance the signal travels; is the signal wavelength; this formula is based on the free space path loss model of wireless signal propagation, and its principles and formulas are public and widely used in communication theory.

[0011] These obstacle-free path data are used to build a reference model and estimate the distance attenuation portion, which usually does not include the characteristics caused by obstacles and only reflects the power loss of the signal in free space.

[0012] Preferably, the sparse signal processing in step S200 is specifically to organize and improve the data before model building to provide high-quality input for subsequent high-definition modeling. This processing step combines physical modeling with signal optimization and is one of the core components of the entire imaging process. The formula is: Among them, TV(R): as the total variation of the target area, : local change of the i,jth point; Y: received signal strength vector; K: observation matrix, generated based on physical modeling; X: target sparse feature vector.

[0013] This optimization formula and technique are derived from the basic theory of sparse signal processing and are classic methods in the field of compressed sensing. They are open and widely used. Total variation minimization is common in the fields of image processing and signal optimization. Related references include: Rudin-Osher-Fatemi (ROF) model: total variation method for image denoising; The basic formula of compressed sensing: Preferably, the volume integral wave equation in step S300 is: Where E(r) is the electric field received at position r, indicating the actual received WIFI signal strength; E inc (r) is the incident electric field, that is, the signal propagated without obstacles; G(r,r') is the Green function, which describes the propagation characteristics of electromagnetic waves from position r' to position r; O(r') is the physical property at position r' in the target area, such as the dielectric constant or reflectivity of the scatterer; E(r') is the electric field generated by each scatterer in the target area; dv' is the integral of each scattering point in the target area D.

[0014] The volume integral wave equation is a classic physical model for electromagnetic wave propagation and scattering; it combines the incident electric field and the scattered electric field caused by obstacles to describe the propagation process of WIFI signals; through this equation, we can predict the signal strength received at a certain location, which is crucial for subsequent signal processing, physical modeling and imaging tasks. A brief explanation of the above formula: The electric field E(r) is composed of two parts: the incident electric field E inc (r): is the original transmitted signal, the ideal signal when there are no obstacles; Scattered electric field (through the integral term): This part of the electric field is caused by scatterers (such as walls, pillars, etc.) in the target area. It takes into account the influence of each scatterer in the target area and integrates the electric field contributed by each scattering point to finally obtain the synthetic scattered signal.

[0015] Preferably, the physical modeling in step S400 is specifically: S401, perform Rytov approximation, the formula is: in, (r) is the phase change of the signal during propagation. This linearization process effectively reduces the nonlinear calculation in the signal propagation model and improves the modeling efficiency; S402, establish a discretized volume integral wave equation, the formula is: in, is the phase change discrete vector, F is the propagation matrix, which represents the propagation relationship of the signal from the scattering source to the receiving point, and O is the physical characteristic vector of the target area. Through this discretization process, numerical optimization and calculation can be performed more efficiently, which is convenient for subsequent data processing and image reconstruction. The specific discretization process is: The target area D is divided into N discrete voxels, and the center point of each voxel is r n , the integral becomes the summation; Φ(r): phase change; g(r,r n): Simplified propagation function; O(r n ): scattering properties of each voxel in the target area; S403, RSSI simplified model, in order to combine the actual WIFI signal data, the RSSI simplified model is adopted, the signal model after the volume integral wave equation and Rytov approximation is matched with the received signal strength RSSI of the WIFI device, and the simplified model is expressed as: Among them, P ryt Indicates the received signal strength change, F R is the real part of the propagation matrix, O R It is the real part of the physical property of the target area, simplifies the signal strength representation, and combines it with the actual measurement data to improve the accuracy of signal modeling. By simplifying the model, the physical modeling results can be combined with the actual measured RSSI data to provide data input for subsequent deep learning algorithms, and ultimately generate high-precision indoor images.

[0016] In order to simplify the calculation and improve the processing efficiency, the step S400 of the present invention uses the Rytov approximation to linearize the volume integral wave equation. The Rytov approximation converts the original nonlinear wave equation into a linear form, making the calculation process simpler; when performing physical modeling, the volume integral wave equation is further discretized and converted into a matrix form for numerical calculation.

[0017] Step S400 introduces Rytov approximate linearization processing and discretized volume integral wave equation, combined with RSSI simplified model, which can significantly improve the propagation modeling accuracy of WIFI signals in complex indoor environments and enhance the adaptability of the model to wall-penetrating signals. Preferably, in step S500, the physical model and the deep learning model are integrated to process RSSI data and generate high-precision images by constructing a neural network. The deep learning model training and image generation are specifically: S501. Extract potential features from RSSI data through MLP: Enhance data representation through MLP network, pass through convolution layer and pooling layer, the convolution kernel size is 3×1, after processing, design fully connected layer for high-dimensional feature mapping, and finally output a fixed-length high-dimensional feature vector as the input of subsequent VAE and MoE; The latent features of RGB images are extracted through MLP: the input RGB image is reduced in resolution, the channel size is increased to 512 using continuous convolution blocks, the mean and standard deviation of each channel are normalized and LeakyReLU activation is performed on each layer, and the convolution layer is normalized to convert it into a latent space vector, which represents the spatial information and texture features in the image; S502, VAE encoder module processing: Use VAE encoder (variational autoencoder) to encode the RSSI data of the input WIFI signal and the multimodal features of the RGB image data to generate the mean and variance of the latent space. The mean and standard deviation represent the Gaussian distribution of the latent space and are used to describe the distribution of the latent variable Z. This process takes the mean and variance as input, and VAE can sample the latent space and obtain the latent vector Z. This step ensures that the latent vector is continuous and smooth after generation. Among them, the variational inference and reconstruction loss formula of the VAE encoder is: Variational inference formula: Where, given input data x, learn a distribution of latent variables z, the goal is to maximize the marginal log-likelihood; p(x|z) is the generative model (such as the decoder network); q(z|x) is the variational distribution (encoder network); D KL [.||.] is the Kullback-Leibler divergence, which measures the difference between the variational distribution and the true posterior distribution; Reconstruction loss: The reconstruction loss in VAE usually uses mean square error (MSE) or cross entropy, depending on the nature of the data; x is the original image; S503, MoE module processing: by dynamically selecting multiple expert networks to perform weighted combination of input potential vectors to generate more complex and diverse images. The specific process is: S5031. Input latent space vector z: The input latent space vector z is obtained from the VAE encoder and represents the feature representation of the WIFI signal RSSI data and the RGB image data; S5032, MoE weighting: Automatically adjust each expert's contribution based on the input features. The formula is: The above MoE output is a weighted sum, w i (z) is the weight calculated by the gating network and satisfies ∑w i (z) = 1 (usually obtained through softmax activation); given the latent space vector z, the MoE module determines the weights of each expert through the expert selection mechanism; it is set to the output of the i-th expert network, and the weight {u; (z)} comes from a gating network (usually a small neural network); the generated latent vector {f i (z)}; It should be noted that the MoE (Mixture of Experts) module uses multiple expert networks to perform weighted combination of latent space vectors to generate more complex and diverse images; each expert network focuses on a certain part of the latent space and can extract different feature information, and generates the final latent space vector through weighted combination to enhance the image generation expressiveness; the purpose of weighting is to automatically adjust the contribution of each expert according to the input features, making the generation process more flexible and targeted; It should be noted that the Gating network determines which experts should participate in data processing by assigning weights to each expert, ensuring that the model can dynamically select the most appropriate expert for different input data; the working method of the Gating network is based on the softmax activation function to calculate the weight of each expert; in this way, the MoE module enhances the model's expressiveness and flexibility, and can better adapt to diverse input data; it is defined as: Among them, g i (z) is the output of the gating network, indicating the importance of each expert; S504, generator module processing: converting the potential vector z output by the VAE encoder and the MoE module into an indoor area image; S505, discriminator module processing: judging the authenticity of the generated image, evaluating the quality of the generated image, and providing feedback to the generator to optimize the generated result; S506, outputting a high-precision image: after optimization by the discriminator feedback, outputting a high-precision image.

[0018] For training and evaluation, according to the unmanned vehicle path planning, sampling is performed at a fixed step size to obtain time-synchronized RSSI data and image data sets, obtain WIFI data packets and images, align images and WIFI data packets, and divide them into training sets, validation sets, and test sets.

[0019] Preferably, the specific process of the generator module in step S504 is: S5041, fully connected layer processing: first, the latent vector z is mapped to a higher-dimensional feature space to ensure that the input latent vector can be fully processed in the generator; S5042, Self-Attention Layer Processing: Through Self-Attention, the generator dynamically pays attention to different parts of the image generation process, calculates the similarity between each part, and adjusts the weight of each part; the self-attention mechanism is used to capture the global dependencies in the image and improve the details of the generated image; Self-Attention plays a key role in image generation, enhancing the generator's modeling of long-range dependencies and capturing global information in the image; S5043, deconvolution layer (transposed convolution) processing: gradually decode the processed features into high-precision images and restore the image's spatial structure; S5044, output layer: Generate the final image through convolution. The output image contains the visual information of the indoor scene.

[0020] In step S504 of the present invention, in order to enhance the details and consistency of the generated image, the present invention introduces (self-attention mechanism) to improve the quality of the generated image. Self-Attention can capture the global dependencies in the image, making the generated image more detailed and avoiding the problem that the generator only focuses on local information and ignores global information.

[0021] Preferably, the specific process of the self-attention layer processing in step S5042 is: (1) Query, key and value: The shape of the input feature matrix X is N×D, where N is the input feature vector and D is the dimension of each feature; Q , W K , and W V is a learnable weight matrix corresponding to the conversion of query, key and value respectively; (2) Calculation of attention weight: The above is to calculate the dot product between the query and the key to obtain the attention weight matrix, where is a scaling factor used to prevent the gradient from disappearing due to excessive dot product; the weight is then calculated using the softmax function, and the calculation formula is as follows: in is the attention weight of position i to position j, and the output is weight V; (3) Weighted summation: The final output is the weighted sum of the values ​​V by the attention weights, where α is the attention weight matrix and V is the value matrix.

[0022] Preferably, in step S505, the discriminator module (Discriminator) includes multiple convolutional layers and Self-Attention layers, which are used to improve the accuracy and stability of the discriminator. The convolutional layers are responsible for extracting features from the image, while the self-attention layers help the discriminator capture the global structure in the image. The specific process of the discriminator module is: S5051, multi-layer convolution layer processing: extract features from the image, the convolution operation is expressed as: Among them, H (l) is the feature map of layer l, W (l) is the convolution kernel, b (l) is the bias term, * indicates the convolution operation; S5052, Self-Attention Layer Processing: Self-Attention is introduced. Self-Attention can help the discriminator better understand the long-range dependencies in the image, improve the sensitivity to image details, and enhance the accuracy of image true and false judgment. The calculation formula is: Among them, Q, K and V are the matrices of query, key and value respectively, representing different transformations of input features, and softmax is used to calculate attention weights; S5053, Fully Connected Layer processing: After being processed by the convolutional layer and the self-attention layer, the discriminator flattens the feature map and inputs it into the fully connected layer to map the high-level features of the image to the final classification space (0 or 1), that is, to determine whether the image is real or generated, expressed as: Among them, W is the weight matrix, b is the bias term, and σ is the activation function (usually sigmoid), which is used to output a value between 0 and 1, indicating the true probability of the image; S5054, output layer: output the true and false judgment results through the sigmoid activation function, 0 is the generated image, 1 is the real image, the formula is as follows: Among them, p real Represents the probability that the input image is a real image. If p real If it is close to 1, it means the image is more likely to be real; if it is close to 0, it means the image is more likely to be generated.

[0023] The discriminator consists of multiple layers of convolutional layers and self-attention layers. The discriminator module is responsible for judging the authenticity of the generated image. Its goal is to determine whether an image comes from the real data distribution (i.e., the real image) or from the generator (i.e., the generated image). It participates in the image generation process as an auxiliary module, evaluates the quality of the generated image, and provides feedback to the generator to further optimize the generation results.

[0024] Preferably, in the step of generating an image in S506 output, LeakyReLU (rectified linear unit with leakage) is used as the activation function in the neural network architecture for encoding the image and K-divergence loss (KLDivergence Loss) is used to ensure that the latent space distribution conforms to the standard normal distribution; the optimizer adopts the Adam optimizer for complex deep neural networks; The Leak ReLU formula is as follows: Where x: input value, a small constant, usually between [0,1], and the common default value is 0.01; The KL divergence loss formula is as follows: Where μ and σ are the mean and standard deviation output by the VAE encoder, and p(z) is the standard normal distribution; The Adam optimizer formula is as follows: in, are model parameters, is the gradient momentum, v t is the gradient second moment estimate, is the learning rate, is a small constant to prevent division by zero.

[0025] After the generator transforms the latent vector z, i.e., transposed convolution and self-attention, it finally generates an image representing the indoor scene. This image generation is not only based on the structural information of the latent space, but also fully considers the global consistency and details of the image, providing a more realistic image output.

[0026] Compared with the prior art, the present invention has the following technical effects: 1. The present invention deploys two groups of WIFI signal devices and indoor cameras, uses WIFI signals to collect RSSI information data, and combines physical modeling with deep learning technology for phased optimization, thereby significantly improving the accuracy of data imaging; firstly, the volume integral wave equation and Rytov approximation theory are used to linearize and approximate the physical model, and the original WIFI signal data is preprocessed, key features are extracted, and sparse signal processing is performed to remove noise and interference factors, laying the foundation for subsequent high-precision imaging. Secondly, a deep learning architecture is designed. Through multilayer perceptron, variational auto-encoder and mixture of experts, a self-attention mechanism is introduced to generate high-precision indoor area images for data optimized by physical modeling, and further improve the performance of wall-penetrating signal modeling. In the process of collecting WIFI signal data, the collection strategy is optimized through efficient path planning to effectively improve the integrity and accuracy of the data. The present invention solves the indoor imaging problem in a limited data environment by combining the physical model constructed by using volume integral wave equation and Rytov approximation theory and the neural network constructed by using VAE and MoE modules and Self-Attention, and has high technical innovation and practical application value. 2. The present invention introduces multimodal data processing to enhance the expressiveness of deep learning models: by using neural network architectures such as MLP, VAE module, MoE module and Self-Attention, combined with RSSI data and image data of WIFI signals, it effectively processes complex multimodal input data and improves the expressiveness and detail retention of image generation; 3. The present invention improves the precision and accuracy of indoor area imaging. By combining WIFI signals with the neural network architecture in deep learning, especially using Self-Attention, the precision and accuracy of indoor area images are greatly improved; 4. The processing and verification mechanism of the present invention further optimizes the quality of generated images by combining VAE modules with MSE modules, effectively removes noise in the generation process, and improves edge clarity and object recognition accuracy in images; 5. The present invention improves the indoor real-time monitoring capability. By designing an integrity program, it can realize real-time monitoring of indoor structures, personnel distribution, etc., and generate accurate indoor area images, greatly improving the performance and effect of IoT devices in automated management and environmental perception; 6. In terms of security monitoring capabilities, the present invention uses the RSSI data of WIFI signals and deep learning models to penetrate walls to obtain indoor information, providing more accurate indoor object positioning, personnel trajectory tracking and abnormal behavior detection in security monitoring. This technology can effectively make up for the limitations of traditional cameras under light, angle and occlusion conditions, and improve the accuracy and reliability of monitoring; 7. The indoor images generated by the present invention based on WIFI signals and deep learning technology are more flexible, low-cost and easy to deploy, providing accurate indoor maps and object positioning information, and promoting the widespread application of intelligent perception and augmented reality. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 Schematic diagram of introducing VAE module for feature extraction in the present invention; Figure 3 A schematic diagram of a method for generating indoor images by a generator module of the present invention; Figure 4 A schematic diagram of a method for enhancing image precision and accuracy by a discriminator module of the present invention; Figure 5 Schematic diagram of the method for improving indoor image accuracy through self-attention mechanism; Figure 6 This is an imaging effect diagram of the present invention. DETAILED DESCRIPTION

[0028] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.

[0029] Example 1 As attached Figure 1 to Figure 6 The embodiment shown in the figure is based on the improved VAE-MoE architecture WIFI signal indoor imaging method, which includes the following steps: S100, path planning and data acquisition: The WIFI transmitter is installed on an unmanned vehicle, and the WIFI receiver is installed on another mobile device. The two unmanned vehicles are located on opposite sides of the building to be tested, and move around the periphery of the building to be tested in a clockwise or counterclockwise direction. During the movement, the WIFI signal emitted by the WIFI transmitter is received by the WIFI receiver, and the time-synchronized WIFI signal RSSI data and the location data of the unmanned vehicle are obtained; the RGB camera is used to collect indoor RGB images; S200, data preprocessing: preprocess the original WIFI signal data, including distance attenuation compensation and sparse signal processing; in indoor environments, the presence of indoor structures such as walls, furniture, etc. will cause signal attenuation, and the received signal may be affected by noise or other non-ideal factors. At the same time, RGB images are collected indoors as subsequent training objects; therefore, for WIFI signal data, distance attenuation compensation and sparse signal processing are effective methods for preprocessing received RSSI data; S300, simulate signal propagation: after the above processing, the input is a discretized vector, and a volume integral wave equation is constructed for the pre-processed data to establish the relationship between signal propagation and object distribution; this equation comprehensively considers the influence of all scatterers inside the target area, provides an accurate simulation of the signal propagation process from the transmitting source to the receiving point, and provides the necessary physical basis for high-precision imaging of WIFI signals; S400, build a linearized physical model: based on the volume integral wave equation, introduce Rytov approximate linearization and perform physical modeling through discretized volume integral wave equation and RSSI simplified model; S500, deep learning model training and image generation: extract potential features from WIFI signal RSSI data and RGB image data through MLP, convert the RSSI data and RGB image data of WIFI signal into high-dimensional feature vectors of fixed dimension, and then use VAE encoder module to encode multimodal input data to generate the mean and variance of latent space, and then use MoE module to perform weighted combination on latent space, and then generate RGB image through generator module, and then use discriminator module to judge the authenticity of generated image, and optimize generator through training, and finally generate high-precision image through output.

[0030] Example 2 This embodiment is based on the improved VAE-MoE architecture WIFI signal indoor imaging method on the basis of embodiment 1. In step S100, the unmanned vehicle moves and collects data according to multiple layers of moving paths. Specifically, the initial layer moving path of the unmanned vehicle is 0.5 meters away from the outer wall of the building to be tested. The initial layer movement is completed by circling the building to be tested. Then, a layer of moving path is set every 1 meter in the direction away from the initial layer, and each layer of moving path circles the building to be tested. After each layer of moving path is completed, each mobile device scans the target area again at the 0° starting point, 45° point, 90° point and 135° point of the layer path to collect wall-penetrating signal characteristics at different angles.

[0031] Example 3 This embodiment is based on the improved VAE-MoE architecture WIFI signal indoor imaging method on the basis of Embodiment 2. In step S100, the unmanned vehicle collects RSSI data and corresponding position data every time it moves 0.1 meters.

[0032] Example 4 The method for indoor imaging of WIFI signals based on the improved VAE-MoE architecture in this embodiment is based on the third embodiment. The distance attenuation compensation in step S200 is specifically: by measuring the original RSSI measurement value of the WIFI signal under barrier-free conditions and without any obstruction between the two unmanned vehicles, a distance-related attenuation model of the signal in an open space is obtained; the formula is: in, : Received signal strength (RSSI) raw measurement value; :Derive the distance-dependent power attenuation value based on the obstacle-free path signal measurement; : Remove the signal after distance attenuation for subsequent modeling and imaging.

[0033] Example 5 The WIFI signal indoor imaging method based on the improved VAE-MoE architecture in this embodiment is based on Embodiment 4. The sparse signal processing in step S200 is specifically: before the model is built, the data is sorted and improved to provide high-quality input for subsequent high-definition modeling. The formula is: Among them, TV(R): as the total variation of the target area, : local change of the i,jth point; Y: received signal strength vector; K: observation matrix, generated based on physical modeling; X: target sparse feature vector.

[0034] Example 6 The indoor imaging method of WIFI signals based on the improved VAE-MoE architecture in this embodiment is based on Embodiment 5, and the volume integral wave equation in step S300 is: Where E(r) is the electric field received at position r, indicating the actual received WIFI signal strength; E inc (r) is the incident electric field, that is, the signal propagated without obstacles; G(r,r') is the Green function, which describes the propagation characteristics of electromagnetic waves from position r' to position r; O(r') is the physical property at position r' in the target area, such as the dielectric constant or reflectivity of the scatterer; E(r') is the electric field generated by each scatterer in the target area; dv' is the integral of each scattering point in the target area D.

[0035] Example 7 The method for indoor imaging of WIFI signals based on the improved VAE-MoE architecture in this embodiment is based on Embodiment 6, and the physical modeling in step S400 is specifically as follows: S401, perform Rytov approximation, the formula is: in, (r) is the phase change of the signal during propagation; this linearization process effectively reduces the nonlinear calculation in the signal propagation model; S402, establish a discretized volume integral wave equation, convert it into a matrix form for numerical calculation, the calculation process is more efficient, the formula is: in, is the phase change discrete vector, F is the propagation matrix, which represents the propagation relationship of the signal from the scattering source to the receiving point, and O is the physical characteristic vector of the target area; The specific discretization process is: The target area D is divided into N discrete voxels, and the center point of each voxel is r n , the integral becomes the summation; Φ(r): phase change; g(r,r n ): Simplified propagation function; O(r n ): scattering properties of each voxel in the target area; S403, RSSI simplified model, the signal model after volume integral wave equation and Rytov approximation processing corresponds to the received signal strength RSSI of the WIFI device, and the simplified model is expressed as: Among them, P ryt Indicates the received signal strength change, F R is the real part of the propagation matrix, O R is the real part of the physical property of the target area; this method provides data support for subsequent deep learning algorithms.

[0036] Example 8 The indoor imaging method of WIFI signals based on the improved VAE-MoE architecture in this embodiment is based on Embodiment 7. The deep learning model training and image generation in step S500 are specifically as follows: S501. Extract potential features from RSSI data through MLP: Enhance data representation through MLP network, pass through convolution layer and pooling layer, the convolution kernel size is 3×1, after processing, design fully connected layer for high-dimensional feature mapping, and finally output a fixed-length high-dimensional feature vector as the input of subsequent VAE and MoE; The latent features of RGB images are extracted through MLP: the input RGB image is reduced in resolution, the channel size is increased to 512 using continuous convolution blocks, the mean and standard deviation of each channel are normalized and LeakyReLU activation is performed on each layer, and the convolution layer is normalized to convert it into a latent space vector, which represents the spatial information and texture features in the image; S502, VAE encoder module processing: The VAE encoder is composed of a convolutional layer, an MLP and a fully connected layer. The final fully connected layer outputs the mean and standard deviation, which are generated by variational inference and sampled by reparameterization techniques. Specifically, the VAE encoder is used to encode the multimodal features of the RSSI data and RGB image data of the input WIFI signal to generate the mean and variance of the latent space. The mean and standard deviation represent the Gaussian distribution of the latent space and are used to describe the distribution of the latent variable Z. The variational inference and reconstruction loss formula of the VAE encoder is: Variational inference formula: Where, given input data x, learn a distribution of latent variables z, the goal is to maximize the marginal log-likelihood; p(x|z) is the generative model (such as the decoder network); q(z|x) is the variational distribution (encoder network); D KL [.||.] is the Kullback-Leibler divergence, which measures the difference between the variational distribution and the true posterior distribution; Reconstruction loss: The reconstruction loss in VAE usually uses mean square error (MSE) or cross entropy, depending on the nature of the data; x is the original image; this term measures the signal reconstruction error, and the goal is to enable the generative model to reconstruct the original input data; S503, MoE module processing: by dynamically selecting multiple expert networks to perform weighted combination of input potential vectors to generate more complex and diverse images. The specific process is: S5031. Input latent space vector z: The input latent space vector z is obtained from the VAE encoder and represents the feature representation of the WIFI signal RSSI data and the RGB image data; this vector contains the core information of the input data; S5032, MoE weighting: Automatically adjust each expert's contribution according to the input features, making the generation process flexible and targeted. The formula is: The above MoE output is a weighted sum, w i (z) is the weight calculated by the gating network and satisfies ∑w i(z) = 1 (usually obtained through softmax activation); given the latent space vector z, the MoE module determines the weights of each expert through the expert selection mechanism; it is set to the output of the i-th expert network, and the weight {u; (z)} comes from a gating network; the generated latent vector {f i (z)}; The MoE module is used to enhance the latent space representation by dynamically selecting multiple expert networks to perform weighted combinations of input latent vectors to generate more complex and diverse images. S504, generator module processing: converting the potential vector z output by the VAE encoder and the MoE module into an indoor area image; S505, discriminator module processing: judging the authenticity of the generated image, evaluating the quality of the generated image, and providing feedback to the generator to optimize the generated result; S506, output the generated image: after being processed by the discriminator, a high-precision image is output.

[0037] Example 9 The method for indoor imaging of WIFI signals based on the improved VAE-MoE architecture in this embodiment is based on Embodiment 8. In step S504, the generator module includes multiple deconvolution layers and convolution layers for decoding the latent space representation and generating an image that meets the expectations. The specific process of the generator module is as follows: S5041, fully connected layer processing: first, the latent vector z is mapped to a higher-dimensional feature space to ensure that the input latent vector can be fully processed in the generator, laying the foundation for subsequent image generation; S5042, Self-Attention Layer Processing: Through Self-Attention, the generator dynamically pays attention to different parts of the image generation process, calculates the similarity between each part, and adjusts the weight of each part; thereby capturing the long-range dependencies and global structures in the image; S5043, deconvolution layer processing: gradually decode the processed features into high-precision images and restore the image's spatial structure; S5044, output layer: Generate the final image through convolution. The output image contains the visual information of the indoor scene and maintains the accuracy and clarity of the image.

[0038] Example 10 This embodiment is based on the improved VAE-MoE architecture WIFI signal indoor imaging method on the basis of embodiment 9, and the specific process of the self-attention layer processing in step S5042 is: (1) Query, key and value: The shape of the input feature matrix X is N×D, where N is the input feature vector and D is the dimension of each feature; Q , W K , and W V is a learnable weight matrix corresponding to the conversion of query, key and value respectively; (2) Calculation of attention weight: The above is to calculate the dot product between the query and the key to obtain the attention weight matrix, where is a scaling factor used to prevent the gradient from disappearing due to excessive dot product; the weight is then calculated using the softmax function, and the calculation formula is as follows: in is the attention weight of position i to position j, and the output is weight V; (3) Weighted summation: The final output is the weighted sum of the values ​​V by the attention weights, where α is the attention weight matrix and V is the value matrix.

[0039] Embodiment 11 This embodiment is based on the improved VAE-MoE architecture WIFI signal indoor imaging method on the basis of embodiment 10, and the specific process of the discriminator module processing in step S505 is: S5051, multi-layer convolutional layer processing: Process the input image structure, texture and other information, and gradually extract deeper feature representations. The mathematical expression is: Among them, H (l) is the feature map of layer l, W (l) is the convolution kernel, b (l) is the bias term, * indicates the convolution operation; S5052, Self-Attention Layer Processing: Self-Attention is introduced to calculate the similarity between input features to adjust the weighted values ​​of different features and improve the sensitivity to image details and structures. The calculation formula is: Among them, Q, K and V are the matrices of query, key and value respectively, representing different transformations of input features, and softmax is used to calculate attention weights; S5053, fully connected layer processing: After being processed by the convolution layer and the self-attention layer, the discriminator flattens the feature map and inputs it into the fully connected layer to map the high-level features of the image to the final classification space and judge the authenticity of the image, which is expressed as: Among them, W is the weight matrix, b is the bias term, and σ is the activation function (usually sigmoid), which is used to output a value between 0 and 1, indicating the true probability of the image; S5054, output layer: output the true and false judgment results through the sigmoid activation function, 0 is the generated image, 1 is the real image, the formula is as follows: Among them, p real Represents the probability that the input image is a real image. If p real If it is close to 1, it means that the image is more likely to be real; if it is close to 0, it means that the image is more likely to be generated; S5055, Optimizer: Use the Adam optimizer to update the parameters of the network. The Adam optimizer accelerates convergence by adaptively adjusting the learning rate and avoids the problem of gradient explosion or gradient disappearance. The formula is as follows: in, are the parameters of the model, is the momentum of the gradient, v t is the second-order moment estimate of the gradient, is the learning rate, is a small constant to prevent division by zero.

[0040] Example 12 The method for indoor imaging of WIFI signals based on the improved VAE-MoE architecture in this embodiment is based on the embodiment 11. S506 outputs the step of generating an image. The activation function in the neural network architecture uses Leaky ReLU (rectified linear unit with leakage) for encoding the image and K-divergence loss (KLDivergence Loss) to ensure that the potential space distribution conforms to the standard normal distribution; the optimizer uses the Adam optimizer for complex deep neural networks; The Leak ReLU formula is as follows: Where x: input value, a small constant, usually between [0,1], and the common default value is 0.01; The KL divergence loss formula is as follows: Where μ and σ are the mean and standard deviation output by the VAE encoder, and p(z) is the standard normal distribution; The Adam optimizer formula is as follows: in, are model parameters, is the gradient momentum, v t is the gradient second moment estimate, is the learning rate, is a small constant to prevent division by zero.

Claims

1. A method for indoor imaging of WIFI signals based on an improved VAE-MoE architecture, characterized in that The following steps are involved: S100, path planning and data acquisition: The WIFI transmitter is installed on a mobile device, and the WIFI receiver is installed on another mobile device. The two mobile devices are located on opposite sides of the building to be tested, and move around the periphery of the building to be tested in a clockwise or counterclockwise direction. During the movement, the WIFI signal emitted by the WIFI transmitter is received by the WIFI receiver, and the time-synchronized WIFI signal RSSI data and the location data of the mobile device are obtained; Use RGB camera to collect indoor RGB images; S200, data preprocessing: preprocessing the original WIFI signal data, including distance attenuation compensation and sparse signal processing; S300, simulating signal propagation: constructing a volume integral wave equation for the preprocessed data, and establishing a relationship between signal propagation and object distribution; S400, build a linearized physical model: based on the volume integral wave equation, introduce Rytov approximate linearization and perform physical modeling through discretized volume integral wave equation and RSSI simplified model; S500, deep learning model training and image generation: extract potential features from WIFI signal RSSI data and RGB image data through MLP, convert the RSSI data and RGB image data of WIFI signal into high-dimensional feature vectors of fixed dimension, and then use VAE encoder module to encode multimodal input data to generate the mean and variance of latent space, and then use MoE module to perform weighted combination on latent space, and then generate RGB image through generator module, and then use discriminator module to judge the authenticity of generated image, and optimize generator through training, and finally generate high-precision image through output.

2. The indoor imaging method of WIFI signals based on the improved VAE-MoE architecture according to claim 1 is characterized in that In step S100, the mobile device moves and collects data along multiple layers of moving paths. Specifically, the initial moving path of the mobile device is 0.5 meters away from the outer wall of the building to be tested. The initial moving path is completed by circling the building to be tested. Then, a moving path is set every 1 meter in the direction away from the initial layer, and each moving path circles the building to be tested. After each moving path is completed, each mobile device scans the target area again at the 0° starting point, 45° point, 90° point and 135° point of the path to collect the wall-penetrating signal characteristics at different angles.

3. The indoor imaging method of WIFI signals based on the improved VAE-MoE architecture according to claim 2 is characterized in that In step S100 , the mobile device collects RSSI data and corresponding position data every time it moves 0.1 meter to 0.3 meter.

4. The method for indoor imaging of WIFI signals based on improved VAE-MoE architecture according to claim 1 is characterized in that The distance attenuation compensation in step S200 is specifically: by measuring the original RSSI measurement value of the WIFI signal under barrier-free conditions and without any obstruction between the two unmanned vehicles, the distance-related attenuation model of the signal in the open space is obtained; the formula is: in, : Received signal strength (RSSI) raw measurement value; :Derive the distance-dependent power attenuation value based on the obstacle-free path signal measurement; : Remove the signal after distance attenuation for subsequent modeling and imaging.

5. The indoor imaging method of WIFI signals based on the improved VAE-MoE architecture according to claim 1 is characterized in that The specific sparse signal processing in step S200 is to organize and improve the data before model building to provide high-quality input for subsequent high-definition modeling. The formula is: Among them, TV(R): as the total variation of the target area, : local change of the i,jth point; Y: received signal strength vector; K: observation matrix, generated based on physical modeling; X: target sparse feature vector.

6. The indoor imaging method of WIFI signals based on the improved VAE-MoE architecture according to claim 1 is characterized in that S300 step volume-integrated wave equation: Where E(r) is the electric field received at position r, indicating the actual received WIFI signal strength; E inc (r) is the incident electric field, that is, the signal propagated without obstacles; G(r,r') is the Green function, which describes the propagation characteristics of electromagnetic waves from position r' to position r; O(r') is the physical property at position r' in the target area, such as the dielectric constant or reflectivity of the scatterer; E(r') is the electric field generated by each scatterer in the target area; dv' is the integral of each scattering point in the target area D.

7. The method for indoor imaging of WIFI signals based on improved VAE-MoE architecture according to claim 1 is characterized in that The physical modeling of the S400 step is as follows: S401, perform Rytov approximation, the formula is: in, (r) is the phase change of the signal during propagation; S402, establish a discretized volume integral wave equation, the formula is: in, is the phase change discrete vector, F is the propagation matrix, which represents the propagation relationship of the signal from the scattering source to the receiving point, and O is the physical characteristic vector of the target area; S403, RSSI simplified model, the signal model after volume integral wave equation and Rytov approximation processing corresponds to the received signal strength RSSI of the WIFI device, and the simplified model is expressed as: Among them, P ryt Indicates the received signal strength change, F R is the real part of the propagation matrix, O R It is the real part of the physical property of the target area.

8. The method for indoor imaging of WIFI signals based on improved VAE-MoE architecture according to claim 1 is characterized in that The specific details of deep learning model training and image generation in step S500 are: S501. RSSI data extracts potential features through MLP: The MLP network is used to enhance data representation. The convolution layer and pooling layer are used. The convolution kernel size is 3×1. After processing, a fully connected layer is designed for high-dimensional feature mapping. Finally, a fixed-length high-dimensional feature vector is output as the input of the subsequent VAE and MoE. The latent features of RGB images are extracted through MLP: the input RGB image is reduced in resolution, the channel size is increased to 512 using continuous convolution blocks, the mean and standard deviation of each channel are normalized and LeakyReLU activation is performed on each layer, and the convolution layer is normalized to convert it into a latent space vector, which represents the spatial information and texture features in the image; S502, VAE encoder module processing: Use the VAE encoder to encode the RSSI data and RGB image data multimodal features of the input WIFI signal to generate the mean and variance of the latent space. The mean and standard deviation represent the Gaussian distribution of the latent space and are used to describe the distribution of the latent variable Z. Among them, the variational inference and reconstruction loss formula of the VAE encoder is: Variational inference formula: Where, given input data x, learn the distribution of a latent variable z, the goal is to maximize the marginal log-likelihood; p(x|z) is the generative model; q(z|x) is the variational distribution; D KL [.||.] is the Kullback-Leibler divergence, which measures the difference between the variational distribution and the true posterior distribution; Reconstruction loss: The reconstruction loss in VAE usually uses mean square error (MSE) or cross entropy, depending on the nature of the data; x is the original image; S503, MoE module processing: by dynamically selecting multiple expert networks to perform weighted combination of input potential vectors to generate more complex and diverse images. The specific process is: S5031. Input latent space vector z: The input latent space vector z is obtained from the VAE encoder and represents the feature representation of the WIFI signal RSSI data and the RGB image data; S5032, MoE weighting: Automatically adjust each expert's contribution based on the input features. The formula is: The above MoE output is a weighted sum, w i (z) is the weight calculated by the gating network and satisfies ∑w i (z) = 1 (usually obtained through softmax activation); given the latent space vector z, the MoE module determines the weights of each expert through the expert selection mechanism; it is set to the output of the i-th expert network, and the weight {u; (z)} comes from a gating network; the generated latent vector {f i (z)}; S504, generator module processing: converting the potential vector z output by the VAE encoder and the MoE module into an indoor area image; S505, discriminator module processing: judging the authenticity of the generated image, evaluating the quality of the generated image, and providing feedback to the generator to optimize the generated result; S506, outputting a high-precision image: after being evaluated and optimized by the discriminator, outputting a high-precision image.

9. The method for indoor imaging of WIFI signals based on improved VAE-MoE architecture according to claim 8 is characterized in that The specific process of the generator module in step S504 is: S5041, fully connected layer processing: first, the latent vector z is mapped to a higher-dimensional feature space to ensure that the input latent vector can be fully processed in the generator; S5042, Self-Attention Layer Processing: Through Self-Attention, the generator dynamically pays attention to different parts of the image generation process, calculates the similarity between each part, and adjusts the weight of each part; S5043, deconvolution layer processing: gradually decode the processed features into high-precision images and restore the image's spatial structure; S5044, output layer: Generate the final image through convolution. The output image contains the visual information of the indoor scene.

10. The method for indoor WIFI signal imaging based on improved VAE-MoE architecture according to claim 9 is characterized in that The specific processing of the self-attention layer in step S5042 is: (1) Query, key and value: The shape of the input feature matrix X is N×D, where N is the input feature vector and D is the dimension of each feature; Among them, W Q , W K , and W V is a learnable weight matrix corresponding to the conversion of query, key and value respectively; (2) Calculation of attention weight: The above is to calculate the dot product between the query and the key to obtain the attention weight matrix, where is a scaling factor used to prevent the gradient from disappearing due to excessive dot product; the weight is then calculated using the softmax function, and the calculation formula is as follows: in is the attention weight of position i to position j, and the output is weight V; (3) Weighted summation: The final output is the weighted sum of the values ​​V by the attention weights, where α is the attention weight matrix and V is the value matrix.

11. The method for indoor WIFI signal imaging based on improved VAE-MoE architecture according to claim 8, characterized in that The specific process of the discriminator module in step S505 is: S5051, multi-layer convolutional layer processing: extract features from the image, the convolution operation is expressed as: Among them, H (l) is the feature map of layer l, W (l) is the convolution kernel, b (l) is the bias term, * indicates the convolution operation; S5052, Self-Attention Layer Processing: Self-Attention is introduced, and the calculation formula is: Among them, Q, K and V are the matrices of query, key and value respectively, representing different transformations of input features, and softmax is used to calculate attention weights; S5053, fully connected layer processing: After being processed by the convolutional layer and the self-attention layer, the discriminator flattens the feature map and inputs it into the fully connected layer to map the high-level features of the image to the final classification space, that is, to determine whether the image is real or generated, expressed as: Among them, W is the weight matrix, b is the bias term, and σ is the activation function, which is used to output a value between 0 and 1, indicating the true probability of the image; S5054, output layer: output the true and false judgment results through the sigmoid activation function, 0 is the generated image, 1 is the real image, the formula is as follows: Among them, p real Represents the probability that the input image is a real image. If p real If it is close to 1, it means the image is more likely to be real; if it is close to 0, it means the image is more likely to be generated.

12. The method for indoor imaging of WIFI signals based on improved VAE-MoE architecture according to claim 8, characterized in that S506 outputs the image generation step. For the activation function in the neural network architecture, Leaky ReLU is used to encode the image and K divergence loss is used to ensure that the potential space distribution conforms to the standard normal distribution. The optimizer uses the Adam optimizer for complex deep neural networks. The Leak ReLU formula is as follows: Where x: input value, a small constant, usually between [0,1], and the common default value is 0.01; The KL divergence loss formula is as follows: Where μ and σ are the mean and standard deviation output by the VAE encoder, and p(z) is the standard normal distribution; The Adam optimizer formula is as follows: in, are model parameters, is the gradient momentum, v t is the gradient second moment estimate, is the learning rate, is a small constant to prevent division by zero.

Images