Channel fingerprint transmission method and device and computer equipment

Abstract

The invention relates to the technical field of positioning, in particular to a channel fingerprint transmission method and device and computer equipment. The method and the device are used for encrypting and compressing channel fingerprints through an encoding network and a decoding network in a transmission process of the channel fingerprints, reducing storage requirements and communication overhead of the channel fingerprints and enhancing protection on user privacy. An approximate posteriori of the channel fingerprint is generated by introducing a pre-trained coding network and a compressed channel fingerprint is generated by sampling. A pre-designed coding network can compress channel fingerprints, and resources consumed by transmission and storage are reduced. And the privacy protection of the user using the location service is enhanced by avoiding transmission of accurate location information.

Description

Technical Field

[0001] This application relates to the field of positioning technology, and in particular to a channel fingerprint transmission method, apparatus and computer equipment. Background Technology

[0002] Currently, the demand for location-based services is increasing. With the continuous development of modern communication technologies, positioning methods based on converged communication have emerged. One implementation of this converged communication-based positioning method uses channel fingerprints based on channel state information. Channel fingerprints consist of channel data describing the device's location and can effectively characterize the wireless environment features of a given location.

[0003] However, this technology currently has the following problems: Channel fingerprints themselves do not explicitly provide location information, requiring the pre-construction of a channel fingerprint database for comparison. The complexity of channel fingerprints leads to significant storage overhead for the database, and determining how to set an update strategy to ensure the accuracy of stored channel fingerprints is also a problem. Furthermore, the exchange of precise location data between the terminal and the base station during channel fingerprint acquisition can easily lead to privacy leaks.

[0004] Therefore, a new method for transmitting channel fingerprints is urgently needed to solve the above problems. Summary of the Invention

[0005] This application provides a channel fingerprint transmission method, apparatus, and computer device, which encrypt and compress the channel fingerprint through an encoding network and a decoding network during the transmission process, thereby reducing the storage requirements and communication overhead of the channel fingerprint and enhancing the protection of user privacy.

[0006] To achieve the above objectives, the main technical solutions adopted in this application include: In a first aspect, embodiments of this application provide a channel fingerprinting transmission method, applied to a terminal, the method comprising: Obtain the channel information of the terminal's location and generate a channel fingerprint; The approximate posterior of the channel fingerprint is determined by a pre-trained coding network, and the approximate posterior is deterministically sampled to obtain a compressed channel fingerprint. The compressed channel fingerprint is sent to the server for processing.

[0007] Secondly, embodiments of this application provide a channel fingerprinting transmission method, applied to a server, the method comprising: The receiving terminal sends a compressed channel fingerprint, which is generated by the terminal based on the channel information of its location. The terminal then uses a pre-trained coding network to determine the approximate posterior of the channel fingerprint and performs deterministic sampling on the approximate posterior to generate the compressed channel fingerprint. Perform at least one of the following processes based on the compressed channel fingerprint: The compressed channel fingerprint is reconstructed using a pre-trained decoding network to obtain the channel fingerprint; The location of the terminal is determined by searching for a record corresponding to the compressed channel fingerprint in a pre-established channel fingerprint database.

[0008] Thirdly, embodiments of this application provide a channel fingerprinting transmission device, the device comprising: A channel fingerprint generation unit is used to obtain channel information of the location of the terminal and generate a channel fingerprint; A posterior sampling unit is used to determine the approximate posterior of the channel fingerprint through a pre-trained coding network and perform deterministic sampling on the approximate posterior to obtain a compressed channel fingerprint; The first transmission unit is used to send the compressed channel fingerprint to the server so that the server can process the compressed channel fingerprint.

[0009] Fourthly, embodiments of this application provide a channel fingerprinting transmission device, the device comprising: The second information transmission unit is used to receive the compressed channel fingerprint sent by the terminal. The compressed channel fingerprint is generated by the terminal based on the channel information of its location, and the approximate posterior of the channel fingerprint is determined by the pre-trained coding network, and then the approximate posterior is deterministically sampled to generate the compressed channel fingerprint. The data processing unit is configured to perform at least one of the following processes based on the compressed channel fingerprint: reconstructing the compressed channel fingerprint using a pre-trained decoding network to obtain the channel fingerprint; searching for a record corresponding to the compressed channel fingerprint in a pre-established channel fingerprint database; and determining the location of the terminal based on the record.

[0010] Fifthly, embodiments of this application provide a computer device, including: The system includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the channel fingerprinting transmission method described in the first or second aspect.

[0011] The channel fingerprint transmission method proposed in this application generates an approximate posterior of the channel fingerprint by introducing a pre-trained coding network and generates a compressed channel fingerprint through sampling. The pre-designed coding network can compress the channel fingerprint, reducing the resources consumed in transmission and storage. Furthermore, it enhances the privacy protection of users of location services by avoiding the transmission of precise location information.

[0012] Another channel fingerprint transmission method, channel fingerprint transmission device, and computer equipment proposed in this application embodiment also have the above-mentioned technical effects. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0014] Figure 1 This is a step diagram of the channel fingerprint transmission method provided in the embodiments of this application; Figure 2 An example of a variational autoencoder provided in this application embodiment; Figure 3 An explanatory diagram of the encoding and decoding network framework provided in the embodiments of this application; Figure 4 This is another flowchart example of the channel fingerprint transmission method provided in the embodiments of this application.

[0015] Figure 5 This is a flowchart illustrating the steps of another channel fingerprint transmission method provided in an embodiment of this application.

[0016] Figure 6 This is a structural diagram of the channel fingerprint transmission device provided in the embodiments of this application; Figure 7 A structural diagram of another channel fingerprint transmission device provided in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] Currently, the demand for location-based services is increasing, and various emerging applications are placing more stringent requirements on wireless positioning. For example, augmented reality (AR) may require centimeter-level accuracy to locate the position and angle of objects; applications such as robotics, digital twins, and smart cities require decimeter-level positioning accuracy; and traffic monitoring requires meter-level positioning accuracy.

[0019] Currently, Massive MIMO technology is widely used. Supported by massive MIMO antenna arrays, Massive MIMO base station systems can provide fine spatial and angular resolution. Based on these communication facilities and combined with techniques such as narrower directional beamforming, current communication systems already possess the foundation for achieving high-precision positioning capabilities. It is foreseeable that in existing communication networks, such as 5G networks, and in future communication networks, such as sixth-generation (6G) wireless networks, communication systems will play a crucial role in providing high-precision positioning capabilities.

[0020] Integrated Communication and Positioning (ICAP) technology can reuse basic communication infrastructure to achieve various functions, enabling seamless integration of positioning, sensing, and communication. A key application of ICAP technology is precise positioning based on channel fingerprints. For communication systems based on User Terminals (UTs) and base stations, channel fingerprints can include channel characteristics reflecting the wireless environment, such as channel path power, angle of arrival (AoA), and multipath delay. These wireless channel parameters are influenced by the location of the UT and its surrounding environment, providing a detailed characteristic description of a specific location within the target area.

[0021] In wireless communication, base stations continuously monitor the Channel State Information (CSI) of user terminals to achieve high-speed data transmission. This process essentially captures a large amount of CSI data related to the environment and the location of the user terminal. Based on CSI data from known locations as channel fingerprints describing the location, base stations can achieve high-precision positioning without additional infrastructure, providing spatial awareness and enhanced communication capabilities. Furthermore, by mapping known user terminal locations to their corresponding channel fingerprints, Massive MIMO systems can enhance wireless communication and optimize transmission. Applying the concept of channel fingerprints, CSI information can generate dual value (i.e., simultaneously used for communication and positioning tasks), becoming a key advantage of ICAP in Massive MIMO systems.

[0022] However, current channel fingerprint-based positioning technology still faces several key challenges. One major challenge lies in the need to pre-build a channel fingerprint database of known locations within the target area. This database stores a large amount of location and fingerprint data, resulting in significant storage overhead. Determining a reasonable update strategy to ensure the accuracy of the database is also a problem that needs to be solved. Another challenge is that during communication, user terminals need to report their location to the base station to generate corresponding statistical CSI. However, this exchange of precise location data is prone to privacy breaches.

[0023] The first embodiment of this application provides a channel fingerprinting transmission method to solve the above-mentioned problems. The method provided in this embodiment is applied to a terminal. Figure 1 This is a flowchart of the channel fingerprint transmission method according to an embodiment of this application, as follows: Figure 1 As shown, the process includes the following steps: Step 110: Obtain the channel information of the terminal's location and generate a channel fingerprint.

[0024] In this embodiment, the terminal receives the original communication signal at its location and can directly use the original communication signal as channel information; or, based on the data format requirements of the communication protocol, it can perform basic processing such as data format conversion on the original communication signal to obtain channel information that meets the format requirements. The channel information can reflect the channel characteristics of the communication environment at the terminal's location. For example, the channel information may include data such as multipath delay (ToA), angle of arrival (AoA), and angle of departure (AoD) at the terminal's location.

[0025] Then, the channel information is processed to generate a corresponding channel fingerprint. Specifically, the processing methods may include integrating and constructing the channel information based on the data structure of the channel fingerprint, or further analyzing the channel information. The data structure of the channel fingerprint can be set according to actual needs, and this embodiment does not impose specific limitations. For example, according to transmission requirements, the structure of the channel fingerprint can be set as a one-dimensional vector or matrix; further analysis of the channel information may include generating an information-rich CSI fingerprint by using signal strength statistics, power statistics, or fine-grained estimation of the wireless channel frequency response, based on the receiver's requirements for data type.

[0026] Step 120: Determine the approximate posterior of the channel fingerprint through a pre-trained coding network and perform deterministic sampling on the approximate posterior to obtain the compressed channel fingerprint.

[0027] In this embodiment, an approximate a posteriori of the channel fingerprint is determined through a pre-trained coding network to achieve compression and encryption of the channel fingerprint.

[0028] In traditional communication technologies, such problems are typically addressed using deterministic mapping. That is, a deterministic mapping, such as a pre-designed encryption function, is used to encrypt the channel fingerprint.

[0029] However, the above-mentioned technical approach is not suitable for the technical problem to be solved in this application. The reason is that wireless channels themselves are time-varying, random, and environmentally dependent. Even at the same physical location, channel fingerprints obtained from multiple measurements will differ due to factors such as human movement, vehicle passage, and weather changes. This difference is not noise, but rather the legitimate range of variation inherent in the channel itself. However, deterministic mapping assumes that the channel fingerprint observed at the same location should be unique, ignoring the inherent diversity of the channel. This results in only an "average" or "most common" channel state being output after reconstruction, losing the multiple reasonable forms that the real channel may possess. If modifications and optimizations are made based on deterministic mapping, at least two or more deterministic mapping results must be stored to reflect the idea that "the channel has a range of variation," further increasing resource overhead.

[0030] Therefore, this embodiment proposes to solve the above problem using a posterior distribution. Specifically, the following definitions are given: Channel fingerprint, mathematically speaking, can be understood as a high-dimensional, observable random vector.

[0031] Compressed channel fingerprints, mathematically speaking, can be understood as representing AND... Associated low-dimensional continuous latent variables. When applied to matrix-form channel fingerprints, core, low-rank structural information (such as the angle / delay components of the dominant scattering clusters, relative position information, etc.) in the channel fingerprint can be encapsulated. The compact representation to be learned and transmitted, denoted as channel fingerprint. Dimensions for Then there is That is, the dimension of the compressed channel fingerprint is much smaller than the dimension of the initial channel fingerprint.

[0032] : The prior distribution. Its form can be set according to physical priors, for example, it can be set to a standard Gaussian distribution. Or Gaussian mixtures related to known location distribution or propagation scenarios.

[0033] Conditional distribution, describing a given... Time generation The probability distribution. In this embodiment, it can be implemented using a decoder neural network. This conditional distribution can be set to a Gaussian distribution, i.e. in, For variance, It is the identity matrix. This is a decoder neural network, and the output of the decoder neural network is the reconstructed mean. .

[0034] True posterior represents the actual posterior after observation. Later about The conditional distribution.

[0035] Approximate posterior (variational / encoder). Used to approximate the true posterior. Its form can be a diagonal covariance Gaussian: in, This represents the construction of a diagonal covariance matrix.

[0036] In this embodiment, the above-mentioned approximate posterior can be achieved through an encoder neural network, and the mean of the encoder network output is the mean value. and log-variance .

[0037] Based on the above definition, according to Bayes' theorem, we have: In practical applications, the ideal situation is to obtain observations. Later about True posterior distribution However, since the denominator p(x) needs to be adjusted... and exist Integrating the joint distribution over the given area is not only complex but may also involve multiple layers of nested operations, making it practically impossible to implement. Therefore, the true posterior is often difficult to compute analytically. As a fallback option, we can obtain an approximate posterior through an encoding network. .

[0038] By rationally designing the architecture of the coding network and continuously optimizing and adjusting its parameters during training, it is possible to obtain a compressed channel fingerprint using an approximate posterior distribution. The dimension is significantly smaller than that of the original channel fingerprint. Or rather The processing and storage resources consumed are significantly less than .

[0039] The use of an encoding network to generate an approximate posterior in this embodiment has the following advantages: 1. Reflecting observation noise and non-uniqueness: The physical propagation process contains measurement noise, short-time fading, and many unobservable random factors in the environment. (Similar or related) It does not necessarily correspond to a unique one. —From the perspective of generative models, multiple different They may all generate the same statistically. Therefore, the approximate posterior, as a distribution with a certain variance, can reflect this characteristic.

[0040] 2. Demonstrates multimodality: If the conditional distribution or prior distribution If multimodality exists, such as when multiple possible scattering configurations in the propagation environment lead to similar power distributions, then the approximate posterior will naturally also exhibit multimodality, a capability that deterministic mappings do not possess.

[0041] 3. Solving the problem of non-analytical normalization constants: By Bayes' theorem, It needs to be based on the normalization factor, i.e. Solve this problem. This integral typically lacks a closed-form solution in high-dimensional or complex generative models, making it impossible to directly derive its explicit form. However, it can be obtained through learning by the encoding network. The approximate distribution parameters are used to determine an approximate posterior, from which a specific posterior is obtained by sampling. It can be used as a compressed channel fingerprint.

[0042] This embodiment preserves the randomness of the channel fingerprint through an approximate posterior. However, the approximate posterior is a distribution and cannot be used as a compressed channel fingerprint.

[0043] Suppose there exists an approximate posterior, which has the following form: Since z is a random variable whose value depends on random noise, the gradient cannot be directly backpropagated to the coding network parameters through random sampling (because sampling is a nondeterministic and non-differentiable operation). Therefore, it is necessary to introduce noise to specifically sample a value from the approximate posterior as a compressed channel fingerprint.

[0044] This embodiment proposes that, when the channel fingerprinting transmission method is applied to the terminal and the coding network performs inference, deterministic sampling is performed on the approximate posterior. Assume a distribution... The formula for sampling it is: and For distribution The parameters, Noise introduced into the sampling process.

[0045] In deterministic sampling, Values ​​are obtained through a deterministic method rather than random values. The deterministic method can... The value is always set to a fixed value, such as zero, or determined by calculation using a deterministic mapping. The value of noise. Reasonable design. The determination method can obtain a compressed channel fingerprint that accurately reflects the characteristics of the original channel fingerprint.

[0046] If the approximate distribution covers the various forms that a channel fingerprint may exist, then deterministic sampling is the process of selecting the most representative compressed channel fingerprint from the approximate distribution that has the core features of the original channel fingerprint.

[0047] Through the above steps, a compressed channel fingerprint can be obtained based on the original channel fingerprint. The processing and storage of the compressed channel fingerprint consumes significantly fewer resources than the original channel fingerprint and is encrypted. Without a decoding network corresponding to the encoding network, it is impossible to restore the compressed channel fingerprint to the original channel fingerprint, thus effectively protecting user privacy.

[0048] Step 130: Send the compressed channel fingerprint to the server so that the server can process the compressed channel fingerprint.

[0049] After obtaining the compressed channel fingerprint, it is sent from the terminal to the server. The server can process the compressed channel fingerprint according to specific needs to achieve various functions. For example, if it is necessary to determine the location of the terminal, a channel fingerprint database storing compressed fingerprint data and the corresponding locations can be pre-established, and the terminal location can be determined by directly comparing the compressed fingerprint data. If it is necessary to obtain specific information about the terminal's location channel, the original channel fingerprint can be obtained by restoring the compressed channel fingerprint through the decoding network corresponding to the encoding network.

[0050] The second embodiment of this application further specifies the channel fingerprint transmission method in the first embodiment in a more detailed and specific way. Some or all of the technical features in the second embodiment can be combined with or replaced by the first embodiment, either individually or in combination, to obtain more feasible channel fingerprint transmission methods.

[0051] The channel fingerprinting transmission method in the second embodiment of this application is described in detail below: Optionally, obtaining channel information of the terminal's location and generating a channel fingerprint includes: obtaining channel information of the terminal's location and constructing a basic channel matrix; performing an angle-domain to time-delay-domain transformation on the basic channel matrix to obtain a beam-domain channel matrix; extracting the long-term statistical feature distribution of the beam-domain channel matrix to obtain a refined beam-domain channel power matrix as a channel fingerprint.

[0052] This embodiment proposes using a matrix form, specifically a refined beam domain channel power matrix, as the original channel fingerprint.

[0053] Specifically, the channel information of the terminal's location is first collected, such as multipath delay, angle of arrival, time of arrival and departure angle, and then assembled according to requirements to construct a basic channel matrix.

[0054] In one alternative embodiment, the power, angle of arrival, and time of arrival of each scattering path are assembled to construct a basic channel matrix.

[0055] Then, the basic channel matrix is ​​transformed from the angle domain to the time delay domain to obtain the beam domain channel matrix. The specific method of the angle domain to time delay domain transformation can be the Discrete Fourier Transform or a more refined oversampling angle domain transformation, etc. This embodiment does not make specific limitations.

[0056] Let the refined beam domain channel power matrix be... Let the beam domain channel matrix be... Therefore, the calculation method for the refined beam domain channel power matrix can be expressed as: In the above formula, E{ } represents the operation of calculating mathematical expectation. ( ) is the Kronecker function, and the matrix The subscripts i and j represent the positions of the elements in the refined beam domain channel power matrix.

[0057] Beam domain channel matrix The subscripts k and t represent the position of the element in the beam domain channel matrix, which is essentially still part of the complete CSI channel fingerprint, where each element The complex channel coefficients between the k-th received beam of the base station and the t-th delay unit of the arrival path of the user terminal. Beam-domain channel matrix. It can reveal the sparse structure of physical propagation paths in the angular and time-delay dimensions.

[0058] In the beam domain channel matrix Based on this, by analyzing the beam domain channel matrix at different times... Perform a statistical average to calculate the mathematical expectation, and then... The long-term statistical feature distribution is extracted to form a more stable feature matrix that is more suitable for positioning, namely the refined beam domain channel power matrix. (Fine-grained Beam-domain Channel PowerMatrix), corresponding to the channel fingerprint in the aforementioned embodiments. (Channel Fingerprint).

[0059] The elements are independent. Each component is directly related to the propagation environment and the characteristics of the scatterer. This characteristic is further verified by the principle of channel sparsity, namely... Most elements in the cluster are approximately zero, and its non-zero cluster structure corresponds to the distribution of a specific scatterer in physical space. It is a statistically stable characteristic that describes the energy distribution pattern of multipath signals in the angular and delay dimensions. It directly reflects the user's geometric position relative to the base station, the propagation path structure, and the surrounding scattering environment, and can characterize the distribution characteristics of channel power in the beam direction and delay dimensions. Therefore, It contains highly location-related structural information and has strong temporal stability, making it an important foundation for building wireless positioning and environmental awareness.

[0060] Compared to traditional RSS channel fingerprinting, As a channel fingerprint, it has the following advantages: Higher dimensionality and richer information: RSS channel fingerprints only represent a single scalar feature of received power, while Representing multipath energy distribution in multidimensional spaces such as angle and delay can provide environmental fine-grained structural information far exceeding that of RSS.

[0061] More environmentally dependent and more discriminative: RSS channel fingerprints often change gradually in adjacent positions, making it difficult to provide sufficient discriminative power; while It is highly sensitive to user location and scatterer layout, and the fingerprint spectrum shows significant differences in different locations, resulting in a significant improvement in positioning resolution.

[0062] More stable and less susceptible to temporary obstruction: RSS channel fingerprints are significantly affected by temporary obstruction and short-term changes; while It is the beam domain channel matrix Its long-term statistical characteristics reflect a relatively stable propagation structure and are not sensitive to short-term rapid decay, making it more suitable as input for database construction and model training.

[0063] Directly and tightly integrated with the physical layer: RSS channel fingerprints are often used as higher-layer information, while Derived from physical layer CSI, it can naturally support key tasks such as communication-sensing integration (ICAP), precoding design, and statistical CSI generation. This invention utilizes... Its physical interpretability and sparsity enable it to serve simultaneous tasks of localization, environment reconstruction, and communication enhancement.

[0064] In summary, the beam domain channel matrix As a high-dimensional beam-domain CSI channel fingerprint, it contains complete space-frequency complex coefficient information, while the refined beam-domain channel power matrix... As its statistical power spectrum, stable structural features strongly correlated with user terminal location are further extracted. The relationship between the two lays the theoretical foundation for the channel fingerprinting transmission method provided in this embodiment, enabling... Channel fingerprinting, as its core, plays a crucial role in wireless positioning, digital twinning, and communication enhancement tasks. The encoder-decoder structure proposed in this invention is used to... Performing low-dimensional compression and high-fidelity reconstruction can not only significantly reduce the storage overhead of fingerprint databases, but also provide more reliable and robust statistical support.

[0065] Optionally, obtaining a compressed channel fingerprint by determining the approximate posterior of the channel fingerprint through a pre-trained coding network and performing deterministic sampling includes: inputting the channel fingerprint into the pre-trained coding network to determine the approximate posterior of the channel fingerprint; setting the noise in the reparameterization to zero to perform deterministic sampling on the approximate posterior, and using the result of the deterministic sampling as the compressed channel fingerprint.

[0066] This embodiment specifically defines the implementation method of deterministic sampling.

[0067] Specifically, this embodiment proposes sampling based on reparameterization techniques, where reparameterization can be expressed as: To compress channel fingerprints, and The mean and log-variance of the encoding network output with respect to the approximate posterior. This represents element-wise multiplication. It is noise.

[0068] In scenarios such as channel fingerprint compression in terminal devices, the desired outcomes are typically: reproducible results (reducing randomness), low computational overhead (avoiding the generation of random numbers), and stable compressed representation (facilitating matching or transmission). Therefore, this embodiment proposes setting random noise to zero to reduce randomness, thus transforming the deterministic sampling of the compressed channel fingerprint into: That is, compressed channel fingerprints directly use approximate posterior mean. Skip random perturbations.

[0069] The reason for proposing this deterministic sampling method in this embodiment is that in a well-trained encoder-decoder network structure, information is already concentrated in the mean, that is, the mean μ has already encoded the main semantic information of the input x. Furthermore, σ is usually small (especially in high-dimensional data), and random perturbations have a limited impact on reconstruction. This deterministic sampling method can obtain the most representative samples while reducing randomness.

[0070] Optionally, a pre-trained encoding network and a pre-trained decoding network located on the server constitute a variational autoencoder; the pre-trained encoding network includes at least one convolutional layer and at least one fully connected layer, wherein at least one convolutional layer in the encoding network is activated using a nonlinear non-saturating activation function, and the output of the last convolutional layer is flattened and used as the input of the first fully connected layer; the pre-trained decoding network includes at least one fully connected layer and at least one transposed convolutional layer, wherein the output of the last fully connected layer in the decoding network is reshaped and used as the input of the first transposed convolutional layer, and at least one transposed convolutional layer is activated using a nonlinear non-saturating activation function or a nonlinear saturating activation function.

[0071] This embodiment defines the structure of the coding network.

[0072] Specifically, a variational autoencoder consists of a pre-trained encoder network and a pre-trained decoder network located on the server. Figure 2 This is an example of a variational autoencoder implemented based on the neural network structure described in this embodiment.

[0073] Figure 2 In the example shown, the encoder network includes two convolutional layers. The input to the first convolutional layer is a refined beam-domain channel power matrix. : Where H represents a matrix The number of rows, W represents the matrix. The number of columns.

[0074] like Figure 2 In the example shown, the processing of the first convolutional layer is as follows: In one optional embodiment, the parameters of the convolutional layer are: a 3*3 kernel; a stride of 2; a number of output channels of c1; and an activation function of ReLU (Recurrent Unsaturated Activation Function).

[0075] The processing of the second convolutional layer is as follows: In one optional embodiment, the parameters of the convolutional layer are: a 3*3 kernel; a stride of 2; a number of output channels of c2; and an activation function of ReLU (a non-linear, non-saturating activation function).

[0076] Figure 2 In the example shown, the convolutional and fully connected layers in the encoder network are flattened using Flatten, allowing... After flattening, we get a vector. .

[0077] The vector h is processed sequentially through three fully connected (Dense) layers: In one alternative embodiment, the number of neurons in the fully connected layers of the coding network is set to 512, 128, and 16, respectively.

[0078] The final encoding network outputs approximate posterior mean and log-variance: in, Represents the fully connected layer d3. The dimension represents the approximate posterior.

[0079] Figure 2 In the example shown, the approximate posterior mean and log-variance are sampled using a reparameterization technique to generate samples and input into the decoding network. The decoding network consists of two fully connected layers and is processed as follows: Combined with the reshape operation, z is mapped to a dimension lower than 1. tensor Among them This represents the number of channels in the first fully connected layer of the decoder.

[0080] In one alternative embodiment, the number of neurons in the fully connected layers of the decoding network is set to 16 and 32, respectively.

[0081] Figure 2 In the example shown, the fully connected layer of the decoder is followed by three transposed convolutional layers. Each transposed convolutional layer processes data as follows: First transposed convolutional layer: In one optional embodiment, the parameters of the transposed convolutional layer are: a 3*3 kernel; a stride of 2; a number of output channels of c2; and an activation function of ReLU (a non-linear, non-saturating activation function).

[0082] Second transposed convolutional layer: In one optional embodiment, the parameters of the transposed convolutional layer are: a 3*3 kernel; a stride of 2; a number of output channels of c1; and an activation function of ReLU (a non-linear, non-saturating activation function).

[0083] Output of the third transposed convolutional layer: in The reconstructed channel fingerprint, provided that the accuracy requirements are met, can be considered as... equal This means that the channel fingerprint is completely restored. The third layer can use either the nonlinear non-saturating activation function ReLU or the nonlinear saturating activation function Sigmoid, depending on the input normalization method.

[0084] Figure 3 The diagram illustrates the framework of the encoding and decoding networks. The parameters of the encoding and decoding networks are learned together. Represents compressed channel fingerprint (latent variable). This represents the original channel fingerprint. To encode network parameters, This is for decoding network parameters. For the prior approximate distribution, It is a posterior approximate distribution (encoding network). As a prior distribution, For the likelihood function (decoding network). This represents the marginal likelihood.

[0085] The number of the aforementioned convolutional layers, transposed convolutional layers, and fully connected layers can be set according to the accuracy requirements for approximate posterior and the accuracy requirements for compressed fingerprint reconstruction, etc., and this embodiment does not impose specific limitations.

[0086] In one alternative embodiment, the Adam optimizer is used during training with a learning rate of Batch size is 32 or 64; latent space dimension is 16, 32, or 64; training epochs are set to 200 to 500; and logarithmic compression can be used. Normalize the input.

[0087] Optionally, during the training phase of the pre-trained encoding network and the pre-trained decoding network located on the server, at least one training is performed as follows: training samples are input into the encoding network to determine the approximate posterior of the training samples; based on noise randomly sampled from a standard Gaussian distribution, the approximate posterior of the training samples is randomly sampled through reparameterization, and the result of the random sampling is used as the training compressed channel fingerprint; the training compressed channel fingerprint is reconstructed through the decoding network to obtain the reconstructed signal corresponding to the training samples; based on the training samples and the reconstructed signal, the total loss is calculated based on a preset loss function, and the network parameters of the encoding network and the decoding network are updated based on the total loss.

[0088] This embodiment limits the training method for the encoding and decoding networks. The encoding and decoding networks are obtained through joint training.

[0089] In this embodiment, the training samples are first input into the encoding network to determine the approximate posterior of the training samples.

[0090] Then, based on noise randomly sampled from a standard Gaussian distribution, the approximate posterior of the training samples is randomly sampled through reparameterization, and the result of the random sampling is used as the training compressed channel fingerprint. The specific implementation of reparameterization can be found in the description of the foregoing embodiments. This embodiment proposes that, in order to introduce more randomness during training, deterministic sampling is not used; instead, noise randomly sampled from a standard Gaussian distribution is used. The random sampling process is as follows: Then, the training compressed channel fingerprint is reconstructed using the decoding network to obtain the reconstructed signal corresponding to the training sample. Finally, based on the training sample and the reconstructed signal, the total loss is calculated using a preset loss function, and the network parameters of the encoding and decoding networks are updated based on the total loss.

[0091] In practical applications, the above steps can be repeated multiple times until the model converges, the preset number of training iterations is reached, or the accuracy meets the requirements.

[0092] Optionally, the preset loss function is the sum of the first loss element and the second loss element; the first loss element is the product of the first weight coefficient and the reconstruction loss, and the reconstruction loss measures the difference between the restored signal output by the decoding network and the training samples; the second loss element is the product of the second weight coefficient and the regularization loss, and the regularization loss measures the difference between the approximate posterior distribution output by the encoding network and the standard prior distribution; the first weight coefficient and the second weight coefficient are hyperparameters.

[0093] This embodiment further defines the loss function used during training.

[0094] In one embodiment, a refined beam domain channel power matrix is ​​used. As a channel fingerprint Given the specific form, the conditional distribution in the aforementioned embodiments becomes... The true posterior becomes Approximate posterior becomes Therefore, the loss function can be expressed as: The loss function in this embodiment is obtained by modifying the ELBO loss of the variational autoencoder. The ELBO loss of the variational autoencoder is obtained as follows: To quantify the similarity between two distributions, calculate and The Kullback-Leibler (KL) divergence between them. It should be noted that... This corresponds to the encoder in a neural network. According to the definition of KL divergence, it can be expanded as: Define the integral term as This quantity is called the "variable lower bound". By rearranging the terms, we can obtain the following expression: because With variables Irrelevant, the left side of the equation is relative to It is a constant. Considering the non-negativity of the KL divergence, minimize Equivalent to maximizing Therefore, the problem is transformed into finding the variational lower bound. The optimization problem.

[0095] For in-depth analysis, focus on parameters and Optimization to maximize The variational lower bound can be expanded as follows: This decomposition form is for optimizing encoder parameters. and decoder It provides clear direction: First item : Represents latent variables Generate channel fingerprints The contribution of this term. By maximizing this term, the accuracy and consistency of the channel fingerprint generated by the decoder can be improved. The second term... : Represents an approximate distribution With the prior distribution of latent variables The difference between them. By minimizing this term, we can ensure that the distribution of latent variables generated by the encoder is consistent with the prior distribution, thereby improving the expressiveness and generalization of the overall model.

[0096] To solve for the variational lower bound The optimal conditions require calculating about The gradient is obtained as follows: in Expanding the gradient, we get: Since directly processing this expression is quite complex, this embodiment proposes a solution using a neural network. Assumptions: 1. The probability density function of the prior distribution Standard Gaussian distribution 2. Approximate posterior distribution A Gaussian distribution with a diagonal covariance matrix The approximate posterior mean Covariance The encoder is trained with positional fingerprints. The output is the input. Let it be the output. The dimension is Then all training data The KL divergence can be expressed as: For the second term, the reconstruction loss term, a Gaussian distribution is used to model the decoding process. Specifically, the conditional probability density function is assumed to be... obey Distribution, among which Indicates the decoder output. Let be a constant covariance matrix. Based on the Markov chain Monte Carlo (MCMC) method, we can obtain: In the formula From the mean Covariance is The first generation generated from the Gaussian distribution One sample. In practical applications, usually one sample is taken. Generate a single sample. Therefore, for all For each training data set, maximizing the log-likelihood expectation simplifies to: In the formula, Indicates the decoder output. Input to the encoder. This is achieved by comparing the input position fingerprint. Its decoder reconstructed output The mean squared error (MSE) between the input and output is used to measure reconstruction accuracy. This MSE loss function reflects the decoder's ability to capture the true structure of the input. By minimizing the MSE, the decoder parameters are optimized. This ensures the reconstructed location fingerprint. As close as possible to the original input .

[0097] In addition, to improve the adaptability of the variational codec framework, hyperparameter weight coefficients are introduced into the loss function. and This is used to balance the contributions of reconstruction loss and regularization loss, thereby achieving precise control over the learning process. Through this mechanism, the model can flexibly adjust the constraint strength on latent variables—typically choosing... This can enhance latent space regularization, prompting the model to learn meaningful representations. The hyperparameters α and β can be dynamically adjusted during training to strike a tradeoff between compression ratio and reconstruction quality.

[0098] In one alternative embodiment, if stronger privacy requirements are needed, it can be achieved by... Adding noise to the middle will improve This can reduce the risk of inferring hidden locations, but this comes at the cost of reconstruction accuracy; therefore, a special privacy / information leakage regularization term can be introduced into the loss function to balance this.

[0099] like Figure 4 The figure shows a summary of the channel fingerprint transmission method provided in the foregoing embodiments. Figure 4 The left side represents the channel fingerprint. Dao generates approximate posterior through an encoding network. The process. Figure 4 Intermediate representation by introducing Combined with the encoder output and The process of generating compressed channel fingerprints based on reparameterized sampling.Figure 4 The right side shows the generated compressed channel fingerprint input to the decoder and then restored to... The process. Figure 4 The following illustrates the process of establishing a loss function by introducing hyperparameters α and β and maximizing the variational lower bound ELBO.

[0100] The third embodiment of this application provides a channel fingerprint transmission method, applied to a server. Figure 5 This is a flowchart of the channel fingerprint transmission method according to an embodiment of this application, as follows: Figure 5 As shown, the process includes the following steps: Step 510: Receive the compressed channel fingerprint sent by the terminal. The compressed channel fingerprint is generated by the terminal based on the channel information of its location. The terminal then uses a pre-trained coding network to determine the approximate posterior of the channel fingerprint and performs deterministic sampling on the approximate posterior to generate the compressed channel fingerprint.

[0101] Step 520: Perform at least one of the following processes based on the compressed channel fingerprint: reconstruct the compressed channel fingerprint using a pre-trained decoding network to obtain the channel fingerprint; search for the record corresponding to the compressed channel fingerprint in a pre-established channel fingerprint database, and determine the location of the terminal based on the record.

[0102] The channel fingerprint transmission method in this embodiment proposes that after the server receives the compressed channel fingerprint sent by the terminal, it can process the compressed channel fingerprint according to specific needs to achieve various functions. Among these functions, some do not require the compressed channel fingerprint to be restored.

[0103] For example, if it is necessary to determine the location of a terminal, a channel fingerprint database that stores compressed fingerprint data and the location corresponding to the compressed fingerprint data can be established in advance. The location of the terminal can be determined by directly comparing the compressed fingerprint data. No reconstruction calculation is performed on the server side during the whole process, which saves communication system resources.

[0104] In addition, if it is necessary to obtain specific information about the terminal location channel, then a restoration operation must be performed. The compressed channel fingerprint is restored by the decoding network corresponding to the encoding network to obtain the original channel fingerprint.

[0105] The fourth embodiment of this application also proposes a channel fingerprinting transmission device, such as... Figure 6 As shown, the device includes: The channel fingerprint generation unit 610 is used to obtain channel information of the terminal's location and generate a channel fingerprint; The posterior sampling unit 620 is used to determine the approximate posterior of the channel fingerprint through a pre-trained coding network and perform deterministic sampling on the approximate posterior to obtain a compressed channel fingerprint. The first transmission unit 630 is used to send the compressed channel fingerprint to the server so that the server can process the compressed channel fingerprint.

[0106] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0107] In this embodiment, the channel fingerprint transmission device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0108] The fifth embodiment of this application also proposes a channel fingerprinting transmission device, such as... Figure 7 As shown, the device includes: The second information transmission unit 710 is used to receive the compressed channel fingerprint sent by the terminal. The compressed channel fingerprint is generated by the terminal based on the channel information of its location, and the approximate posterior of the channel fingerprint is determined by a pre-trained coding network, and then the approximate posterior is deterministically sampled to generate the compressed channel fingerprint. The data processing unit 720 is configured to perform at least one of the following processes based on the compressed channel fingerprint: reconstructing the compressed channel fingerprint using a pre-trained decoding network to obtain the channel fingerprint; searching for a record corresponding to the compressed channel fingerprint in a pre-established channel fingerprint database; and determining the location of the terminal based on the record.

[0109] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0110] In this embodiment, the channel fingerprint transmission device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0111] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 8As shown, the computer device includes one or more processors 810, memory 820, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 8 Take the 810 processor as an example.

[0112] The processor 810 may be a central processing unit, a network processor, or a combination thereof. The processor 810 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GPA), or any combination thereof.

[0113] The memory 820 stores instructions executable by at least one processor 810 to cause the at least one processor 810 to perform the method shown in the above embodiments.

[0114] The memory 820 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 820 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 820 may optionally include memory remotely located relative to the processor 810, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0115] The memory 820 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 820 may also include a combination of the above types of memory.

[0116] The computer device also includes a communication interface 830 for communicating with other devices or communication networks.

[0117] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0118] This application also provides a computer program product including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method of any embodiment of this application.

[0119] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

[0120] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0121] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0122] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0123] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0124] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0125] It is understood that in the specific implementation of this application, data such as user information, location information, and navigation data are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0126] The apparatus, computer device, computer storage medium, and computer program product described in the above embodiments can be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0127] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0128] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, computer devices, computer storage media, and computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, computer devices, computer storage media, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0132] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0133] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, computer equipment, computer storage media, and computer program products are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0134] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0135] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A channel fingerprinting transmission method, characterized in that, Applied to a terminal, the method includes: Obtain the channel information of the terminal's location and generate a channel fingerprint; The approximate posterior of the channel fingerprint is determined by a pre-trained coding network, and the approximate posterior is deterministically sampled to obtain a compressed channel fingerprint. The compressed channel fingerprint is sent to the server for processing.

2. The method according to claim 1, characterized in that, The step of obtaining the channel information of the terminal's location and generating a channel fingerprint includes: Obtain the channel information of the terminal's location and construct a basic channel matrix; An angle-delay-domain transformation is performed on the basic channel matrix to obtain the beam-domain channel matrix; The long-term statistical feature distribution of the beam domain channel matrix is ​​extracted to obtain a refined beam domain channel power matrix as the channel fingerprint.

3. The method according to claim 1, characterized in that, The step of determining the approximate posterior of the channel fingerprint through a pre-trained coding network and performing deterministic sampling on the approximate posterior to obtain a compressed channel fingerprint includes: The channel fingerprint is input into the pre-trained coding network to determine the approximate posterior of the channel fingerprint; The noise in the reparameterization is set to zero to perform deterministic sampling of the approximate posterior, and the result of the deterministic sampling is used as the compressed channel fingerprint.

4. The method according to claim 1, characterized in that, The pre-trained encoding network and the pre-trained decoding network located on the server constitute a variational autoencoder; The pre-trained encoding network includes at least one convolutional layer and at least one fully connected layer. In the encoding network, the at least one convolutional layer is activated using a non-linear, non-saturating activation function, and the output of the last convolutional layer is flattened and used as the input of the first fully connected layer. The pre-trained decoding network includes at least one fully connected layer and at least one transposed convolutional layer. In the decoding network, the output of the last fully connected layer is reshaped and used as the input of the first transposed convolutional layer. The at least one transposed convolutional layer is activated using a non-linear non-saturated activation function or a non-linear saturated activation function.

5. The method according to claim 1, characterized in that, During the training phase of the pre-trained encoding network and the pre-trained decoding network located on the server, training is performed at least once in the following manner: The training samples are input into the encoding network to determine the approximate posterior of the training samples; Based on noise randomly sampled from a standard Gaussian distribution, the approximate posterior of the training samples is randomly sampled through reparameterization, and the result of the random sampling is used as the training compressed channel fingerprint. The training compressed channel fingerprint is restored by a decoding network to obtain the restored signal corresponding to the training sample; Based on the training samples and the restored signal, the total loss is calculated using a preset loss function, and the network parameters of the encoding network and the decoding network are updated based on the total loss.

6. The method according to claim 5, characterized in that, The preset loss function is the sum of the first loss element and the second loss element; The first loss element is the product of the first weight coefficient and the reconstruction loss, wherein the reconstruction loss measures the difference between the restored signal output by the decoding network and the training samples; The second loss element is the product of the second weight coefficient and the regularization loss, which measures the difference between the approximate posterior distribution of the output of the coding network and the standard prior distribution. The first weighting coefficient and the second weighting coefficient are hyperparameters.

7. A channel fingerprinting transmission method, characterized in that, Applied to the server side, the method includes: The receiving terminal sends a compressed channel fingerprint, which is generated by the terminal based on the channel information of its location. The terminal then uses a pre-trained coding network to determine the approximate posterior of the channel fingerprint and performs deterministic sampling on the approximate posterior to generate the compressed channel fingerprint. Perform at least one of the following processes based on the compressed channel fingerprint: The compressed channel fingerprint is reconstructed using a pre-trained decoding network to obtain the channel fingerprint; The location of the terminal is determined by searching for a record corresponding to the compressed channel fingerprint in a pre-established channel fingerprint database.

8. A channel fingerprinting transmission device, characterized in that, The device includes: A channel fingerprint generation unit is used to obtain channel information of the location of the terminal and generate a channel fingerprint; A posterior sampling unit is used to determine the approximate posterior of the channel fingerprint through a pre-trained coding network and perform deterministic sampling on the approximate posterior to obtain a compressed channel fingerprint; The first transmission unit is used to send the compressed channel fingerprint to the server so that the server can process the compressed channel fingerprint.

9. A channel fingerprinting transmission device, characterized in that, The device includes: The second information transmission unit is used to receive the compressed channel fingerprint sent by the terminal. The compressed channel fingerprint is generated by the terminal based on the channel information of its location, and the approximate posterior of the channel fingerprint is determined by the pre-trained coding network, and then the approximate posterior is deterministically sampled to generate the compressed channel fingerprint. The data processing unit is configured to perform at least one of the following processes based on the compressed channel fingerprint: reconstructing the compressed channel fingerprint using a pre-trained decoding network to obtain the channel fingerprint; searching for a record corresponding to the compressed channel fingerprint in a pre-established channel fingerprint database; and determining the location of the terminal based on the record.

10. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the channel fingerprinting transmission method of any one of claims 1 to 6 or 7.

Images