Channel estimation method, apparatus, device, storage medium, and program product

Abstract

Embodiments of the present application provide a channel estimation method, device, equipment, storage medium and program product, relating to the technical field of wireless communication. The method comprises: determining energy spectrum data according to a received signal of an antenna array; taking the energy spectrum data as an input of a convolutional neural network model to obtain a probability heat map; the probability heat map is used to represent the probability of a scatterer peak on a grid; the convolutional neural network model is a model optimized by a focal loss function and a dice loss function; determining a direction parameter of the scatterer according to the probability heat map; and determining a channel estimation result according to the direction parameter of the scatterer. The present scheme solves the problem that in the prior art, due to the energy dispersion effect of a near-field spherical wave, the sidelobe of a strong scatterer leaks and covers a neighboring weak scatterer, resulting in missed detection or false detection, which seriously reduces the direction estimation accuracy and further causes deterioration of channel estimation performance.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, specifically to a channel estimation method, apparatus, device, storage medium, and program product. Background Technology

[0002] To achieve Tbit / s-level throughput, sixth-generation mobile communication (6G) is evolving towards "Very Large Scale Array (XL-MIMO)" and "High Frequency Bands (Millimeter Wave / Terahertz)". These two trends together lead to a fundamental shift in physical paradigms: the Rayleigh distance, which distinguishes between near and far fields, has drastically expanded to hundreds of meters. This means that the vast majority of users in 6G will be in the "near-field" of the antenna array's radiation. In the near field, electromagnetic waves exhibit "spherical wave" characteristics, rendering the "far-field plane wave" model and its core algorithms (such as OMP / SOMP) relied upon by 5G completely ineffective.

[0003] In the prior art, an electromagnetic information theory (EIT) near-field channel modeling and estimation scheme is proposed to address the above problems. However, this method relies on traditional signal processing algorithms (such as multi-peak threshold detection) to estimate the scatterer direction, which leads to a serious decrease in the direction estimation accuracy under multi-scatterer interference and low signal-to-noise ratio environments, thereby causing a sharp deterioration in channel estimation performance. Summary of the Invention

[0004] At least one embodiment of this application provides a channel estimation method, apparatus, device, storage medium, and program product to solve the problem in the prior art where, due to the energy dispersion effect of near-field spherical waves, the side lobes of strong scatterers leak and mask nearby weak scatterers, leading to missed detections or false detections, severely reducing the accuracy of direction estimation, and thus causing a deterioration in channel estimation performance.

[0005] To solve the above-mentioned technical problems, this application is implemented as follows:

[0006] In a first aspect, embodiments of this application provide a channel estimation method, including:

[0007] Based on the received signals from the acquired antenna array, determine the energy spectrum data;

[0008] The energy spectrum data is used as input to a convolutional neural network model to obtain a probability heatmap; the probability heatmap is used to characterize the probability of the scatterer peak on the grid; the convolutional neural network model is an optimized model using a focus loss function and a dice loss function.

[0009] The directional parameters of the scatterer are determined based on the probability heatmap.

[0010] The channel estimation result is determined based on the directional parameters of the scatterer.

[0011] Optionally, the convolutional neural network model is a U-Net neural network model;

[0012] The supervision signal for the convolutional neural network model is a hard label;

[0013] The output of the convolutional neural network model is a binary classification logical score graph.

[0014] Optionally, based on the received signal from the acquired antenna array, energy spectrum data is determined, including:

[0015] The spatial domain signal of the received signal is mapped to the wavenumber domain to generate a two-dimensional energy spectrum.

[0016] The two-dimensional energy spectrum is normalized to obtain the energy spectrum data;

[0017] The antenna array is a uniform planar array.

[0018] Optionally, determining the directional parameters of the scatterer based on the probability heatmap includes:

[0019] The pixel coordinates corresponding to the probability heatmap are obtained using an extraction algorithm;

[0020] The pixel coordinates are subjected to non-maximum suppression processing to obtain pixel coordinate data;

[0021] The pixel coordinate data is mapped to the physical wavenumber domain to obtain the direction parameters.

[0022] Optionally, the channel estimation result is determined based on the directional parameters of the scatterer, including:

[0023] The directional parameters are input into the channel function of electromagnetic information theory to obtain the reconstructed channel covariance matrix;

[0024] The channel covariance matrix is ​​processed by the minimum mean square error estimator to obtain the channel estimation result.

[0025] Optionally, the convolutional neural network model includes an encoder and a decoder;

[0026] The encoder has multiple downsampling modules, each including a max pooling layer and multiple cascaded convolutional layers, with each convolutional layer connected to a normalization layer and an activation function.

[0027] The decoder includes multiple upsampling modules, each upsampling module including multiple transposed convolutional layers and skip connections.

[0028] Secondly, embodiments of this application provide a channel estimation apparatus, comprising:

[0029] The first determining module is used to determine the energy spectrum data based on the received signal from the acquired antenna array;

[0030] The second determining module is used to take the energy spectrum data as input to the convolutional neural network model to obtain a probability heatmap; the probability heatmap is used to characterize the probability of the scatterer peak on the grid; the convolutional neural network model is a model optimized by the focus loss function and the dice loss function;

[0031] The third determining module is used to determine the orientation parameters of the scatterer based on the probability heatmap;

[0032] The fourth determining module is used to determine the channel estimation result based on the directional parameters of the scatterer.

[0033] Thirdly, embodiments of this application provide a channel estimation device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method described in the second aspect.

[0034] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that, when executed by a processor, implements the steps of the method described above.

[0035] Fifthly, embodiments of this application provide a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method described above.

[0036] Compared with existing technologies, the channel estimation method, apparatus, device, storage medium, and program product provided in this application can determine the energy spectrum data corresponding to the received signal from the acquired antenna array, and then process the energy spectrum data through a convolutional neural network model to determine the scatterer peak, ultimately obtaining a more accurate estimation result of the scatterer direction. This solves the problem in existing technologies where, due to the energy dispersion effect of near-field spherical waves, the sidelobes of strong scatterers leak and mask nearby weak scatterers, leading to missed or false detections, severely reducing the direction estimation accuracy, and consequently degrading channel estimation performance. Attached Figure Description

[0037] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0038] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application;

[0039] Figure 2 This is a flowchart illustrating the channel estimation method according to an embodiment of this application;

[0040] Figure 3 This is a schematic diagram of the channel estimation system according to an embodiment of this application;

[0041] Figure 4 This is a comparison diagram between the channel estimation method of this application embodiment and the traditional peak detection method;

[0042] Figure 5 This is a schematic diagram of the channel estimation device according to an embodiment of this application;

[0043] Figure 6 This is a schematic diagram of the structure of a channel estimation device according to an embodiment of this application. Detailed Implementation

[0044] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, without limiting the number of objects; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, "A or B" covers three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0045] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc.; an indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on the judgment result.

[0046] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used in the systems and radio technologies mentioned above, as well as in other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th Generation (6G) communication systems.

[0047] Figure 1This diagram illustrates a block diagram of a wireless communication system applicable to embodiments of this application. The wireless communication system includes a terminal 11 and a network device 12. The terminal 11 can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home (home devices with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game console, personal computer (PC), ATM, or self-service machine, etc. Wearable devices include: smartwatches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among these, in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in this application embodiment. Network device 12 may include access network devices or core network devices, wherein access network devices may also be referred to as Radio Access Network (RAN) devices, radio access network functions, or radio access network units. Access network devices may include base stations, Wireless Local Area Network (WLAN) access points (APs), or Wireless Fidelity (WiFi) nodes, etc.In this context, a base station may be referred to as a Node B (NB), an Evolved Node B (eNB), a Next Generation Node B (gNB), a New Radio Node B (NR Node B), an Access Point, a Relay Base Station (RBS), a Serving Base Station (SBS), a Base Transceiver Station (BTS), a Radio Base Station, a Radio Transceiver, a Basic Service Set (BSS), an Extended Service Set (ESS), a Home Node B (HNB), a Home Evolved Node B, a Transmission Reception Point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved. The base station is not limited to any specific technical terminology. It should be noted that in this application embodiment, only a base station in an NR system is used as an example for introduction, and the specific type of base station is not limited.

[0048] Core network equipment may include, but is not limited to, at least one of the following: core network node, core network function, Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), Binding Support Function (BSF), and Application Function. Function (AF), etc. It should be noted that the embodiments of this application only use the core network equipment in the NR system as an example for introduction, and do not limit the specific type of core network equipment.

[0049] The core foundation of 5G algorithms is "angular sparsity," meaning that channel energy exhibits sparse spikes in the Fourier domain (angular domain). However, the nonlinear phase characteristics of near-field spherical waves cause severe "energy dispersion" in the angular domain, rendering the angular sparsity assumption completely invalid.

[0050] The academic community has proposed physical models, such as a near-field channel modeling and estimation scheme based on electromagnetic information theory (EIT); this scheme consists of two core parts:

[0051] (1) EIT Near-Field Channel Modeling (Based on Existing Models) This technique is based on electromagnetic scattering theory and models the near-field channel as a non-stationary Gaussian random field. Its core feature is that the spatial correlation function of the channel... Analytically derived as a function of the scatterer's physical parameters, under the weak near-field assumption, this correlation function can be approximated as:

[0052] ;

[0053] in, For the location of the receiving antenna, The distance to the scatterer. Let be the size of the scatterer, and the phase term in this formula. And amplitude term coefficient All are highly dependent on the direction vector of the scatterer ;

[0054] In other words, if we want to use this model to reconstruct the channel covariance matrix and perform minimum mean square error (MMSE) estimation, we must first obtain high-precision scatterer-related information as prior information.

[0055] To obtain the aforementioned crucial prior direction information, existing technologies typically employ peak detection methods based on signal processing, with the following specific steps:

[0056] Step A (Transformation): Perform a two-dimensional discrete Fourier transform (DFT) on the received signal to map the spatial domain signal to the wavenumber domain, thereby obtaining the energy spectrum matrix.

[0057] Step B (Detection): A traditional multi-peak threshold detection algorithm is used to search for power peak points in the energy spectrum. A dynamic threshold (such as a multiple of the average energy) is typically set, and grid points exceeding the threshold are identified as potential scatterer peaks.

[0058] Step C (Mapping): Using the discrete grid index of the detected peak, calculate the physical direction parameters using the linear mapping formula. .

[0059] Step D (Estimation): Substitute the calculated direction parameters into the relevant function formulas mentioned above. In this process, the channel covariance matrix is ​​reconstructed, and the channel is recovered using the minimum mean square error (MMSE) estimator.

[0060] Although the aforementioned existing technologies theoretically solve the near-field modeling problem, in practical applications, this invention, through rigorous simulation reproduction, has revealed serious shortcomings:

[0061] Poor multipath resolution (core bottleneck): The traditional peak detection algorithm used in step B of the existing technology is prone to failure in multi-scatterer scenarios. Due to the energy diffusion effect of near-field spherical waves, the side lobes of strong scatterers will leak and mask the nearby weak scatterers, leading to missed detections or false detections;

[0062] Quantitative empirical evidence: Simulation results show that in scenarios with multiple scatterers (2-5), the traditional method suffers from incorrect direction estimation, resulting in a significant decrease in the normalized mean square error (NMSE) of the final channel estimation compared to the ideal case (where the direction is known).

[0063] Insufficient robustness: The fixed or linear threshold mechanism of traditional methods is difficult to adapt to low signal-to-noise ratio environments and is easily interfered with by noise, resulting in a large number of false alarms.

[0064] This application relates to an AI-assisted near-field channel estimation method based on prior information. The core idea is to reconstruct the scatterer orientation estimation problem from the traditional signal processing (peak search) into an "image-to-image translation" task. It uses a deep learning model to accurately segment the scatterer peaks from the energy spectrum, replacing the peak detection step in the traditional method, thereby providing accurate prior information for subsequent physical channel reconstruction.

[0065] As described in the background section, in the prior art, traditional signal processing algorithms (such as multi-peak threshold detection) for scatterer direction estimation result in a severe decrease in direction estimation accuracy under multi-scatterer interference and low signal-to-noise ratio environments, which in turn leads to a sharp deterioration in channel estimation performance. To solve the above problems, this application provides a channel estimation method that can reduce or avoid the occurrence of the above situations and achieve "sub-pixel level" direction estimation accuracy.

[0066] This application provides a channel estimation method and apparatus. The method and apparatus are based on the same concept, and since the principles by which they solve the problem are similar, their implementations can be referred to interchangeably; repeated details will not be repeated.

[0067] like Figure 2 As shown in the embodiment of this application, a channel estimation method includes the following steps:

[0068] Step 201: Determine the energy spectrum data based on the received signal from the acquired antenna array;

[0069] Step 202: The energy spectrum data is used as input to the convolutional neural network model to obtain a probability heatmap; the probability heatmap is used to characterize the probability of the scatterer peak on the grid; the convolutional neural network model is a model optimized by the focus loss function and the dice loss function.

[0070] Step 203: Determine the orientation parameters of the scatterer based on the probability heatmap;

[0071] Step 204: Determine the channel estimation result based on the directional parameters of the scatterer.

[0072] Optionally, the antenna array includes:

[0073] Uniform Linear Array (ULA);

[0074] Uniform Planar Array (UPA);

[0075] Uniform Circular Array (UCA);

[0076] Non-uniform array;

[0077] Sparse array.

[0078] The channel estimation method of this application embodiment can determine the energy spectrum data corresponding to the received signal of the acquired antenna array, and then process the energy spectrum data through a convolutional neural network model to determine the scatterer peak, thus obtaining a relatively accurate estimation result of the scatterer direction. This solves the problem in the prior art where, due to the energy dispersion effect of near-field spherical waves, the sidelobes of strong scatterers leak and mask nearby weak scatterers, leading to missed or false detections, severely degrading the direction estimation accuracy, and consequently causing a deterioration in channel estimation performance.

[0079] Optionally, the convolutional neural network model is a U-Net neural network model;

[0080] The supervision signal for the convolutional neural network model is a hard label;

[0081] The output of the convolutional neural network model is a binary classification logical score graph.

[0082] Optionally, the convolutional neural network model includes an encoder and a decoder;

[0083] The encoder has multiple downsampling modules, each including a max pooling layer and multiple cascaded convolutional layers, with each convolutional layer connected to a normalization layer and an activation function.

[0084] The decoder includes multiple upsampling modules, each upsampling module including multiple transposed convolutional layers and skip connections.

[0085] In this embodiment, the neural network model includes an encoder (shrinking path) and a decoder (expanding path). The encoder contains multiple downsampling modules, each consisting of a 2×2 max-pooling layer and a DoubleConv structure (i.e., two cascaded 3×3 convolutional layers, each followed by BatchNorm normalization and ReLU activation), used to extract deep semantic features. The decoder contains corresponding upsampling modules, each first upsampling using a 2×2 transposed convolution, then concatenating with high-resolution feature maps from the same layer of the encoder via skip connections, and finally fusing the features through a DoubleConv structure. The output layer maps the feature channels to two classes (foreground / background) using 1×1 convolutions, outputting a binary classification logistic score map (Logits).

[0086] The training strategy for the neural network model is as follows: the model is trained using energy spectrum data generated by simulation.

[0087] A "hard label" is used as the supervision signal (i.e., it is 1 only at the center pixel of the true peak).

[0088] To address the extreme class imbalance caused by the extremely sparse peak pixels in the energy spectrum and to improve the peak contour overlap, a combination of Focal Loss and Dice Loss is employed. As a loss function.

[0089] It's important to note that Focal Loss is introduced to address the imbalance between easy and difficult samples. Traditional cross-entropy loss functions are easily dominated by a large number of simple, easily classified background samples, preventing the model from effectively learning sparse peak features. Focal Loss addresses this by introducing a modulation factor. To reduce the weight of simple samples, the mathematical expression is:

[0090] ;

[0091] in, This represents the model's predicted probability of the true class. This is a focusing parameter used to adjust the degree of attention given to difficult-to-classify samples (such as those with weak peaks). The balancing parameter is used to adjust the weight ratio of positive and negative samples. Through this loss function, the model can automatically suppress the gradient contribution of massive background noise, focusing the training effort on hard-to-detect peak pixels.

[0092] Dice Loss is introduced to optimize the ensemble similarity between predicted regions and ground truth labels. Dice Loss is directly constructed based on the Dice coefficient, is insensitive to the size of the foreground region, and is particularly suitable for small object segmentation tasks. Its mathematical expression is defined as:

[0093] ;

[0094] Where N is the total number of pixels. and These represent the predicted probability value and the true label value (0 or 1) of the i-th pixel, respectively. To prevent smoothing terms with a denominator of zero, Dice Loss constrains the geometry of the predicted peak from a global perspective, ensuring that the predicted peak location highly coincides with the true location.

[0095] Optionally, based on the received signal from the acquired antenna array, energy spectrum data is determined, including:

[0096] The spatial domain signal of the received signal is mapped to the wavenumber domain to generate a two-dimensional energy spectrum.

[0097] The two-dimensional energy spectrum is normalized to obtain the energy spectrum data;

[0098] The antenna array is a uniform planar array.

[0099] In this embodiment of the application, the received signal of the antenna array is acquired, and Erwin discrete Fourier transform (2D-DFT) is performed on it to map the spatial domain signal to the wavenumber domain, generating a two-dimensional energy spectrum (e.g., with a size of 41×41), which is then normalized to construct the input image of the AI ​​model (the energy spectrum data).

[0100] Optionally, the convolutional neural network model is a pre-trained model. The convolutional neural network model, through its powerful feature extraction capability, suppresses sidelobe interference and outputs a high-precision probability heatmap (the foreground channel extracted after Softmax processing of the output Logits). This heatmap characterizes the probability of the scatterer peak existing on the grid.

[0101] Optionally, determining the directional parameters of the scatterer based on the probability heatmap includes:

[0102] The pixel coordinates corresponding to the probability heatmap are obtained using an extraction algorithm;

[0103] The pixel coordinates are subjected to non-maximum suppression processing to obtain pixel coordinate data;

[0104] The pixel coordinate data is mapped to the physical wavenumber domain to obtain the direction parameters.

[0105] Optionally, non-maximum suppression processing is performed on the pixel coordinates to obtain pixel coordinate data, including:

[0106] Non-maximum suppression is performed on pixel coordinates that are greater than a preset threshold to identify redundant points and obtain discrete pixel coordinate data.

[0107] Optionally, the pixel coordinate data is mapped to the physical wavenumber domain to obtain the direction parameters, including:

[0108] The pixel coordinates are converted into direction parameters in the physical wavenumber domain using a linear mapping formula.

[0109] The channel estimation method of this application improves the accuracy of the direction parameters by converting the discrete pixel coordinate data into the direction parameters through the wavenumber domain mathematical properties of the Fourier transform.

[0110] Optionally, the channel estimation result is determined based on the directional parameters of the scatterer, including:

[0111] The directional parameters are input into the channel function of electromagnetic information theory to obtain the reconstructed channel covariance matrix;

[0112] The channel covariance matrix is ​​processed by the minimum mean square error estimator to obtain the channel estimation result.

[0113] The channel estimation method of this application embodiment reconstructs the channel covariance matrix based on the direction parameters using the channel function of electromagnetic information theory, and calculates the channel estimation result using a minimum mean square error estimator, replacing the peak detection step in the traditional process, and achieving "sub-pixel level" direction estimation accuracy.

[0114] The channel estimation method in this application embodiment is as follows: Figure 3 The channel estimation system shown performs the following, specifically:

[0115] Step 1: Constructing the energy spectrum image

[0116] Obtain the observation signal vector of the received UPA antenna array A two-dimensional discrete Fourier transform (2D-DFT) is performed on the signal to map the spatial domain signal to the wavenumber domain, resulting in a two-dimensional energy spectrum matrix. This matrix is ​​then normalized and used as input to the AI ​​model.

[0117] Step 2: AI Model Prediction

[0118] The normalized energy spectrum is input into the trained U-Net model. The model outputs binary classification Logits, which are then subjected to a Softmax operation along the channel dimension to extract the foreground channels as a probability heatmap. This heatmap represents the probability of a scatterer peak at each grid point.

[0119] Step 3: Post-processing to obtain pixel coordinates:

[0120] Non-maximum suppression (NMS): Filters out candidate points above a threshold and sorts them by confidence level. For the point with the highest confidence level, it suppresses other non-maximum candidate points in its neighborhood, outputting the final list of discrete pixel coordinates (y, x).

[0121] Step 4: Coordinate Reverse Mapping

[0122] Using the properties of the Fourier transform, the pixel coordinates obtained in step 3 are converted into direction parameters in the physical wavenumber domain. The mapping formula is:

[0123] ;

[0124] Step 5: Physical Model Channel Estimation

[0125] The high-precision orientation parameters obtained in step 4 As prior information, it is substituted into the analytical correlation function of the EIT channel model to reconstruct the channel covariance matrix. Finally, the final channel estimation result is obtained by using the MMSE estimator. :

[0126] .

[0127] It should be noted that, in order to verify the accuracy of the convolutional neural network model described in this application, simulation tests were conducted in a scenario with multiple scatterers (e.g., 2-5 scatterers). Figure 4 As shown, compared with traditional peak detection-based methods, the method of the present invention exhibits significant advantages in low signal-to-noise ratio.

[0128] The various methods of the embodiments of this application have been described above. Apparatus for implementing the above methods will now be provided.

[0129] like Figure 5 As shown in the illustration, this application also provides a channel estimation device 500, comprising:

[0130] The first determining module 501 is used to determine the energy spectrum data based on the received signal of the acquired antenna array;

[0131] The second determining module 502 is used to take the energy spectrum data as input to the convolutional neural network model to obtain a probability heatmap; the probability heatmap is used to characterize the probability of the scatterer peak on the grid; the convolutional neural network model is a model optimized by the focus loss function and the dice loss function;

[0132] The third determining module 503 is used to determine the direction parameters of the scatterer based on the probability heatmap;

[0133] The fourth determining module 504 is used to determine the channel estimation result based on the direction parameters of the scatterer.

[0134] The channel estimation device of this application embodiment can determine the energy spectrum data corresponding to the received signal of the acquired antenna array, and then process the energy spectrum data through a convolutional neural network model to determine the scatterer peak value, finally obtaining a relatively accurate estimation result of the scatterer direction. This solves the problem in the prior art where, due to the energy dispersion effect of near-field spherical waves, the sidelobes of strong scatterers leak and mask nearby weak scatterers, leading to missed or false detections, severely reducing the direction estimation accuracy, and consequently degrading the channel estimation performance.

[0135] The channel estimation device in the embodiments of this application, such as Figure 6 As shown, it includes a transceiver 610, a processor 600, a memory 620, and a program or instructions stored in the memory 620 and executable on the processor 600; when the processor 600 executes the program or instructions, it implements the various processes of the above-described channel estimation method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0136] The transceiver 610 is used to receive and send data under the control of the processor 600.

[0137] Among them, Figure 6 In this context, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 600) and memory (memory 620). The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 610 may be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over a transmission medium. The processor 600 is responsible for managing the bus architecture and general processing, and the memory 620 may store data used by the processor 600 during operation.

[0138] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the above-described channel estimation method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0139] This application also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the various processes of the above-described channel estimation method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0140] It should be noted that the collection, gathering, updating, analysis, processing, use, transmission, and storage of user personal information involved in this disclosed technical solution all comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken to prevent unauthorized access to user personal information data and to safeguard user personal information security and network security.

[0141] It should be noted that, in this document, 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. Unless otherwise specified, 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 that element.

[0142] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0143] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A channel estimation method, characterized in that, include: Based on the received signals from the acquired antenna array, determine the energy spectrum data; The energy spectrum data is used as input to a convolutional neural network model to obtain a probability heatmap; The probability heatmap is used to characterize the probability of the scatterer peak on the grid; the convolutional neural network model is an optimized model using a focus loss function and a dice loss function. The directional parameters of the scatterer are determined based on the probability heatmap. The channel estimation result is determined based on the directional parameters of the scatterer.

2. The method according to claim 1, characterized in that, The convolutional neural network model is the U-Net neural network model; The supervision signal for the convolutional neural network model is a hard label; The output of the convolutional neural network model is a binary classification logical score graph.

3. The method according to claim 1, characterized in that, Based on the received signals from the acquired antenna array, the energy spectrum data is determined, including: The spatial domain signal of the received signal is mapped to the wavenumber domain to generate a two-dimensional energy spectrum. The two-dimensional energy spectrum is normalized to obtain the energy spectrum data; The antenna array is a uniform planar array.

4. The method according to claim 1, characterized in that, Based on the probability heatmap, the orientation parameters of the scatterer are determined, including: The pixel coordinates corresponding to the probability heatmap are obtained using an extraction algorithm; The pixel coordinates are subjected to non-maximum suppression processing to obtain pixel coordinate data; The pixel coordinate data is mapped to the physical wavenumber domain to obtain the direction parameters.

5. The method according to claim 1, characterized in that, Based on the directional parameters of the scatterer, the channel estimation result is determined, including: The directional parameters are input into the channel function of electromagnetic information theory to obtain the reconstructed channel covariance matrix; The channel covariance matrix is ​​processed by the minimum mean square error estimator to obtain the channel estimation result.

6. The method according to claim 1, characterized in that, The convolutional neural network model includes an encoder and a decoder; The encoder has multiple downsampling modules, each including a max pooling layer and multiple cascaded convolutional layers, with each convolutional layer connected to a normalization layer and an activation function. The decoder includes multiple upsampling modules, each upsampling module including multiple transposed convolutional layers and skip connections.

7. A channel estimation device, characterized in that, include: The first determining module is used to determine the energy spectrum data based on the received signal from the acquired antenna array; The second determining module is used to take the energy spectrum data as input to the convolutional neural network model to obtain a probability heatmap; The probability heatmap is used to characterize the probability of the scatterer peak on the grid; the convolutional neural network model is an optimized model using a focus loss function and a dice loss function. The third determining module is used to determine the orientation parameters of the scatterer based on the probability heatmap; The fourth determining module is used to determine the channel estimation result based on the directional parameters of the scatterer.

8. A channel estimation device, characterized in that, include: Transceiver, processor, memory, and programs or instructions stored in the memory and executable on the processor; When the processor executes the program or instructions, it implements the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 6.

Images