Intelligent wireless channel estimation method and system

Abstract

The invention discloses an intelligent wireless channel estimation method and system, and belongs to the technical field of wireless communication and artificial intelligence crossing, and the method comprises the steps: obtaining a wireless input signal to be subjected to channel estimation; inputting the wireless input signal into a pre-trained channel estimation model; wherein the channel estimation model is of a multi-branch structure, and a loss function of the channel estimation model is fused with physical knowledge constraints of a wireless communication system; and outputting a wireless channel estimation result by using the channel estimation model. By the adoption of the intelligent wireless channel estimation method and system, the channel estimation performance can be effectively improved, and the requirement for real-time channel estimation in a large-bandwidth multi-user receiving scene is met.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of wireless communication and artificial intelligence, and in particular to an intelligent wireless channel estimation method and system. Background Technology

[0002] In the field of wireless communication, whether it's 5G, Wi-Fi, or various systems based on OFDM (Orthogonal Frequency Division Multiplexing) technology, all must meet core communication requirements such as high speed, low latency, and multi-user concurrency. To this end, these systems generally employ key technologies such as large-bandwidth transmission, OFDMA (Orthogonal Frequency Division Multiple Access) resource allocation, and multi-antenna cooperation (e.g., MU-MIMO (Multiple-Input Multiple-Output)) to improve spectral efficiency and communication capacity. Against this backdrop, the real-time requirements for high-precision channel estimation and symbol detection at the receiver are becoming increasingly stringent. Traditional pilot-based channel estimation methods (such as LS (Least Squares Method) and MMSE (Minimum Mean Square Error)) not only suffer from limited accuracy in complex environments such as multipath fading, frequency offset interference, and noise enhancement, but also struggle to adapt to dynamically changing channel conditions and support high-performance communication scenarios.

[0003] In recent years, end-to-end deep learning methods have been gradually applied to wireless physical layer design, with some studies attempting to replace traditional modules such as channel estimation, equalization, and demodulation with neural networks. However, due to the lack of inherent perception of the time-frequency structure of the communication physical layer (such as pilot patterns, symbol power constraints, and phase continuity), neural networks often face problems such as poor generalization performance (e.g., weak cross-scenario adaptability), high computational complexity (difficult to meet real-time requirements), and substandard inference latency. These shortcomings are particularly prominent in scenarios such as 5G high-frequency bands, Wi-Fi high bandwidth, or OFDM high-speed transmission.

[0004] In the field of wireless communication (covering 5G, Wi-Fi, and various OFDM systems), existing receiver solutions rely on traditional signal processing frameworks. This involves obtaining the channel response through pilot-assisted channel estimation (such as LS and MMSE), and then combining this with equalization algorithms like ZF (Zero-Forcing) and MMSE to achieve symbol detection. These methods perform stably in low-noise, relatively stable channel scenarios (such as static or slow-fading environments), but they have limitations under complex communication conditions. In the face of strong path loss and beamforming errors in 5G high-frequency bands, co-channel interference in Wi-Fi multi-user scenarios, and high-frequency selective fading and carrier frequency offset in OFDM systems, their estimation accuracy and detection performance drop significantly. Furthermore, traditional methods rely on highly complex operations such as matrix inversion. When large-scale MIMO or OFDMA resource blocks (such as RBs in 5G and RUs in Wi-Fi) are discontinuously allocated, the computational complexity increases dramatically, making it difficult to meet real-time requirements.

[0005] In recent years, deep learning technology has been gradually introduced into channel estimation and symbol detection tasks, with some studies attempting to replace traditional modules with models such as CNN (Convolutional Neural Network) and LSTM (Long Short-Term Memory). However, existing solutions neglect the physical knowledge of wireless communication systems: for example, key features such as the time slot structure and reference signal pattern of 5G, the frame format and pilot distribution of Wi-Fi, and the power non-negativity and phase continuity of OFDM symbols are not effectively incorporated into network design. This results in the model being able to converge under the channel scenarios and modulation schemes corresponding to the training set, but its generalization performance is severely degraded in cross-channel environments (such as from urban macro base stations to indoor micro base stations), different modulation orders (such as from QPSK to 4096QAM), or variable resource allocation modes. At the same time, the high computational overhead brought about by the complex network structure makes it difficult for its inference latency to meet the time constraints of 5G low-latency services (such as URLLC), Wi-Fi high-bandwidth real-time transmission (such as 8K video), and OFDM high-speed communication, thus limiting its practical deployment value. Summary of the Invention

[0006] This invention provides an intelligent wireless channel estimation method and system to address the severe degradation of generalization performance in existing wireless channel estimation methods under cross-channel environments, different modulation orders, or variable resource allocation modes. Simultaneously, the high computational overhead caused by complex network structures leads to inference latency, making it difficult to meet the time constraints of 5G low-latency services, Wi-Fi high-bandwidth real-time transmission, and OFDM high-speed communication.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] On one hand, the present invention provides an intelligent wireless channel estimation method, comprising:

[0009] Acquire the wireless input signal for which channel estimation is to be performed;

[0010] The wireless input signal is input into a pre-trained channel estimation model; wherein the channel estimation model has a multi-branch structure and its loss function incorporates the physical knowledge constraints of the wireless communication system.

[0011] The wireless channel estimation results are output using the aforementioned channel estimation model.

[0012] Furthermore, the channel estimation model includes: a global frequency domain branch, a local frequency domain branch, a time-frequency joint branch, and a decoding module; wherein,

[0013] The global frequency domain branch is used to model the long-range correlation between multiple subcarriers to obtain global frequency domain features;

[0014] The local frequency domain branch is used to extract short-range frequency variation features to obtain local frequency domain features;

[0015] The time-frequency joint branch is used to extract time-frequency structural features to obtain cross-time-frequency joint features;

[0016] After the global frequency domain features, local frequency domain features, and cross-time-frequency joint features are fused, they are input into the decoding module, which directly maps the fused features into the channel frequency response of the required dimension.

[0017] Furthermore, the data processing steps for modeling the long-range correlation between multiple subcarriers using global frequency domain branching to obtain global frequency domain features include:

[0018] The frequency domain symbol on the pilot subcarrier in the wireless input signal is decomposed into real and imaginary parts to construct a real input. in, Indicates the real part, The imaginary part is represented by N, which is the sign dimension in the frequency domain; subsequently, the learnable parameter matrix is ​​used. Generate query, key, and value vector Q f =X in W q K f =X in W k V f =X in W v It employs multi-head attention computation, where each attention head outputs:

[0019]

[0020] in, and Let represent the query, key, and value vectors of the i-th attention head in the global frequency domain branch, respectively; d represents the total dimension of the feature vectors of the input multi-head attention; and h represents the number of attention heads. By concatenating the outputs of all attention heads, the global frequency domain features are obtained.

[0021] Furthermore, the data processing steps for extracting short-range frequency-varying features from local frequency domain branches to obtain local frequency domain features include:

[0022] The wireless input signal is reconstructed in the frequency domain as a two-dimensional tensor of length S. Where S is the number of subcarriers; Z1 is input into a convolutional layer, and the local subcarrier neighborhood information Y is extracted through the convolutional layer. conv ; then Y conv The system is divided into multiple subcarrier sequences. Each subcarrier sequence is input as a sequence into the attention layer. The attention layer applies attention weights to each subcarrier sequence to capture short-range frequency change patterns and obtain local frequency domain features.

[0023] Furthermore, the data processing procedure for extracting time-frequency structural features through the time-frequency joint branch to obtain cross-time-frequency joint features includes:

[0024] The OFDM symbol sequence of the wireless input signal is reconstructed into a three-dimensional time-frequency feature tensor X containing real and imaginary components. TF X TF Input a convolutional layer, and extract shallow features Y in the local time-frequency two-dimensional neighborhood through the convolutional layer. cnn Y cnn After flattening, the input is a bidirectional LSTM, and combined with 1×1 convolutional mapping residuals to establish long-range dependencies across frames and resource units, thus obtaining cross-time-frequency joint features.

[0025] Furthermore, the decoding module is a multilayer perceptron or a deconvolution.

[0026] Furthermore, the loss function of the channel estimation model is expressed as:

[0027]

[0028] Where β, γ, γ, δ, ∈, ζ are adjustable weight parameters used to balance the contributions of different constraints:

[0029] The error function for receiving the reference signal is expressed as:

[0030]

[0031] Where, N sc N represents the number of subcarriers. symFor time-domain symbolic numbers, Let y be the predicted value of the k-th subcarrier and the n-th time-domain symbol. k,n This represents the true value of the k-th subcarrier and the n-th time-domain symbol;

[0032] The subcarrier orthogonality constraint is expressed as:

[0033]

[0034] in, Let be the cross-correlation coefficient between subcarrier k and k+1 in the nth symbol; This represents the true value of the (k+1)th subcarrier and the nth time-domain symbol;

[0035] Cyclic prefix correlation constraints are expressed as follows:

[0036]

[0037] in, The prediction cyclic prefix part of the nth OFDM symbol, N is the tail of the nth symbol body. cp N sampling points cp The cyclic prefix length is given by corr(·), and the correlation coefficient is given by corr(·).

[0038] For pilot subcarrier constraints, it is expressed as:

[0039]

[0040] in, Let N be the set of pilot subcarriers. pilot For pilot numbers, The pilot prior values ​​are known. k represents the channel estimate predicted by the network at the pilot location. p The frequency index of the pilot subcarrier, n p Indicates the time-domain symbol index of the pilot signal;

[0041] For time-frequency domain phase constraints, it is expressed as:

[0042]

[0043] in, For time-domain phase constraints, For frequency domain phase constraint, in, This represents the predicted symbol phase angle at the nth OFDM symbol and the kth subcarrier. This represents the predicted symbol phase angle at the (n-1)th OFDM symbol and the kth subcarrier. This represents the predicted symbol phase angle at the nth OFDM symbol and the (k-1)th subcarrier;

[0044] The modulation domain constellation constraint is represented as:

[0045]

[0046] in, N is the set of data subcarriers. data For the number of data subcarriers, For the modulation constellation point set; k represents the subcarrier frequency index, n represents the OFDM symbol time index, and c represents the modulation constellation point set. A symbol of an ideal constellation;

[0047] For amplitude constraints, it is expressed as:

[0048] On the other hand, the present invention also provides an intelligent wireless channel estimation system, comprising:

[0049] The signal acquisition module is used to acquire the wireless input signal to be channel estimated.

[0050] Signal processing module, used for:

[0051] The wireless input signal is input into a pre-trained channel estimation model; wherein the channel estimation model has a multi-branch structure and its loss function incorporates the physical knowledge constraints of the wireless communication system.

[0052] The wireless channel estimation results are output using the aforementioned channel estimation model.

[0053] In another aspect, the present invention also provides an electronic device comprising a processor and a memory; wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described method.

[0054] In another aspect, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the above method.

[0055] The beneficial effects of the technical solution provided by this invention include at least the following:

[0056] 1. Strong anti-interference capability: Compared with the traditional LS / MMSE estimation scheme, the end-to-end learning structure can adaptively model non-ideal channel environments such as multipath fading, frequency offset, and co-channel interference, such as 5G high-frequency bands, dense Wi-Fi scenarios, and OFDM multi-user scenarios.

[0057] 2. Low latency and high throughput: Fast signal recovery can be achieved without relying on pilot signals, adapting to the low latency requirements of 5G services, high bandwidth transmission of Wi-Fi and high speed requirements of OFDM, effectively improving spectrum efficiency;

[0058] 3. Strong physical interpretability: The loss function incorporates physical constraints such as non-negative power, continuous phase, and matching modulation constellation to ensure that the output conforms to the laws of communication signals and avoids performance degradation caused by unreasonable data;

[0059] 4. Strong structural scalability: The network can be flexibly expanded to 5G multi-user MIMO, Wi-Fi multi-RU allocation and different modulation order scenarios of various OFDM systems, adapting to diverse time and frequency resource configurations. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a schematic diagram of the overall structure of the multi-scale, multi-branch feature fusion network provided in an embodiment of the present invention;

[0062] Figure 2 This is a comparison chart of the loss functions of the model proposed in this invention with those of DNN and LSTM models;

[0063] Figure 3 These are simulation results of the proposed model provided in the embodiments of the present invention;

[0064] Figure 4 This is a system block diagram of the electronic device provided in the embodiments of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0066] First, it should be noted that in the embodiments of the present invention, the words "exemplarily," "for example," etc., are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the term "exemplarily" is intended to present the concept in a specific manner. Furthermore, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either one or the other.

[0067] First Embodiment

[0068] This embodiment provides an intelligent wireless channel estimation method, applicable to intelligent channel estimation in multi-user OFDM communication scenarios; the method can be implemented by an electronic device and includes the following steps:

[0069] S1, acquire the wireless input signal to be channel estimated;

[0070] S2, input the wireless input signal into a pre-trained channel estimation model; wherein, the channel estimation model is a multi-branch structure, and its loss function incorporates the physical knowledge constraints of the wireless communication system;

[0071] S3, use the channel estimation model to output the wireless channel estimation result;

[0072] It should be noted that, in order to accurately model the multi-resolution characteristics of large-bandwidth wireless signals in the frequency and time domains of a wireless communication system, this embodiment proposes a method such as... Figure 1 The multi-branch, multi-scale fusion network shown perceives global frequency domain features, local frequency domain features, and cross-time-frequency joint features, as detailed below:

[0073] 1) Global frequency domain branch

[0074] The frequency domain symbols on the pilot subcarriers of the input signal of the OFDM system (such as the EHT-LTF frequency domain symbols in 802.11be) are used as N-dimensional complex eigenvectors. It is then decomposed into real and imaginary parts to construct a real number input. in, Indicates the real part, The imaginary part is represented; subsequently, the learnable parameter matrix is ​​used. Generate query, key, and value vector Q f =X in W q K f =X in W k V f =X in W v And it employs multi-head attention computation, where each attention head The output represents the weighted characteristics of the inter-carrier interference pattern. and Let be the query and key vectors of the i-th attention head in the global frequency domain branch, respectively, where d is the total dimension of the feature vector and h is the number of attention heads. The global frequency domain features are obtained by calculating all attention heads and concatenating them. The structure explicitly models interference patterns among multiple subcarriers and can learn global features such as subcarrier power allocation and frequency offset compensation.

[0075] 2) Local frequency domain branch

[0076] The input signal is reconstructed in the frequency domain as a two-dimensional tensor of length S. Here, 2 represents the two channels, real and imaginary, and S is the number of subcarriers; specifically: the input signal is reconstructed into a complex vector of length S (corresponding to the frequency domain symbols of S subcarriers); the complex value of each subcarrier is decomposed into a real and an imaginary part, resulting in two real vectors, each of length S; these two vectors are stacked row-wise to form a two-dimensional tensor of shape (2, S). Each row represents the real and imaginary parts of all subcarriers.

[0077] The extraction of local subcarrier neighborhood information through convolutional layers can be represented as Y. conv =ReLU(W c *Z1+b c (convolution kernel) k is the size of the neighborhood window); then Y... conv After dividing the subcarrier sequence into multiple small segments (e.g., each segment is of length L), each segment is input into the attention layer as a sequence. The attention layer applies attention weights to each small subcarrier sequence, and the output is the local frequency domain feature L. local By reducing the number of attention heads, computational efficiency and frequency domain resolution are balanced, thereby effectively capturing short-range frequency variation patterns.

[0078] 3) Time-frequency joint branch

[0079] The OFDM symbol sequence of the input signal is reconstructed into a three-dimensional time-frequency feature tensor X containing real and imaginary components. TF ∈R T×F×2 (Where T is the number of time-domain frames and F is the number of subcarriers), specifically: an OFDM time-domain frame of length T, each frame containing F subcarriers, each subcarrier being a complex symbol, each complex symbol being decomposed into a real part and an imaginary part, and treated as two independent channels, to obtain X. TF ∈R T×F×2 , where channel 0 represents the real part of all (t,f) positions and channel 1 represents the imaginary part of all (t,f) positions.

[0080] shallow features within the local time-frequency two-dimensional neighborhood are extracted using convolutional layers. (convolution kernel) k t For the time dimension window size, k f C1 is the subcarrier-dimensional window size, C1 is the number of output channels, and b c (where F is the effective length in the frequency dimension after convolution, and is the bias), flattening Afterwards, via hidden dimension d lstm Bidirectional LSTM Using 1×1 convolution (W res As the weight, W res ,b res (for bias) Mapped residual Z t =vec(W res *Y cnn [t,:,:]+b res This establishes long-range dependencies across frames / resource units (RUs), ultimately yielding time-frequency characteristics. This enables joint modeling of time-frequency structural features such as pilot patterns, data symbols, and guard intervals. Among them, Y... cnn [t,:,:] represents the two-dimensional feature matrix (of shape F'×C1) of all subcarriers and all convolutional channels corresponding to the t-th time frame (OFDM symbol), and vec(.) represents the vectorization operation of expanding the matrix or tensor into a one-dimensional vector by columns (or by rows); This represents the hidden state vector output by the forward LSTM at time step t, with dimension d. lstm , This represents the hidden state vector output by the inverse LSTM at time step t, with dimension d. lstm The `reshape(.)` function rearranges a given vector or matrix into a new tensor with specified dimensions. This represents the set of sequences consisting of the residual connection results (the sum of the bidirectional LSTM output and the 1×1 convolutional residual) corresponding to T time steps; it should be noted that d lstm The hidden state dimension is 1 for a unidirectional LSTM; the dimension after bidirectional concatenation is 2d. lstm Hidden dimension d lstm This refers to a hidden state with dimension d. lstm A unidirectional LSTM.

[0081] 4) Feature fusion and decoding

[0082] Frequency domain attention features output by the global frequency domain branch Short-range frequency-varying features L extracted by local frequency domain branching local And the time-frequency feature H generated by the time-frequency joint branch TFPerform feature fusion and construct a fusion vector:

[0083]

[0084] Where, d f The total feature dimension is denoted as . This fused vector is directly mapped to the channel frequency response (CFR) of the desired dimension by a decoding module (multilayer perceptron or deconvolution). The decoding process can be represented as follows: Among them, W d Let σ be the decoding weight matrix, and σ be the activation function.

[0085] Based on the above, for feature extraction of OFDM signals, this embodiment constructs a multi-branch network architecture based on a multi-scale attention mechanism. Frequency domain features are extracted through the collaborative processing of global frequency domain branch (modeling long-range correlation between multiple subcarriers) and local frequency domain branch (extracting short-range frequency variation features). Time-frequency structural features such as pilots, data symbols and guard intervals are extracted through time-frequency joint branch.

[0086] Furthermore, this embodiment constructs a loss function based on physical structure awareness, using subcarriers, cyclic prefixes, pilots, time-frequency domain phase, constellation point distance, reference signal error function, and power as constraints. It should be noted that OFDM signals achieve high spectral efficiency through subcarrier orthogonality, with core characteristics including: strict orthogonality between subcarriers (resisting inter-carrier interference, ICI); time-domain symbols containing cyclic prefixes (CP) to combat inter-symbol interference (ISI); continuity of channel characteristics among adjacent resource units (REs) in the time-frequency resource grid (subcarrier-symbol dimension) (smoothness of multipath fading); known pilot subcarrier positions and powers (for synchronization and channel estimation); data subcarriers conforming to modulation constellation constraints; and non-negativity of signal power and continuous evolution of phase over time / frequency (phase gradient caused by multipath delay and Doppler effect). Combining these constraints, the total loss function is:

[0087]

[0088] Where α, β, γ, δ, ∈, ζ are adjustable weight parameters used to balance the contributions of different constraints: in high-frequency mobile scenarios (such as 5G vehicle communication), δ (strengthening phase continuity) and α (countering ICI) can be increased; in multipath severe scenarios (such as indoor Wi-Fi), β (strengthening CP constraints) can be increased; in low signal-to-noise ratio scenarios, ∈ (strengthening constellation constraints) and γ (depending on pilot priors) can be increased.

[0089] 1) Received reference signal error function

[0090]

[0091] Where, N sc N represents the number of subcarriers. sym For time-domain symbolic numbers, Let y be the predicted value of the k-th subcarrier and the n-th time-domain symbol. k,n This represents the true value of the k-th subcarrier and the n-th time-domain symbol.

[0092] 2) Subcarrier orthogonality constraint:

[0093]

[0094] in, Let be the cross-correlation coefficient between subcarrier k and k+1 at the nth symbol (value range [-1, 1]); This represents the true value of the (k+1)th subcarrier and the nth time-domain symbol.

[0095] 3) Cyclic prefix (CP) correlation constraint (anti-ISI):

[0096]

[0097] in, The prediction CP part (N) of the nth OFDM symbol cp (where the length of the cyclic prefix is) N is the tail of the nth symbol body. cp There are 10 sampling points, and corr(·) is the correlation coefficient.

[0098] 4) Pilot subcarrier constraints:

[0099]

[0100] in, Let N be the set of pilot subcarriers. pilot For pilot numbers, The pilot prior values ​​are known (e.g., fixed power, sign of a specific phase); k represents the channel estimate predicted by the network at the pilot location. p The frequency index of the pilot subcarrier, n p This indicates the time-domain symbol index where the pilot signal is located.

[0101] 5) Time-frequency domain phase constraint:

[0102]

[0103] Among them, the time-domain phase constraint (for the same subcarrier k, adjacent symbols n and n-1):

[0104]

[0105] in, This represents the predicted symbol phase angle at the nth OFDM symbol and the kth subcarrier. This represents the predicted symbol phase angle at the (n-1)th OFDM symbol and the kth subcarrier.

[0106] Frequency domain phase constraint (for the same symbol n, adjacent subcarriers k and k-1):

[0107]

[0108] in, This represents the predicted symbol phase angle at the nth OFDM symbol and the (k-1)th subcarrier;

[0109] 6) Modulation domain constellation constraints:

[0110]

[0111] in, N is the set of data subcarriers. data For the number of data subcarriers, To modulate constellation point sets (such as QPSK) k represents the subcarrier frequency index, n represents the OFDM symbol time index, and c represents the modulation constellation point set. A symbol of an ideal constellation.

[0112] 7) Amplitude (power) constraint:

[0113]

[0114] The loss function is constrained by subcarriers, cyclic prefixes, pilots, time-frequency domain phase, constellation point distance, reference signal error function, and power. Adjustable weights α, β, γ, δ, ∈, ζ form a cooperative feedback loop with the multi-branch network, and the global frequency domain branch is constrained by subcarrier orthogonality. Pilot prior constraints Direct backpropagation, with local frequency domain branches constrained by power. and the distance of constellation points Common constraints, time-frequency joint branch due to cyclic prefix correlation With time-frequency phase constraint The gradient is updated synchronously in both the time and frequency domains across frames using a bidirectional LSTM. Its comparison with the loss functions of DNN and LSTM models is as follows: Figure 2 As shown.

[0115] The effectiveness of the method of the present invention will be verified by simulation below.

[0116] The simulation, conducted using the MATLAB platform, is based on a multi-user MU-MIMO scenario within IEEE 802.11be (Wi-Fi 7). It employs a 320MHz bandwidth, 6 transmit antennas, and a 4-user, 4-space-time stream configuration to perform ultra-high-speed multi-user (EHTMU) signal transmission, channel estimation, and bit error rate (BER) calculation. The simulation settings include MCS 4 modulation (QPSK, 3 / 4 code rate), a packet length of 512 bytes, and ultra-high-speed long training field (EHT-LTF) symbol type. The TGax ChannelModel B model is used, considering a 5-meter non-line-of-sight distance and independent channels for multiple users. Singular value decomposition (SVD) feedback is obtained through the Neighbor Discovery Protocol (NDP) to calculate the MU-MIMO beamforming matrix. Multiple random packet transmissions are performed, and an AWGN is added. The synchronized EHT-LTF frequency domain symbols are used as training data, and the transmitted EHT-LTF frequency domain symbols are used as tag data for channel estimation, outputting the channel frequency response (CFR). Specific simulation results are shown below. Figure 3 As shown.

[0117] In summary, this embodiment proposes an intelligent channel estimation method that combines physical information with the adaptive capabilities of neural networks. This method can improve the accuracy and robustness of channel estimation in wireless signals under multi-frequency selectivity and multi-user interference scenarios; enhance the performance of traditional channel estimation methods to meet the real-time channel estimation requirements in high-bandwidth multi-user reception scenarios; and ensure the physical rationality and interpretability of the neural network structure, thereby improving its generalization and accuracy in channel estimation under different channel conditions.

[0118] Second Embodiment

[0119] This embodiment provides an intelligent wireless channel estimation system, which includes the following modules:

[0120] The signal acquisition module is used to acquire the wireless input signal to be channel estimated.

[0121] Signal processing module, used for:

[0122] The wireless input signal is input into a pre-trained channel estimation model; wherein the channel estimation model has a multi-branch structure and its loss function incorporates the physical knowledge constraints of the wireless communication system.

[0123] The wireless channel estimation results are output using the aforementioned channel estimation model.

[0124] It should be noted that the intelligent wireless channel estimation system of this embodiment corresponds to the intelligent wireless channel estimation method of the first embodiment described above; the functions implemented by each functional module in the intelligent wireless channel estimation system of this embodiment correspond one-to-one with the process steps in the intelligent wireless channel estimation method of the first embodiment described above; therefore, they will not be described again here.

[0125] Third Embodiment

[0126] This embodiment provides an electronic device, such as... Figure 4 As shown, the electronic device includes a processor and a memory; wherein the processor and the memory can be connected via a communication bus; the memory stores at least one instruction, which is loaded and executed by the processor to implement the method of the first embodiment described above. Furthermore, the electronic device may also include a transceiver, the processor and the transceiver can be connected via a communication bus, and the transceiver is used to communicate with other devices.

[0127] Below, in conjunction with Figure 4 A detailed introduction to each component of this electronic device is provided below:

[0128] The processor is the control center of the electronic device. The electronic device may include multiple processors, each of which can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The term "processor" can refer to a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), other general-purpose processors, application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), one or more field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor can perform various functions of the electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0129] In a specific implementation, as one example, the processor may include one or more CPUs, for example... Figure 4CPU0 and CPU1 shown are, of course, merely illustrative examples.

[0130] The memory is used to store the software program that executes the solution of the present invention, and the processor controls its execution. For specific implementation methods, please refer to the above method embodiments, which will not be repeated here.

[0131] Optionally, the memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory may be integrated with the processor or may exist independently, and may be accessed through the interface circuit of the electronic device (…). Figure 4 (Not shown in the image) is coupled to the processor; however, this embodiment of the invention does not impose specific limitations on this.

[0132] The transceiver may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function. The transceiver can be integrated with the processor or exist independently, and can be connected through the interface circuit of the electronic device (…). Figure 4 (Not shown in the image) is coupled to the processor, and this embodiment of the invention does not specifically limit this.

[0133] In addition, it should be noted that, Figure 4 The structure of the electronic device shown is not intended to limit the device. Actual devices may include more or fewer components than shown, or combine certain components, or have different component arrangements. Furthermore, the technical effects achieved by this electronic device when performing the method of the first embodiment described above can be referenced to the technical effects described in the first embodiment; therefore, they will not be repeated here.

[0134] Fourth embodiment

[0135] This embodiment provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the method of the first embodiment described above. The computer-readable storage medium may be a ROM, random access memory, CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc. The instruction stored therein can be loaded and executed by a processor in a terminal.

[0136] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely or partially hardware embodiment, a completely or partially software embodiment, or an embodiment combining software and hardware aspects. Moreover, when implemented in software, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any usable medium accessible to a computer or a data storage device such as a server or data center containing one or more sets of usable media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive (SSD).

[0137] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. 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, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate 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 functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal 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.

[0139] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device 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 terminal device. 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 terminal device that includes said element. Furthermore, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Additionally, the character " / " in this text generally indicates an "or" relationship between the preceding and following objects, but it can also indicate an "AND / OR" relationship; please refer to the context for specific interpretations. "At least one" refers to one or more items, while "more than" refers to two or more items. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can be expressed as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0140] Furthermore, it is understood that in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0141] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0142] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of functional modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Additionally, the functional units in the various embodiments of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0143] If the method is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0144] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments of the present invention have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make several improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A smart wireless channel estimation method, characterized in that, include: Acquire the wireless input signal for which channel estimation is to be performed; The wireless input signal is input into a pre-trained channel estimation model; wherein the channel estimation model has a multi-branch structure and its loss function incorporates the physical knowledge constraints of the wireless communication system. The wireless channel estimation results are output using the aforementioned channel estimation model.

2. The intelligent wireless channel estimation method as described in claim 1, characterized in that, The channel estimation model includes: a global frequency domain branch, a local frequency domain branch, a time-frequency joint branch, and a decoding module; wherein... The global frequency domain branch is used to model the long-range correlation between multiple subcarriers to obtain global frequency domain features; The local frequency domain branch is used to extract short-range frequency variation features to obtain local frequency domain features; The time-frequency joint branch is used to extract time-frequency structural features to obtain cross-time-frequency joint features; After the global frequency domain features, local frequency domain features, and cross-time-frequency joint features are fused, they are input into the decoding module, which directly maps the fused features into the channel frequency response of the required dimension.

3. The intelligent wireless channel estimation method as described in claim 2, characterized in that, The data processing steps for obtaining global frequency domain features by modeling the long-range correlation between multiple subcarriers using global frequency domain branching include: The frequency domain symbol on the pilot subcarrier in the wireless input signal is decomposed into real and imaginary parts to construct a real input. in, Indicates the real part, The imaginary part is represented by N, which is the sign dimension in the frequency domain; subsequently, the learnable parameter matrix is ​​used. Generate query, key, and value vector Q f =X in W q K f =X in W k V f =X in W v It employs multi-head attention computation, where each attention head outputs: in, and Let represent the query, key, and value vectors of the i-th attention head in the global frequency domain branch, respectively; d represents the total dimension of the feature vectors of the input multi-head attention; and h represents the number of attention heads. By concatenating the outputs of all attention heads, global frequency domain features are obtained.

4. The intelligent wireless channel estimation method as described in claim 2, characterized in that, The data processing steps for extracting short-range frequency variation features using local frequency domain branching to obtain local frequency domain features include: The wireless input signal is reconstructed in the frequency domain as a two-dimensional tensor of length S. Where S is the number of subcarriers; Z1 is input into a convolutional layer, and the local subcarrier neighborhood information Y is extracted through the convolutional layer. conv ; then Y conv The system is divided into multiple subcarrier sequences. Each subcarrier sequence is input as a sequence into the attention layer. The attention layer applies attention weights to each subcarrier sequence to capture short-range frequency change patterns and obtain local frequency domain features.

5. The intelligent wireless channel estimation method as described in claim 2, characterized in that, The data processing steps for extracting time-frequency structural features through time-frequency joint branching to obtain cross-time-frequency joint features include: The OFDM symbol sequence of the wireless input signal is reconstructed into a three-dimensional time-frequency feature tensor X containing real and imaginary components. TF X TF Input a convolutional layer, and extract shallow features Y in the local time-frequency two-dimensional neighborhood through the convolutional layer. cnn Y cnn After flattening, the input is a bidirectional LSTM, and combined with 1×1 convolutional mapping residuals to establish long-range dependencies across frames and resource units, thus obtaining cross-time-frequency joint features.

6. The intelligent wireless channel estimation method as described in claim 2, characterized in that, The decoding module is a multilayer perceptron or a deconvolution module.

7. The intelligent wireless channel estimation method as described in claim 1, characterized in that, The loss function of the channel estimation model is expressed as: Where α,β,γ,δ,∈,ζ are adjustable weight parameters used to balance the contributions of different constraints: The error function for receiving the reference signal is expressed as: Where, N sc N is the number of subcarriers. sym For time-domain symbolic numbers, Let y be the predicted value of the k-th subcarrier and the n-th time-domain symbol. k,n This represents the true value of the k-th subcarrier and the n-th time-domain symbol; The subcarrier orthogonality constraint is expressed as: in, Let be the cross-correlation coefficient between subcarrier k and k+1 in the nth symbol; This represents the true value of the (k+1)th subcarrier and the nth time-domain symbol; Cyclic prefix correlation constraints are expressed as follows: in, The prediction cyclic prefix part of the nth OFDM symbol, N is the tail of the nth symbol body. cp N sampling points cp The cyclic prefix length is given by corr(·), and the correlation coefficient is given by corr(·). For pilot subcarrier constraints, it is expressed as: in, Let N be the set of pilot subcarriers. pilot For pilot numbers, The pilot prior values ​​are known. k represents the channel estimate predicted by the network at the pilot location. p The frequency index of the pilot subcarrier, n p Indicates the time-domain symbol index of the pilot signal; For time-frequency domain phase constraints, it is expressed as: in, For time-domain phase constraints, For frequency domain phase constraint, in, This represents the predicted symbol phase angle at the nth OFDM symbol and the kth subcarrier. This represents the predicted symbol phase angle at the (n-1)th OFDM symbol and the kth subcarrier. This represents the predicted symbol phase angle at the nth OFDM symbol and the (k-1)th subcarrier; The modulation domain constellation constraint is represented as: in, N is the set of data subcarriers. data For the number of data subcarriers, For the modulation constellation point set; k represents the subcarrier frequency index, n represents the OFDM symbol time index, and c represents the modulation constellation point set. A symbol of an ideal constellation; For amplitude constraints, it is expressed as:

8. An intelligent wireless channel estimation system, characterized in that, include: The signal acquisition module is used to acquire the wireless input signal to be channel estimated. Signal processing module, used for: The wireless input signal is input into a pre-trained channel estimation model; wherein the channel estimation model has a multi-branch structure and its loss function incorporates the physical knowledge constraints of the wireless communication system. The wireless channel estimation results are output using the aforementioned channel estimation model.

Images