Channel estimation method, system and device based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network, and medium

Abstract

The invention discloses a channel estimation method, system and device based on signal-to-noise ratio adaptive modulation and a bidirectional LSTM network and a medium, and relates to the field of wireless communication, and the method comprises the steps: carrying out the preliminary channel estimation through employing a least square method based on a received signal and a pilot signal, and obtaining a coarse channel estimation value at each moment; performing statistical analysis on the coarse channel estimation value and the residual signal to obtain a multi-dimensional statistical feature; inputting the multi-dimensional statistical features into a multi-layer perceptron network, and predicting a signal-to-noise ratio at each moment; inputting the signal-to-noise ratio into an MLP-FiLM network to generate a corresponding characteristic linear modulation parameter; inputting the received signal into the bidirectional LSTM network, and performing affine transformation on the hidden state of the bidirectional LSTM network by using the characteristic linear modulation parameter to obtain a modulated hidden characteristic; and inputting the modulated hidden features into a full connection layer, and outputting a final channel estimation value through linear mapping. According to the invention, the channel estimation precision and the system robustness under the complex time-varying channel can be improved.

Description

Technical Field

[0001] This application relates to the field of wireless communication, and in particular to a channel estimation method, system, device, and medium based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network. Background Technology

[0002] Channel estimation refers to the process by which the receiver in a communication system uses certain methods and algorithms to acquire or infer the impact of the wireless channel on the transmitted signal (i.e., channel characteristics). More specifically, it is the process of determining the mathematical description (such as the channel impulse response CIR or channel frequency response CFR) of the channel that the signal traverses from the transmitter to the receiver.

[0003] The performance of wireless communication systems is largely influenced by the wireless channel, factors such as shadowing fading and frequency-selective fading, which make the propagation path between the transmitter and receiver highly complex. Unlike wired channels, which are fixed and predictable, wireless channels exhibit significant randomness, posing a considerable challenge to receiver design. If the receiver is unaware of these channel effects, it becomes difficult to accurately recover the original information transmitted by the transmitter, especially under adverse channel conditions or with complex modulation and coding schemes. In coherent detection of OFDM systems, channel estimation is necessary, and the accuracy of this estimation directly impacts the overall system performance.

[0004] Traditional channel estimation algorithms mainly include least squares (LS), minimum mean square error (MMSE), and compressed sensing algorithms. LS estimates the channel by minimizing the sum of squared errors based on pilot information from the transmitter; however, it requires pilot signals as an aid, leading to high pilot overhead, especially in large-array antenna scenarios, where estimation accuracy decreases. MMSE utilizes prior channel knowledge and statistical information for channel estimation, requiring matrix inversion. As the dimension of the channel matrix increases in large-array antennas, the inversion process becomes more complex, resulting in high computational complexity. While compressed sensing algorithms can effectively acquire sparse channel information and recover the original signal by leveraging signal sparsity, they have strict requirements on channel sparsity and can only be applied to millimeter-wave massive MIMO systems with specific sparsity characteristics. Summary of the Invention

[0005] The purpose of this application is to provide a channel estimation method, system, device, and medium based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network. By combining the signal-to-noise ratio adaptive modulation mechanism (SNR-Adaptive Modulation) and the deep temporal feature learning network (FiLM-BiLSTM), dynamic weighting and feature adjustment are achieved at different signal-to-noise ratio times during signal reception and channel estimation, thereby significantly improving the channel estimation accuracy and system robustness under complex time-varying channels.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a channel estimation method based on signal-to-noise ratio adaptive modulation and a bidirectional LSTM network, including: Based on the received signal and pilot signal, the least squares method is used to perform preliminary channel estimation and obtain coarse channel estimates at each time point. Statistical analysis was performed on the coarse channel estimate and the residual signal to obtain multidimensional statistical characteristics; The multidimensional statistical features are input into a multilayer perceptron network to predict the signal-to-noise ratio at each time step. The signal-to-noise ratio is input into the MLP-FiLM network to generate corresponding characteristic linear modulation parameters; the characteristic linear modulation parameters include characteristic scaling coefficients and characteristic offset coefficients; The received signal is input into a bidirectional LSTM network, and the hidden state of the bidirectional LSTM network is subjected to an affine transformation using the characteristic linear modulation parameters to obtain the modulated hidden features. The modulated hidden features are input into a fully connected layer, and the final channel estimate is output through linear mapping.

[0007] Secondly, this application provides a channel estimation system based on signal-to-noise ratio adaptive modulation and a bidirectional LSTM network, comprising: The preliminary channel estimation unit is used to perform preliminary channel estimation based on the received signal and pilot signal using the least squares method to obtain coarse channel estimates at each time step. The statistical analysis unit is used to perform statistical analysis on the coarse channel estimate and the residual signal to obtain multidimensional statistical characteristics. The signal-to-noise ratio prediction unit is used to input the multidimensional statistical features into the multilayer perceptron network and predict the signal-to-noise ratio at each time step. The characteristic linear modulation parameter generation unit is used to input the signal-to-noise ratio into the MLP-FiLM network and generate the corresponding characteristic linear modulation parameters; the characteristic linear modulation parameters include characteristic scaling coefficients and characteristic offset coefficients; The modulated hidden feature generation unit is used to input the received signal into the bidirectional LSTM network and use the feature linear modulation parameters to perform an affine transformation on the hidden state of the bidirectional LSTM network to obtain the modulated hidden features. The final channel estimation unit is used to input the modulated hidden features into the fully connected layer and output the final channel estimation value through linear mapping.

[0008] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described channel estimation method based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network.

[0009] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned channel estimation method based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network.

[0010] According to the specific embodiments provided in this application, this application has the following technical effects: (1) Reduce pilot overhead while ensuring the accuracy of preliminary estimation: This application uses LS (Least Square) as the basis for preliminary channel estimation. At the same time, it optimizes the subsequent bidirectional LSTM network and signal-to-noise ratio adaptive modulation. It can improve the final estimation accuracy without relying on a large number of pilots. It retains the advantage of simple implementation of LS algorithm and overcomes its defects of large pilot overhead and insufficient accuracy in large array scenarios. (2) Achieving accurate prediction of signal-to-noise ratio and providing a reliable basis for adaptive modulation: This application performs statistical analysis on coarse channel estimates and residual signals, extracts multidimensional statistical features and inputs them into a multilayer perceptron (MLP) to predict the signal-to-noise ratio, which can accurately capture the real-time change characteristics of the channel. The output signal-to-noise ratio parameters can directly provide data support for subsequent feature modulation, realizing dynamic matching between channel state and modulation strategy. (3) Based on the adaptive feature modulation of MLP-FiLM network, the feature representation capability is improved: The predicted signal-to-noise ratio is input into the MLP-FiLM network to generate feature scaling coefficients and feature offset coefficients. The hidden state of the bidirectional LSTM network is modulated by affine transformation, so that the network can dynamically adjust the feature weights according to the real-time signal-to-noise ratio, strengthen the effective features and suppress the noise features, and solve the problems of fixed feature processing and poor adaptability to complex channels in traditional algorithms. (4) Improve the accuracy of channel estimation by utilizing the timing modeling capability of bidirectional LSTM network: This application uses bidirectional LSTM network to process the received signal. This network can simultaneously utilize the forward and backward timing information of the signal to fully explore the timing correlation of the channel. It is especially suitable for non-sparse channel scenarios, breaking through the strict limitation of compressed sensing algorithm on channel sparsity and improving the versatility of the algorithm. At the same time, combined with the hidden state characteristics of adaptive modulation, the estimation error caused by random channel changes is further reduced. (5) Reduced computational complexity, suitable for large array antenna scenarios: The core computational module of this application is a bidirectional LSTM network and an MLP (Multilayer Perceptron), which does not require complex matrix inversion operations. Moreover, the modulation process of the MLP-FiLM network is a linear operation, and the computational overhead is controllable. At the same time, the preliminary estimate shows that the use of the LS algorithm further simplifies the early computation process, and the overall computational complexity of the scheme is reduced, making it more suitable for complex communication scenarios such as large-scale MIMO and large array antennas. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a channel estimation method based on signal-to-noise ratio adaptive modulation and a bidirectional LSTM network, provided as an embodiment of this application; Figure 2 A schematic diagram of the overall process of a channel estimation method based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network provided in an embodiment of this application; Figure 3 This is a schematic diagram of the transmitted signal structure. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0015] In one exemplary embodiment, such as Figures 1-2 As shown, a channel estimation method based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is described using a server as an example, including the following steps 201 to 208. Wherein: S1: Based on the received signal and pilot signal, the least squares method is used to perform preliminary channel estimation to obtain coarse channel estimates at each time point.

[0016] S2: Perform statistical analysis on the coarse channel estimate and the residual signal to obtain multidimensional statistical characteristics.

[0017] S3: Input the multidimensional statistical features into the multilayer perceptron network to predict the signal-to-noise ratio at each time step.

[0018] S4: Input the signal-to-noise ratio into the MLP-FiLM network to generate corresponding characteristic linear modulation parameters; the characteristic linear modulation parameters include characteristic scaling coefficients and characteristic offset coefficients.

[0019] S5: Input the received signal into the bidirectional LSTM network, and use the characteristic linear modulation parameters to perform an affine transformation on the hidden state of the bidirectional LSTM network to obtain the modulated hidden features.

[0020] S6: Input the modulated hidden features into the fully connected layer and output the final channel estimate through linear mapping.

[0021] The core idea of ​​this application is to combine the signal-to-noise ratio adaptive modulation (SNR-Adaptive Modulation) mechanism with a bidirectional LSTM network to achieve dynamic weighting and feature modulation at different SNR times during signal reception and channel estimation, thereby significantly improving the channel estimation accuracy and system robustness under complex time-varying channels.

[0022] In a specific embodiment, step S1 specifically includes: constructing an objective function based on the received signal and the pilot signal using the least squares criterion; solving the objective function to obtain coarse channel estimates at each time step.

[0023] like Figure 3 As shown, the transmitted signal structure follows the existing IEEE standard. One frame of signal contains K blocks, which can be divided into two parts. The first part is the preamble training block, and the second part is the data block with pilot symbols inserted. Each data block contains... Data symbols and There are pilot symbols. Therefore, one data block has a total of There are symbols, and 1 frame of signal has _ symbols_. Pilot symbols are inserted into each block at equal intervals; the pilot density is defined as... The relationship between channel input and output can be expressed as: ,in: In order to receive signals, For the channel matrix, In order to transmit signals, This is a noise signal.

[0024] LS channel estimation is a channel estimation method based on the least squares criterion. In wireless systems, when the transmitted signal is a pilot signal... At that time, receive signal It can be represented as: in, Pilot signal, For the channel matrix, This is a noise signal.

[0025] The estimated channel can be expressed as: According to the least squares criterion, we have the following objective function: J To obtain the squared error J smallest Taking the derivative of the objective function and setting its derivative to zero, we get: in, This is a coarse channel estimate. This is the conjugate transpose of the matrix.

[0026] In a specific embodiment, step S2 specifically includes: statistically modeling the coarse channel estimate and the residual signal within a sliding time window, and extracting multidimensional statistical features. These multidimensional statistical features include the mean residual energy, the variance of the residual energy, the median of the residual modulus, the estimated power, the differential energy between adjacent time points, the pilot power, the residual ratio, and the residual moving average.

[0027] residual signal The calculation formula is: Mean residual energy (mean_resid): ; Residual energy variance (var_resid): ; Median residual modular square (median_resid): ; LS estimated power ( ): ; Energy difference between adjacent time steps (adj_time_diff): ; Pilot power (p_power): ; Signal-to-residual ratio (mean_ratio): ; Residual moving average (ema_resid): ; in, Let N represent the residual signal at each time point within the sliding time window, and let N represent the length of the sliding time window. This represents the channel estimate at each time point within the sliding time window. This represents the difference between the channel estimates at the previous and next time steps within the sliding time window. The pilot signal represents the signal at each moment within the sliding time window. This indicates the signal received within the sliding time window. This represents the smoothing factor.

[0028] These features comprehensively reflect the energy stability, noise level, and temporal correlation of the signal at the current moment, and are the core inputs for subsequent signal-to-noise ratio estimation.

[0029] In a specific embodiment, step S3 specifically includes: inputting the above-mentioned multidimensional statistical features into a lightweight multilayer perceptron network (MLP-SNR), and obtaining the predicted value of the signal-to-noise ratio (SNR) at each time step through nonlinear mapping.

[0030] The MLP-SNR module employs a two-layer hidden layer structure with ReLU activation. This module features a small number of parameters, low computational complexity, and strong real-time performance, enabling window-level signal-to-noise ratio estimation. Mathematically, it is expressed as: in, for Multidimensional statistical characteristics of time. This represents the corresponding signal-to-noise ratio (SNR). The SNR indicates the confidence level of the current signal sample: the higher the SNR, the stronger the signal confidence; conversely, the lower the SNR, the lower the confidence level when noise dominates.

[0031] In addition, convolutional neural networks (CNNs) or temporal convolutional networks (TCNs) can be used to replace MLP-SNR to enhance the ability to extract local temporal features; or statistical signal-to-noise ratio estimation methods based on Kalman filtering can be used to reduce network dependence.

[0032] In one specific embodiment, step S4 specifically includes: To achieve signal-to-noise ratio-driven adaptive memory modulation, this application adds an MLP-FiLM network before the LSTM network to generate two characteristic linear modulation parameters based on the current signal-to-noise ratio: Specifically, the signal-to-noise ratio is processed through two fully connected layers. The first layer outputs: Second layer output: in, This is the feature scaling factor (controlling the memory strength). These are the feature offset coefficients (control state offset). These two parameters are used to linearly modulate the hidden state of the bidirectional LSTM network. For MLP-FiLM networks by arrive The weight matrix, for The output after the first fully connected layer For MLP-FiLM networks by arrive The weight matrix, This represents the activation function. For MLP-FiLM networks, the signal-to-noise ratio arrive The weight matrix, , , All are biased. Let be the signal-to-noise ratio at time t.

[0033] Furthermore, gated attention mechanisms (such as the Squeeze-and-Excitation module or self-attention layers) can be used to implement feature weighting, and... , The modulation is equivalent; or a parameter-shared Sigmoid control function can be used to replace affine modulation to achieve a similar signal-to-noise ratio adaptive effect.

[0034] In one specific embodiment, step S5 specifically includes: In the temporal feature extraction stage, this application employs a bidirectional LSTM network to learn the temporal dependency characteristics of the signal simultaneously from both the forward and backward directions, obtaining the following results: and ; The forward temporal hidden feature at time t, The backward temporal hidden feature is at time t.

[0035] The concatenated temporal hidden features are obtained by splicing the two. : Then use the data generated in the previous step and The hidden state is FiLM modulated to obtain the modulated hidden features. : This enhances the memory writing of current features at high signal-to-noise ratios (SNR) and reduces the impact of noise interference at low SNR. In this way, the LSTM state update changes from "fixed gating" to "SNR-driven adaptive gating," enhancing the model's dynamic adaptability to time-varying channels.

[0036] when When the value is greater than 1, the network enhances current memory writes under high signal-to-noise ratio; when When the signal-to-noise ratio (SNR) is less than 1, noise propagation is suppressed under low SNR conditions. This design enables the network to have dynamic channel awareness, allowing it to adaptively adjust the timing memory update rate according to the SNR, thus achieving SNR-level memory writing.

[0037] In addition, GRU (Gated Cyclic Unit) or Transformer encoder bidirectional LSTM networks can be used to implement time-dependent modeling with different structures; when hardware resources are limited, unidirectional LSTM combined with delay compensation modules can also be used to approximate bidirectional information fusion.

[0038] In one specific embodiment, step S6 specifically includes: Hidden features after FiLM modulation The input is fed into a fully connected linear mapping layer (FC), which maps it to the target channel dimension: in, Let be the final channel estimate at time t. The weight matrix is ​​the linear mapping. This is the bias of the linear mapping.

[0039] This represents the real and imaginary parts of the frequency domain channel estimate. The final output channel estimate is... With actual channel The error between them is optimized through backpropagation using normalized mean square error (NMSE): This output layer implements the mapping from time-series features to frequency-domain channel response, enabling the entire network to learn the nonlinear relationship from the input signal to the channel estimate end-to-end.

[0040] Furthermore, fully connected layers (FC) can be replaced by 1D convolutional layers or attention fusion layers to enhance frequency domain smoothness and multi-subcarrier correlation modeling.

[0041] This application has the following significant advantages: Highly adaptive: Traditional LSTM treats high-noise and low-noise moments equally, which can easily lead to noise accumulation. This application uses FiLM modulation to dynamically adjust the update intensity of memory cells according to the signal-to-noise ratio, enabling the model to strengthen learning at high signal-to-noise ratios and suppress noise propagation at low signal-to-noise ratios.

[0042] High generalization ability: By introducing the signal-to-noise ratio as a control factor, the model can automatically adapt to different scenarios and different signal-to-noise ratio ranges, thereby maintaining stable performance in various complex channel environments.

[0043] Lightweight structure: Two small-scale MLP networks are used to complete the signal-to-noise ratio estimation and FiLM parameter generation, which significantly reduces the computational complexity and number of parameters of the model, making it suitable for real-time deployment on terminal devices.

[0044] Robustness enhancement: By introducing statistical features as prior input, the network's ability to perceive noise changes is enhanced; bidirectional LSTM captures temporal dependency information, making the estimation results smoother and more stable.

[0045] Based on the same inventive concept, this application also provides a system for implementing the channel estimation method based on adaptive signal-to-noise ratio modulation and bidirectional LSTM network described above. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more channel estimation system embodiments based on adaptive signal-to-noise ratio modulation and bidirectional LSTM network provided below can be found in the limitations of the channel estimation method based on adaptive signal-to-noise ratio modulation and bidirectional LSTM network described above, and will not be repeated here.

[0046] In one exemplary embodiment, a channel estimation system based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network is provided, comprising the following units.

[0047] The preliminary channel estimation unit is used to perform preliminary channel estimation based on the received signal and pilot signal using the least squares method to obtain coarse channel estimates at each time point.

[0048] The statistical analysis unit is used to perform statistical analysis on the coarse channel estimate and the residual signal to obtain multidimensional statistical characteristics.

[0049] The signal-to-noise ratio prediction unit is used to input the multidimensional statistical features into the multilayer perceptron network to predict the signal-to-noise ratio at each time step.

[0050] The characteristic linear modulation parameter generation unit is used to input the signal-to-noise ratio into the MLP-FiLM network and generate the corresponding characteristic linear modulation parameters; the characteristic linear modulation parameters include characteristic scaling coefficients and characteristic offset coefficients.

[0051] The modulated hidden feature generation unit is used to input the received signal into the bidirectional LSTM network and perform an affine transformation on the hidden state of the bidirectional LSTM network using the feature linear modulation parameters to obtain the modulated hidden features.

[0052] The final channel estimation unit is used to input the modulated hidden features into the fully connected layer and output the final channel estimation value through linear mapping.

[0053] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments. The computer device may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, memory, and I / O are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device stores data to be processed. The I / O interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with an external terminal via a network connection. When the computer program is executed by the processor, it implements the steps in the above-described method embodiments.

[0054] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0055] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0056] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0057] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0058] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A channel estimation method based on SNR adaptive modulation and bidirectional LSTM network, characterized in that, The method comprises the following steps: Based on the received signal and the pilot signal, a preliminary channel estimation is performed using the least squares method to obtain coarse channel estimation values at each time; Statistical analysis is performed on the coarse channel estimation values and the residual signal to obtain multi-dimensional statistical features; The multi-dimensional statistical features are input into a multi-layer perception network to predict the signal-to-noise ratio at each time; The signal-to-noise ratio is input into an MLP-FiLM network to generate corresponding feature linear modulation parameters; The feature linear modulation parameters include feature scaling coefficients and feature offset coefficients; The received signal is input into a bidirectional LSTM network, and the feature linear modulation parameters are used to perform affine transformation on the hidden state of the bidirectional LSTM network to obtain modulated hidden features; The modulated hidden features are input into a fully connected layer to output final channel estimation values through linear mapping.

2. The channel estimation method based on SNR adaptive modulation and bidirectional LSTM network according to claim 1, characterized in that, Based on the received signal and the pilot signal, a preliminary channel estimation is performed using the least squares method to obtain coarse channel estimation values at each time, specifically comprising: Based on the received signal and the pilot signal, a target function is constructed using the least squares criterion; Solving the target function, the coarse channel estimation values at each time are obtained. 3.The channel estimation method based on SNR adaptive modulation and bidirectional LSTM network according to claim 1, wherein, The multi-dimensional statistical features include residual energy mean, residual energy variance, residual modulus median, estimated power, adjacent time difference energy, pilot power, residual ratio, and residual sliding average.

4. The channel estimation method based on SNR adaptive modulation and bidirectional LSTM network according to claim 1, characterized in that, The MLP-SNR module adopts a two-layer hidden layer structure, and the activation function is ReLU.

5. The channel estimation method based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network according to claim 1, wherein the expression of the feature linear modulation parameter is: wherein, is the feature scaling coefficient at time t, is the feature bias coefficient at time t, is the weight matrix from to in the MLP-FiLM network, is the output after the first fully connected layer, is the output after the first fully connected layer, is the weight matrix from to in the MLP-FiLM network, (•) denotes an activation function, is the weight matrix from the signal-to-noise ratio to in the MLP-FiLM network, , , are biases, is the signal-to-noise ratio at time t.

6. The channel estimation method based on SNR adaptive modulation and bidirectional LSTM network according to claim 5, characterized in that, The expression of the modulated hidden feature is: wherein, is the hidden feature modulated at time t, is the time-series hidden feature spliced at time t, is the forward time-series hidden feature at time t, is the backward time-series hidden feature at time t.

7. The channel estimation method based on SNR adaptive modulation and bidirectional LSTM network according to claim 6, characterized in that, The expression of the final channel estimation value is: wherein, is the final channel estimate value at time t, is a weight matrix of the linear mapping, is a bias of the linear mapping. 8.A channel estimation system based on SNR adaptive modulation and bidirectional LSTM network, characterized in that, The method comprises the following steps: A preliminary channel estimation unit is configured to perform a preliminary channel estimation based on a received signal and a pilot signal using the least squares method to obtain coarse channel estimation values at each time; A statistical analysis unit is configured to perform statistical analysis on the coarse channel estimation values and the residual signal to obtain multi-dimensional statistical features; A signal-to-noise ratio prediction unit is configured to input the multi-dimensional statistical features into a multi-layer perception network to predict the signal-to-noise ratio at each time; A feature linear modulation parameter generation unit is configured to input the signal-to-noise ratio into an MLP-FiLM network to generate corresponding feature linear modulation parameters; The feature linear modulation parameters include feature scaling coefficients and feature offset coefficients; A modulated hidden feature generation unit is configured to input the received signal into a bidirectional LSTM network and perform affine transformation on the hidden state of the bidirectional LSTM network using the feature linear modulation parameters to obtain modulated hidden features; A final channel estimation value determination unit is configured to input the modulated hidden features into a fully connected layer to output final channel estimation values through linear mapping.

9. A computer device comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the channel estimation method based on signal-to-noise ratio adaptive modulation and bidirectional LSTM network according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the channel estimation method based on the signal-to-noise ratio adaptive modulation and the bidirectional LSTM network according to any one of claims 1-7.

Images