Channel response model training method and channel estimation method

Abstract

The invention provides a channel response model training method and a channel estimation method, and relates to the technical field of communication. The channel response model training method comprises the following steps: acquiring a training sample set; the training sample set comprises a plurality of training samples, and each training sample comprises an initial channel estimation result of a training communication link and a real channel response label of the training communication link; training a preset neural network based on the training sample, and updating parameters in the preset neural network based on the dynamic learning rate in the training process to obtain a pre-trained channel response model; the dynamic learning rate is negatively correlated with a dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used for representing the signal-to-noise ratio of the training communication link; the channel response model is used for predicting a channel response result according to the input initial channel estimation result. The method is used for improving the adaptability and performance of the channel response model.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a channel response model training method and a channel estimation method. Background Art

[0002] With the rapid development of wireless communication technology, channel estimation, as one of the key technologies to ensure communication quality, has received widespread attention.

[0003] A channel estimation method using a neural network model for channel estimation is proposed in the related art.

[0004] However, the adaptability and performance of neural network models in related technologies need to be improved. Summary of the Invention

[0005] The present application provides a channel response model training method and a channel estimation method, which can improve the adaptability and performance of the channel response model.

[0006] In a first aspect, the present application provides a channel response model training method, the method comprising: obtaining a training sample set; the training sample set comprises multiple training samples, each training sample comprises an initial channel estimation result of a training communication link and a true channel response label of the training communication link; training a preset neural network based on the training samples, and updating the parameters in the preset neural network based on a dynamic learning rate during the training process to obtain a pre-trained channel response model; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; the channel response model is used to predict the channel response result based on the input initial channel estimation result.

[0007] Optionally, a preset neural network is trained based on training samples, and parameters in the preset neural network are updated based on a dynamic learning rate during the training process to obtain a pre-trained channel response model, including: performing a first operation: inputting the training sample into the preset neural network to obtain a predicted channel response label output by the preset neural network; performing a second operation: calculating a loss function value based on the predicted channel response label and the actual channel response label in the training sample; performing a third operation: updating parameters in the model based on the dynamic learning rate and the loss function value; repeating the first operation, the second operation, and the third operation until the preset neural network reaches a first convergence condition to obtain a pre-trained channel response model.

[0008] Optionally, the dynamic learning rate and the dynamic signal-to-noise ratio factor satisfy the following relationship:

[0009]

[0010] Among them, α(t) represents the dynamic learning rate at time t; α0 represents the initial learning rate; m represents the first adjustment coefficient, which is used to control the influence of the dynamic signal-to-noise ratio factor on the dynamic learning rate; SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t.

[0011] Optionally, the dynamic signal-to-noise ratio factor and the signal-to-noise ratio of the training communication link satisfy the following relationship:

[0012]

[0013] Among them, SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t; SNR(t) represents the signal-to-noise ratio of the training communication link at time t; n represents the second adjustment coefficient, which is used to control the influence of the convergence state of the preset neural network on the dynamic signal-to-noise ratio factor; T represents the number of time steps contained in the time window selected forward when calculating the difference between the mean square errors; MSE(ti) represents the mean square error of the preset neural network at time ti; MSE(ti-1) represents the mean square error of the preset neural network at time ti-1.

[0014] It should be understood that in actual communications, the signal-to-noise ratio (SNR) of a channel may fluctuate dramatically over time or in different environments. A fixed learning rate will result in a bias towards training samples with high SNRs while ignoring the learning of training samples with low SNRs.

[0015] The channel response model training method provided in the present application can train a preset neural network based on training samples, and update the parameters in the preset neural network based on the dynamic learning rate during the training process to obtain a pre-trained channel response model. Since the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link, and the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link, in this way, when facing training samples under high signal-to-noise ratio, a smaller learning rate can be used to suppress the overfitting of the preset neural network to clean data and retain the generalization ability to the noisy environment. When facing training samples under low signal-to-noise ratio, a larger learning rate can be used to enhance the sensitivity of the model to noise data and force the preset neural network to learn anti-noise features, so that the preset neural network can effectively learn under high and low signal-to-noise ratios and can adapt to dynamic channel conditions, thereby improving the adaptability and performance of the channel response model.

[0016] In a second aspect, the present application provides a channel response model training device, which includes various functional modules used in the method described in the above first method.

[0017] In a third aspect, the present application provides a channel estimation method, which includes: obtaining a preprocessed received signal, and performing initial channel estimation on the received signal based on a first channel estimation algorithm of compressed sensing to obtain an initial channel estimation result; inputting the initial channel estimation result into a channel response model to obtain an intermediate channel estimation result output by the channel response model; wherein the channel response model is obtained by training a preset neural network based on training samples in a training sample set, and updating the parameters in the preset neural network based on a dynamic learning rate during the training process; the training samples include the channel estimation results of the training communication link and the true channel response labels of the training communication link; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; the intermediate channel estimation result is optimized according to a second channel estimation algorithm based on compressed sensing to determine the final channel estimation result.

[0018] Optionally, performing initial channel estimation on the received signal based on a first channel estimation algorithm using compressed sensing to obtain an initial channel estimation result includes: establishing the following objective function based on the received signal:

[0019]

[0020] Where z(t) represents the received signal; Φ represents the measurement matrix; h(t) represents the channel response estimation result; λ represents the Lagrange multiplier; || ||1 represents the L1 norm. Indicates the square of the L2 norm; It means to find among all possible h(t) such that h(t) reaches its minimum value;

[0021] The optimization problem is solved using the gradient descent method according to the following formula until the second convergence condition is reached, and the current channel estimation result is used as the initial channel estimation result:

[0022] h k+1 =h k -η(-Φ T (z(t)-Φh k )+λ·sign(h k ));

[0023] Among them, h k represents h(t) optimized at the kth iteration, where k is a natural number; h k+1 represents h(t) optimized at the k+1th iteration; η represents the first learning rate of iterative optimization; sign represents the sign function;

[0024] The second convergence condition satisfies the following relationship:

[0025] ||hk+1 -h k ||2<∈;

[0026] Among them, || ||2 means finding the L2 norm; ∈ represents the first convergence threshold.

[0027] Optionally, optimizing the intermediate channel estimation result according to the second channel estimation algorithm based on compressed sensing to determine the final channel estimation result includes iteratively optimizing the intermediate channel estimation result according to the following formula until a third convergence condition is reached, and using the current channel estimation result as the final channel estimation result:

[0028]

[0029] in, represents the channel estimation result of the k+1th iterative optimization; represents the channel estimation result of the k+1th iterative optimization, k is a natural number, and the channel estimation result of the 0th iterative optimization is the intermediate channel estimation result; η p represents the second learning rate of iterative optimization; Φ represents the measurement matrix; z(t) represents the received signal after preprocessing; λ p represents the regularization parameter; sign represents the sign function;

[0030] The third convergence condition satisfies the following relationship:

[0031]

[0032] Among them, || ||2 means finding the L2 norm; ∈ p Represents the second convergence threshold.

[0033] In a fourth aspect, the present application provides a channel estimation device, which includes various functional modules used in the method described in the third aspect above.

[0034] In a fifth aspect, the present application provides a computer program product, comprising: computer instructions; when the computer instructions are executed on an electronic device, the electronic device implements the method described in the first aspect above.

[0035] In a sixth aspect, the present application provides an electronic device comprising: a processor and a memory; the memory stores instructions executable by the processor; when the processor is configured to execute the instructions, the electronic device implements the method described in the first aspect above.

[0036] In a seventh aspect, the present application provides a readable storage medium, which includes: software instructions; when the software instructions are executed in an electronic device, the electronic device implements the method described in the first aspect above.

[0037] The beneficial effects of the second to seventh aspects mentioned above can be referred to the first aspect and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0039] Figure 1 A schematic diagram of the composition of a channel estimation system provided in an embodiment of the present application;

[0040] Figure 2 A flow chart of a channel response model training method provided in an embodiment of the present application;

[0041] Figure 3 A schematic diagram of a flow chart of a channel estimation method provided in an embodiment of the present application;

[0042] Figure 4 A schematic diagram comparing the MSE of the algorithms under different SNRs provided in the embodiment of the present application;

[0043] Figure 5 A schematic diagram of the composition of a channel response model training device provided in an embodiment of the present application;

[0044] Figure 6 A schematic diagram of the composition of a channel estimation device provided in an embodiment of the present application;

[0045] Figure 7 A schematic diagram of the composition of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0047] It should be noted that in the embodiments of this application, words such as "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described in the embodiments of this application as "exemplarily" or "for example" should not be interpreted as being more preferred or advantageous than other embodiments or designs. Rather, the use of words such as "exemplarily" or "for example" is intended to present the relevant concepts in a concrete manner.

[0048] In order to facilitate a clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with basically the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order.

[0049] With the rapid development of wireless communication technology, channel estimation, as one of the key technologies to ensure communication quality, has received widespread attention.

[0050] A channel estimation method using a neural network model for channel estimation is proposed in the related art.

[0051] However, the adaptability and performance of neural network models in related technologies need to be improved.

[0052] Based on this, the embodiments of the present application provide a channel response model training method and a channel estimation method, which can use a dynamic learning rate related to the signal-to-noise ratio to train the neural network model, thereby improving the adaptability and performance of the neural network model.

[0053] The following is an introduction with reference to the accompanying drawings.

[0054] Figure 1 Schematic diagram of the composition of the channel estimation system provided in the embodiment of the present application. Figure 1 As shown, the system may include: a channel response model training apparatus 100, a terminal device 200, and a base station 300. The terminal device 200 is in communication with the channel response model training apparatus 100 and the base station 300. The base station 300 is in communication with the channel response model training apparatus 100.

[0055] The channel response model training device 100 may be an electronic device with computing and processing capabilities, such as a computer or a server.

[0056] Among them, the server can be a single server, or it can be a server cluster composed of multiple servers. In some implementations, the server cluster can also be a distributed cluster. Optionally, the server can also be implemented on a cloud platform. For example, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an inter-cloud, and a multi-cloud, etc., or any combination thereof. The embodiments of the present application are not limited to this.

[0057] The channel response model training device 100 can train the channel response model according to the channel response model training method provided in the following method embodiment. The specific process can be referred to as described in the following embodiment and will not be repeated here.

[0058] In some embodiments, the channel response model training apparatus 100 may also send the pre-trained channel response model to the terminal device 200 and the base station 300 .

[0059] The terminal device 200 may be a mobile phone, a tablet computer, a wearable device, an in-vehicle device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc. The embodiment of the present application does not limit the specific form of the terminal device 200.

[0060] The terminal device 200 can perform channel estimation according to the channel estimation method provided in the following method embodiment. The specific process can be referred to as described in the following embodiment and will not be repeated here.

[0061] The base station 300 may be a base station (BS), a base transceiver station (BTS), a 3G base station (NodeB), a 4G base station (evolved NodeB, eNB), a 5G base station (next generation NodeB, gNB), etc. The embodiments of the present application do not limit the specific form of the base station 300.

[0062] The base station 300 may perform channel estimation according to the channel estimation method provided in the following method embodiment. The specific process may refer to that described in the following embodiment and will not be described again here.

[0063] It should be noted that the above description uses the channel response model training device 100, the terminal device 200, and the base station 300 as independent devices as an example. Optionally, the channel response model training device 100 can also be integrated with the terminal device 200 or the base station 300. That is, the channel response model training device 100 or its corresponding functions can be integrated into the terminal device 200, or the channel response model training device 100 or its corresponding functions can also be integrated into the base station 300. This embodiment of the present application is not limited to this.

[0064] The following first introduces the channel response model training method provided in the embodiment of the present application.

[0065] The executor of the channel response model training method provided in the embodiment of the present application is a channel response model training device (such as the above-mentioned channel response model training device 100). As mentioned above, the channel response model training device can be an electronic device with computing and processing functions such as a computer or a server. Optionally, the channel response model training device can also be a processor (such as a central processing unit (CPU)) in the aforementioned electronic device; or, the channel response model training device can also be a software system or platform in the aforementioned electronic device; or, the channel response model training device can also be an application (application, APP) with a channel response model training function in the aforementioned electronic device; or, the channel response model training device can also be a functional module in the aforementioned electronic device for executing the channel response model training method, etc. The embodiment of the present application does not limit this.

[0066] For the sake of simplicity, the following description will be given by taking the execution entity of the channel response model training method provided in the embodiment of the present application as a channel response model training device as an example.

[0067] Figure 2 Schematic diagram of the flow of the channel response model training method provided in the embodiment of the present application. Figure 2 As shown, the method includes the following steps:

[0068] S101: Obtain a training sample set.

[0069] The training sample set includes multiple training samples, each of which includes an initial channel estimation result of a training communication link and a true channel response label of the training communication link. The initial channel estimation result can be specifically described in the channel estimation method provided in the following method embodiment, and will not be repeated here.

[0070] S102: Training a preset neural network based on the training samples, and updating parameters in the preset neural network based on a dynamic learning rate during the training process to obtain a pre-trained channel response model.

[0071] The dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link, and the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link. The channel response model is used to predict the channel response result based on the input initial channel estimation result.

[0072] In some possible embodiments, the above S102 may specifically include the following steps:

[0073] Step 1a: Execute the first operation: input the training sample into the preset neural network to obtain the predicted channel response label output by the preset neural network.

[0074] As an example, the preset neural network may adopt a convolutional neural network (CNN) structure. The CNN structure may include an input layer, multiple convolutional layers, a pooling layer, and a fully connected layer. The specific structure is as follows:

[0075] (1) Input layer, used to receive the initial channel estimation results in the training samples.

[0076] (2) Convolutional layer, used to extract features from the initial channel estimation results and activated using the ReLU activation function.

[0077] (3) Pooling layer, used for dimensionality reduction and reducing computational complexity.

[0078] (4) Fully connected layer, used to map the extracted features to the optimized channel response results.

[0079] (5) Output layer, used to output channel response results.

[0080] Step 2a: Perform the second operation: calculate the loss function value based on the predicted channel response label and the true channel response label in the training sample.

[0081] As an example, the channel response model training device may use mean squared error (MSE) as a loss function.

[0082] For example, MSE can be specifically expressed as the following formula:

[0083]

[0084] In formula (1), MSE represents the mean square error, which is also the loss function value. N represents the number of training samples. Denotes the predicted channel response label. h(t) denotes the true channel response label.

[0085] Step 3a: Perform the third operation: update the parameters in the model based on the dynamic learning rate and loss function value.

[0086] As an example, the dynamic learning rate and the dynamic signal-to-noise ratio factor satisfy the following relationship:

[0087]

[0088] In formula (2), α(t) represents the dynamic learning rate at time t; α0 represents the initial learning rate; m represents the first adjustment coefficient, which is used to control the influence of the dynamic signal-to-noise ratio factor on the dynamic learning rate; SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t.

[0089] As an example, the dynamic signal-to-noise ratio factor and the signal-to-noise ratio of the training communication link satisfy the following relationship:

[0090]

[0091] In formula (3), SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t. The significance of this factor is that when the signal-to-noise ratio is higher and the neural network converges more stably (that is, the mean square error changes less), the SNR dy The larger the value of (t), the smaller its value will be. Conversely, when the signal-to-noise ratio is low or the neural network is unstable, its value will respond accordingly. SNR(t) represents the signal-to-noise ratio of the training communication link at time t. n represents the second adjustment coefficient, which is used to control the influence of the convergence state of the preset neural network on the dynamic signal-to-noise ratio factor. t represents the number of time steps contained in the time window selected forward when calculating the difference between the mean square errors. MSE(ti) represents the mean square error of the preset neural network at time ti. MSE(ti-1) represents the mean square error of the preset neural network at time ti-1.

[0092] Step 4a: Repeat the first operation, the second operation, and the third operation until the preset neural network reaches a first convergence condition, thereby obtaining a pre-trained channel response model.

[0093] Optionally, the first convergence condition may specifically include: the number of times the training sample is input into the preset neural network reaches a preset number threshold and / or the error of the preset neural network is less than or equal to an error threshold.

[0094] The number threshold can be preset in the channel response model training device. For example, the number threshold can be set to 500, 1000, 2000, or 5000 times. This embodiment of the application does not limit the specific value of the number threshold. The error threshold can also be preset in the channel response model training device. For example, the error threshold can be set to 5%, 10%, or 15%. This embodiment of the application does not limit the specific value of the error threshold.

[0095] It should be understood that in actual communications, the signal-to-noise ratio (SNR) of a channel may fluctuate dramatically over time or in different environments. A fixed learning rate will result in a bias towards training samples with high SNRs while ignoring the learning of training samples with low SNRs.

[0096] In the channel response model training method provided in the embodiment of the present application, the channel response model training device can train the preset neural network based on the training samples, and update the parameters in the preset neural network based on the dynamic learning rate during the training process to obtain a pre-trained channel response model. Since the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link, and the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link, in this way, when facing training samples under high signal-to-noise ratio, a smaller learning rate can be used to suppress the overfitting of the preset neural network to clean data and retain the generalization ability to the noisy environment. When facing training samples under low signal-to-noise ratio, a larger learning rate can be used to enhance the sensitivity of the model to noise data and force the preset neural network to learn anti-noise features, so that the preset neural network can effectively learn under high signal-to-noise ratio and low signal-to-noise ratio, and can adapt to dynamic channel conditions, thereby improving the adaptability and performance of the channel response model.

[0097] The above describes the channel response model training method provided by the embodiment of the present application. Next, the channel estimation method provided by the embodiment of the present application is described.

[0098] The execution subject of the channel estimation method provided in the embodiment of the present application is a channel estimation device, which can be a terminal device (such as the above-mentioned terminal device 200) or a base station (such as the above-mentioned base station 300).

[0099] Figure 3 Schematic diagram of the flow of the channel estimation method provided in the embodiment of the present application. Figure 3 As shown, the method includes the following steps:

[0100] S201 : Acquire a preprocessed received signal, and perform initial channel estimation on the received signal based on a first channel estimation algorithm of compressed sensing to obtain an initial channel estimation result.

[0101] Preprocessing primarily involves signal denoising and normalization. Preprocessing can eliminate noise and interference from the signal, improving the accuracy of subsequent channel estimation. This application proposes a preprocessing algorithm based on adaptive filtering to make the filtered signal as close as possible to the original transmitted signal.

[0102] The specific process is as follows:

[0103] Assuming the received signal y(t), the noise is n(t), and the channel response is h(t), the received signal can be expressed as:

[0104] y(t)=h(t)*x(t)+n(t) Formula (4)

[0105] In formula (4), x(t) represents the transmitted signal and * represents the convolution operation.

[0106] The goal of the adaptive filter is to find a filter w(t) that makes the output signal z(t) as close as possible to the transmitted signal x(t), that is:

[0107] z(t)=w(t)*y(t) Formula (5)

[0108] The filter coefficients are updated by minimizing the error update e(t):

[0109] e(t)=x(t)-z(t) Formula (6)

[0110] Update the filter coefficients using the gradient descent algorithm:

[0111] w(t+1)=w(t)+μe(t)y(t) Formula (7)

[0112] In formula (7), μ represents the learning rate.

[0113] The adaptive filtering algorithm provided in the embodiments of the present application can adaptively adjust the filter coefficients according to the received signal under different signal-to-noise ratio conditions to minimize the error between the filtered output signal and the transmitted signal, thereby effectively reducing noise and improving signal quality.

[0114] In some embodiments, performing initial channel estimation on the received signal using a first channel estimation algorithm based on compressed sensing in S201 to obtain an initial channel estimation result may specifically include the following steps:

[0115] Step 1b: Based on the received signal, establish the following objective function:

[0116]

[0117] In formula (8), z(t) represents the received signal; Φ represents the measurement matrix; h(t) represents the channel response estimation result; λ represents the Lagrange multiplier; || ||1 represents the L1 norm; Indicates the square of the L2 norm; It means to find among all possible h(t) such that h(t) reaches its minimum value.

[0118] As an example, before the above step 1b, the channel estimation apparatus may first establish a channel estimation problem.

[0119] From the preprocessed received signal z(t) obtained from the above preprocessing process, assuming that the channel response h(t) can be expressed as a sparse vector, the channel estimation problem can be expressed as:

[0120] z(t)=Φh(t)+n(t) Formula (9)

[0121] In formula (9), Φ represents the measurement matrix.

[0122] The goal of channel estimation is to find a sparse vector h(t) that minimizes the measurement error ∈:

[0123] min||h(t)||1 bject to||z(t)-Φh(t)||2≤∈ Formula (10)

[0124] In formula (10), ∈ represents the first convergence threshold.

[0125] The channel estimation device can then solve the optimization problem using the Basis Pursuit method. Basis Pursuit is a compressed sensing algorithm based on convex optimization that seeks a sparse solution by minimizing the L1 norm of the channel response vector. Its goal is to find a sparse vector h(t) that minimizes the measurement error within an allowable range.

[0126] For example, the channel estimation device may use the Lagrange multiplier method to convert the constrained problem into an unconstrained problem using the objective function of the above formula (10). The converted objective function is shown in the above formula (8).

[0127] Step 2b: Use the gradient descent method to solve the optimization problem according to the following formula until the second convergence condition is reached, and use the current channel estimation result as the initial channel estimation result:

[0128] h k+1 =h k -η(-Φ T (z(t)-Φh k )+λ·sign(h k )) Formula (11)

[0129] In formula (9), h k represents h(t) optimized at the kth iteration, where k is a natural number; h k+1 represents h(t) of the k+1th iterative optimization; η represents the first learning rate of iterative optimization; sign represents the sign function; the second convergence condition satisfies the following relationship:

[0130] ||h k+1 -h k ||2<∈ Formula (12)

[0131] Among them, || ||2 means finding the L2 norm; ∈ represents the first convergence threshold.

[0132] S202: Input the initial channel estimation result into a channel response model to obtain an intermediate channel estimation result output by the channel response model.

[0133] The channel response model is obtained by training a preset neural network based on training samples in a training sample set and updating the parameters of the preset neural network based on a dynamic learning rate during the training process; the training samples include the channel estimation results of the training communication link and the true channel response label of the training communication link; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; and the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link. The specific training process of the channel response model can be referred to the description in the above embodiment and will not be repeated here.

[0134] S203 : Optimize the intermediate channel estimation result according to the second channel estimation algorithm based on compressed sensing to determine a final channel estimation result.

[0135] In some embodiments, the above S203 may specifically include the following steps:

[0136] Step 1c: Iteratively optimize the intermediate channel estimation result according to the following formula until the third convergence condition is reached, and use the current channel estimation result as the final channel estimation result:

[0137]

[0138] In formula (13), represents the channel estimation result of the k+1th iterative optimization; represents the channel estimation result of the k+1th iterative optimization, k is a natural number, and the channel estimation result of the 0th iterative optimization is the intermediate channel estimation result; η p represents the second learning rate of iterative optimization; Φ represents the measurement matrix; z(t) represents the received signal after preprocessing; λ p represents the regularization parameter; sign represents the sign function; the third convergence condition satisfies the following relationship:

[0139]

[0140] In formula (14), || ||2 means finding the L2 norm; ∈ p Represents the second convergence threshold.

[0141] After the iteration, the channel evaluation device can output the final channel estimation result, which combines the advantages of deep learning model and iterative optimization, and has high estimation accuracy and robustness under low signal-to-noise ratio conditions.

[0142] In an exemplary embodiment, a series of comparative experiments were conducted to verify the performance of the channel estimation algorithm provided in the present embodiment under different signal-to-noise ratio (SNR) conditions. Several commonly used traditional channel estimation algorithms were selected for comparison, including the least squares (LS) method and the least mean square error (LMMSE) method. Different SNR values were used in the experiment, and the MSE performance index of each algorithm was recorded.

[0143] The experimental environment is:

[0144] Number of transmitting antennas: 8;

[0145] Number of receiving antennas: 8

[0146] Modulation method: Binary Phase-Shift Keying (BPSK);

[0147] Channel model: Multiple-Input Multiple-Output (MIMO);

[0148] Signal-to-noise ratio range: 0dB to 40dB, interval 10dB;

[0149] Test data: Run 10,000 simulations at each SNR to obtain the mean squared error.

[0150] For example, Figure 4 The MSE comparison diagram of the algorithms under different SNRs provided in the embodiment of this application is shown in FIG. Figure 4 As shown, Figure 4 The MSE comparison results of each algorithm under different SNR are shown. Figure 4 It can be seen that as the SNR increases, the MSE of each algorithm gradually decreases. The MSE of the channel estimation method provided by the embodiment of the application is significantly better than that of the LS and LMMSE algorithms under high SNR conditions, especially when the SNR exceeds 15dB, the advantage is more obvious.

[0151] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to realize the above functions, the channel response model training device and the channel estimation device include hardware structures and / or software modules corresponding to the execution of each function. It should be easy to realize that the technical goals in this field are combined with the units and algorithm steps of each example described in the embodiments disclosed in this article, and the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technical goals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0152] In an exemplary embodiment, the present application also provides a channel response model training device in the form of a virtual device. Figure 5 This is a schematic diagram of the composition of the channel response model training device provided in the embodiment of the present application. Figure 5 As shown, the channel response model training device includes: a first acquisition module 501 and a first processing module 502.

[0153] The first acquisition module 501 is used to acquire a training sample set; the training sample set includes multiple training samples, and each training sample includes a channel estimation result of a training communication link and a true channel response label of the training communication link.

[0154] The first processing module 502 is used to train a preset neural network based on training samples, and update the parameters in the preset neural network based on a dynamic learning rate during the training process to obtain a pre-trained channel response model; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; the channel response model is used to predict the channel response result based on the input channel estimation result.

[0155] In some embodiments, the first processing module 502 is specifically used to perform a first operation: input the training sample into a preset neural network to obtain a predicted channel response label output by the preset neural network; perform a second operation: calculate the loss function value based on the predicted channel response label and the actual channel response label in the training sample; perform a third operation: update the parameters in the model based on the dynamic learning rate and the loss function value; repeat the first operation, the second operation, and the third operation until the preset neural network reaches a first convergence condition to obtain a pre-trained channel response model.

[0156] In some embodiments, the dynamic learning rate and the dynamic signal-to-noise ratio factor satisfy the following relationship:

[0157]

[0158] Among them, α(t) represents the dynamic learning rate at time t; α0 represents the initial learning rate; m represents the first adjustment coefficient, which is used to control the influence of the dynamic signal-to-noise ratio factor on the dynamic learning rate; SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t.

[0159] In some embodiments, the dynamic signal-to-noise ratio factor and the signal-to-noise ratio of the training communication link satisfy the following relationship:

[0160]

[0161] Among them, SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t; SNR(t) represents the signal-to-noise ratio of the training communication link at time t; n represents the second adjustment coefficient, which is used to control the influence of the convergence state of the preset neural network on the dynamic signal-to-noise ratio factor; T represents the number of time steps contained in the time window selected forward when calculating the difference between the mean square errors; MSE(ti) represents the mean square error of the preset neural network at time ti; MSE(ti-1) represents the mean square error of the preset neural network at time ti-1.

[0162] In an exemplary embodiment, the present application also provides a channel estimation device in the form of a virtual device. Figure 6 Schematic diagram of the composition of the channel estimation device provided in the embodiment of the present application. Figure 6 As shown, the channel estimation device includes: a second acquisition module 601 and a second processing module 602.

[0163] The second acquisition module 601 is configured to acquire the preprocessed received signal and perform initial channel estimation on the received signal based on a first channel estimation algorithm of compressed sensing to obtain an initial channel estimation result;

[0164] The second processing module 602 is used to input the initial channel estimation result into the channel response model to obtain an intermediate channel estimation result output by the channel response model; wherein the channel response model is obtained by training a preset neural network based on training samples in the training sample set, and updating the parameters in the preset neural network based on a dynamic learning rate during the training process; the training samples include the channel estimation result of the training communication link and the true channel response label of the training communication link; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; the intermediate channel estimation result is optimized according to the second channel estimation algorithm based on compressed sensing to determine the final channel estimation result.

[0165] In some embodiments, the second acquisition module 601 is specifically configured to establish the following objective function based on the received signal:

[0166]

[0167] Where z(t) represents the received signal; Φ represents the measurement matrix; h(t) represents the channel response estimation result; λ represents the Lagrange multiplier; || ||1 represents the L1 norm. Indicates the square of the L2 norm; It means to find among all possible h(t) such that h(t) reaches its minimum value;

[0168] The optimization problem is solved using the gradient descent method according to the following formula until the second convergence condition is reached, and the current channel estimation result is used as the initial channel estimation result:

[0169] h k+1 =h k -η(-Φ T (z(t)-Φh k )+λ·sign(h k ));

[0170] Among them, h k represents h(t) optimized at the kth iteration, where k is a natural number; h k+1 represents h(t) of the k+1th iterative optimization; η represents the first learning rate of iterative optimization; sign represents the sign function; the second convergence condition satisfies the following relationship:

[0171] ||h k+1 -h k ||2<∈;

[0172] Among them, || ||2 means finding the L2 norm; ∈ represents the first convergence threshold.

[0173] In some embodiments, the second processing module 602 is specifically configured to iteratively optimize the intermediate channel estimation result according to the following formula until a third convergence condition is reached, and use the current channel estimation result as the final channel estimation result:

[0174]

[0175] in, represents the channel estimation result of the k+1th iterative optimization; represents the channel estimation result of the k+1th iterative optimization, k is a natural number, and the channel estimation result of the 0th iterative optimization is the intermediate channel estimation result; η p represents the second learning rate of iterative optimization; Φ represents the measurement matrix; z(t) represents the received signal after preprocessing; λ p represents the regularization parameter; sign represents the sign function; the third convergence condition satisfies the following relationship:

[0176]

[0177] Among them, || ||2 means finding the L2 norm; ∈ p Represents the second convergence threshold.

[0178] It should be noted that the above Figure 5 and Figure 6 The module division described is illustrative and represents only one logical functional division. Actual implementations may employ different divisions. For example, two or more functions may be integrated into a single processing module. These integrated modules may be implemented as either hardware or software functional modules.

[0179] In an exemplary embodiment, the embodiment of the present application further provides an electronic device, which may be the above-mentioned channel response model training device or channel estimation device. Figure 7 This is a schematic diagram of the composition of the electronic device provided in the embodiment of the present application. Figure 7 As shown, the electronic device includes: a first processor 10 , a memory 20 , a communication line 30 , a communication interface 40 , and an input / output interface 50 .

[0180] The first processor 10 , the memory 20 , the communication interface 40 , and the input / output interface 50 may be connected via a communication line 30 .

[0181] The first processor 10 is used to execute the instructions stored in the memory 20 to implement the channel response model training method or channel estimation method provided in the above embodiments of the present application. The first processor 10 can be a CPU, a general-purpose processor network processor (NP), a digital signal processor (DSP), a microprocessor, a microcontroller (MCU) / single-chip microcomputer / single-chip microcomputer, a programmable logic device (PLD) or any combination thereof. The first processor 10 can also be any other device with processing functions, such as a circuit, a device or a software module, which is not limited in the embodiments of the present application. In one example, the first processor 10 may include one or more CPUs, such as Figure 7 As an optional implementation, the electronic device may include multiple processors, for example, in addition to the first processor 10, it may also include a second processor 60 ( Figure 7 The dashed line is used as an example.

[0182] The memory 20 is used to store instructions. For example, the instruction may be a computer program. Optionally, the memory 20 may be a read-only memory (ROM) or other types of static storage devices that can store static information and / or instructions, or a random access memory (RAM) or other types of dynamic storage devices that can store information and / or instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, etc., and the embodiments of the present application are not limited thereto.

[0183] It should be noted that the memory 20 may exist independently of the first processor 10 or may be integrated with the first processor 10. The memory 20 may be located inside or outside the electronic device, which is not limited in the embodiment of the present application.

[0184] The communication line 30 is used to transmit information between the components included in the electronic device.

[0185] Communication interface 40 is used to communicate with other devices or other communication networks. Such other communication networks may be Ethernet, radio access networks (RAN), wireless local area networks (WLAN), etc. Communication interface 40 may be a module, circuit, transceiver, or any other device capable of communication.

[0186] The input / output interface 50 is used to implement human-computer interaction between a user and the electronic device, for example, to implement action interaction or information interaction between the user and the electronic device.

[0187] For example, the input / output interface 50 may be a mouse, keyboard, display screen, or touch screen screen, etc. Action interaction or information interaction between a user and the electronic device may be achieved through the mouse, keyboard, display screen, or touch screen screen, etc.

[0188] It should be noted that Figure 7 The structure shown in the figure does not constitute a limitation on the electronic device, except Figure 7 In addition to the components shown, the electronic device may include more or fewer components than shown, or a combination of certain components, or a different arrangement of components.

[0189] In an exemplary embodiment, the present application also provides a computer program product, which includes computer instructions. When the computer instructions are executed in an electronic device, the electronic device implements the method in the aforementioned method embodiment.

[0190] In an exemplary embodiment, the present application also provides a computer-readable storage medium including software instructions. When the software instructions are executed in an electronic device, the electronic device implements the method in the aforementioned method embodiment. The computer-readable storage medium can be a non-transitory computer-readable storage medium, for example, a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0191] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using a software program, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer-executable instructions. When the computer-executable instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer-executable instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer-executable instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0192] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple components. A single processor or other unit may implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0193] Although the present application has been described with reference to specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the drawings are merely illustrative of the present application as defined by the appended claims and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, the present application is intended to include such modifications and variations as fall within the scope of the claims of the present application and their equivalents.

[0194] The above is only a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A channel response model training method, characterized in that: The method comprises: Acquire a training sample set; the training sample set includes a plurality of training samples, each of the training samples includes an initial channel estimation result of a training communication link and a true channel response label of the training communication link; A preset neural network is trained based on the training samples, and parameters in the preset neural network are updated based on a dynamic learning rate during the training process to obtain a pre-trained channel response model; the dynamic learning rate is negatively correlated with a dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; and the channel response model is used to predict a channel response result based on an input initial channel estimation result.

2. The method according to claim 1, characterized in that The step of training a preset neural network based on the training samples and updating parameters in the preset neural network based on a dynamic learning rate during the training process to obtain a pre-trained channel response model includes: Performing a first operation: inputting the training sample into the preset neural network to obtain a predicted channel response label output by the preset neural network; Performing a second operation: calculating a loss function value based on the predicted channel response label and the true channel response label in the training sample; Performing a third operation: updating parameters in the model based on the dynamic learning rate and the loss function value; Repeat the first operation, the second operation, and the third operation until the preset neural network reaches a first convergence condition, thereby obtaining a pre-trained channel response model.

3. The method according to claim 2, characterized in that The dynamic learning rate and the dynamic signal-to-noise ratio factor satisfy the following relationship: Among them, α(t) represents the dynamic learning rate at time t; α0 represents the initial learning rate; m represents the first adjustment coefficient, which is used to control the influence of the dynamic signal-to-noise ratio factor on the dynamic learning rate; SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t.

4. The method according to claim 2, characterized in that The dynamic signal-to-noise ratio factor and the signal-to-noise ratio of the training communication link satisfy the following relationship: Among them, SNR dy (t) represents the dynamic signal-to-noise ratio factor at time t; SNR(t) represents the signal-to-noise ratio of the training communication link at time t; n represents the second adjustment coefficient, which is used to control the influence of the convergence state of the preset neural network on the dynamic signal-to-noise ratio factor; T represents the number of time steps contained in the time window selected forward when calculating the difference between the mean square errors; MSE(ti) represents the mean square error of the preset neural network at time ti; MSE(ti-1) represents the mean square error of the preset neural network at time ti-1.

5. A channel estimation method, characterized in that: The method comprises: Acquire the preprocessed received signal, and perform initial channel estimation on the received signal based on a first channel estimation algorithm of compressed sensing to obtain an initial channel estimation result; Inputting the initial channel estimation result into a channel response model to obtain an intermediate channel estimation result output by the channel response model; wherein the channel response model is obtained by training a preset neural network based on training samples in a training sample set, and updating parameters in the preset neural network based on a dynamic learning rate during the training process; the training samples include the channel estimation result of the training communication link and the true channel response label of the training communication link; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; The intermediate channel estimation result is optimized according to a second channel estimation algorithm based on compressed sensing to determine a final channel estimation result.

6. The method according to claim 5, characterized in that The first channel estimation algorithm based on compressed sensing performs initial channel estimation on the received signal to obtain an initial channel estimation result, including: Based on the received signal, the following objective function is established: Wherein, z(t) represents the received signal; Φ represents the measurement matrix; h(t) represents the channel response estimation result; λ represents the Lagrange multiplier; ∥∥1 represents the L1 norm; Indicates finding the square of the L2 norm; It means to find among all possible h(t) such that h(t) reaches its minimum value; The optimization problem is solved using the gradient descent method according to the following formula until the second convergence condition is reached, and the current channel estimation result is used as the initial channel estimation result: h k+1 =h k -η(-Φ T (z(t)-Φh k )+λ·sign(h k )); Among them, h k represents h(t) optimized at the kth iteration, where k is a natural number; h k+1 represents h(t) optimized at the k+1th iteration; η represents the first learning rate of iterative optimization; sign represents the sign function; The second convergence condition satisfies the following relationship: ||h k+1 -h k ||2<∈; Among them, ∥∥2 means finding the L2 norm; ∈ represents the first convergence threshold.

7. The method according to claim 5, characterized in that Optimizing the intermediate channel estimation result according to the second channel estimation algorithm based on compressed sensing to determine the final channel estimation result includes: The intermediate channel estimation result is iteratively optimized according to the following formula until a third convergence condition is reached, and the current channel estimation result is used as the final channel estimation result: in, represents the channel estimation result of the k+1th iterative optimization; represents the channel estimation result of the k+1th iterative optimization, k is a natural number, and the channel estimation result of the 0th iterative optimization is the intermediate channel estimation result; η p represents the second learning rate of iterative optimization; Φ represents the measurement matrix; z(t) represents the received signal after preprocessing; λ p represents the regularization parameter; sign represents the sign function; The third convergence condition satisfies the following relationship: Among them, ∥∥2 means finding the L2 norm; ∈ p Represents the second convergence threshold.

8. A channel response model training device, characterized in that: The device includes: a first acquisition module and a first processing module; The first acquisition module is used to acquire a training sample set; the training sample set includes multiple training samples, each of the training samples includes a channel estimation result of a training communication link and a true channel response label of the training communication link; The first processing module is used to train a preset neural network based on the training samples, and update the parameters in the preset neural network based on a dynamic learning rate during the training process to obtain a pre-trained channel response model; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; the channel response model is used to predict a channel response result based on an input channel estimation result.

9. A channel estimation device, characterized in that The device includes: a second acquisition module and a second processing module; The second acquisition module is used to acquire the preprocessed received signal and perform initial channel estimation on the received signal based on a first channel estimation algorithm of compressed sensing to obtain an initial channel estimation result; The second processing module is used to input the initial channel estimation result into a channel response model to obtain an intermediate channel estimation result output by the channel response model; wherein the channel response model is obtained by training a preset neural network based on training samples in a training sample set, and updating the parameters in the preset neural network based on a dynamic learning rate during the training process; the training samples include the channel estimation result of the training communication link and the true channel response label of the training communication link; the dynamic learning rate is negatively correlated with the dynamic signal-to-noise ratio factor of the training communication link; the dynamic signal-to-noise ratio factor is used to represent the signal-to-noise ratio of the training communication link; the intermediate channel estimation result is optimized according to a second channel estimation algorithm based on compressed sensing to determine a final channel estimation result.

10. An electronic device, characterized in that: include: processor and memory; The memory stores instructions executable by the processor; When the processor is configured to execute the instructions, the electronic device implements the method according to any one of claims 1 to 7.

11. A readable storage medium, characterized in that: include: Software instructions; When the software instructions are executed in an electronic device, the electronic device is enabled to implement the method according to any one of claims 1 to 7.

12. A computer program product, characterized in that include: Computer instructions; When the computer instructions are executed in an electronic device, the electronic device is enabled to implement the method according to any one of claims 1 to 7.

Images