A Joint Estimation Method for Doppler Frequency Offset and Time-Varying Channel

Abstract

The present invention discloses a method for joint estimation of Doppler frequency offset and time-varying channel. Aiming at the problem that Doppler frequency offset (DFO) in a RIS (Reconfigurable Intelligent Surface)-assisted mobile uplink OFDM system will bring great interference to obtaining channel information, the present invention jointly estimates DFO and time-varying channel, fully considering the influence of DFO on channel estimation. In DFO estimation, the present invention uses a Convolutional Neural Network (CNN) to obtain a high-precision DFO estimate to reduce the influence of residual frequency offset on channel estimation. To improve the practicability of the present invention, non-ideal DFO is set as the training target of the network. Based on the time-domain signal after DFO compensation, the present invention adopts a threshold denoising method to obtain a high-precision cascaded channel estimate to reduce the influence of direct channel estimation error and noise. Simulations show that the present invention has a high estimation accuracy and is applicable to the high-precision acquisition of time-varying channel information in a RIS-assisted mobile communication system.

Description

Technical Field

[0001] The present invention relates to a method for jointly estimating Doppler frequency offset and time-varying channel, which is applied to an RIS-assisted OFDM system and belongs to the field of wireless communication technology. Background Art

[0002] Future wireless communication systems will be a system with higher speed, more energy-efficient, more reliable and low latency. Reconfigurable intelligent surface (RIS) is considered to be the most promising technology to achieve this system. RIS is a new type of wireless communication technology. By introducing a large number of programmable passive reflectors, the metasurface composed of these large numbers of passive metamaterials can change the amplitude and phase of the received signal, so as to achieve the desired performance gain. However, the performance gain provided by RIS depends on the accuracy of the channel state information (CSI), and the passivity of the RIS and the high-dimensional cascaded channel make the accurate acquisition of CSI more challenging. Most of the existing methods for obtaining CSI are for static channels. However, the mobility of users makes it difficult to continue using the static channel estimation method. Therefore, it is necessary to study the estimation method of time-varying channels in RIS-assisted systems.

[0003] In recent years, scholars at home and abroad have proposed some time-varying channel estimation methods for RIS-assisted mobile communication systems. Since RIS is generally fixed while users are mobile, Hu Chenxi et al. ("Two-Timescale Channel Estimation for Reconfigurable Intelligent Surface Aided Wireless Communications.") decomposed the channel estimation problem into two time scales, i.e., estimating the quasi-static channel between RIS and the base station on a large time scale and estimating the continuously changing time-varying channel between RIS and users on a small time scale. However, the high-speed movement of users will make the coherence time shorter, which will affect the estimation performance. Huang Zixuan et al. ("Transforming Fading Channel from Fast to Slow: IRS-Assisted High-Mobility Communication.") deployed RIS in moving vehicles. First, the direct link channel between the base station and users was obtained using pilots, and the direct link phase shift was fed back to RIS to adjust the RIS reflection coefficient to compensate for the base station-RIS-user cascaded channel. Finally, the cascaded channel estimation was obtained through the LS method. Vishnu Karthikeya Gorty et al. ("Channel Estimation for Double IRS Assisted Communication for a Mobile User.") used the compressive sensing method to estimate the channels between the base station and two RISs and the Kalman filter to obtain the time-varying channel between RIS and users for a dual RIS system. However, the complexity of this method is relatively high. Chao Xu et al. ("Channel Estimation for Reconfigurable Intelligent Surface Assisted High-Mobility Wireless Systems.") adopted the MMSE method to estimate the time-varying channel information in the RIS-assisted system. However, the performance of this method depends on the prior statistical information of the channel and has a relatively high computational complexity.

[0004] In summary, the existing time-varying channel estimations in RIS-assisted mobile communication systems do not consider the impact of Doppler frequency offset (DFO), that is, it is assumed that precise carrier synchronization is achieved. However, in an actual communication system, due to the existence of DFO estimation errors, there will still be the impact of residual frequency offset in the received signal after frequency offset compensation. This will cause the direct component in the received signal in a Rice channel environment to still have a frequency offset, while the random variation of the scattered component signals will be enhanced, which makes it more challenging to accurately obtain channel estimation in this scenario.

[0005] In view of this, it is necessary to study an efficient estimation method for time-varying channels with residual DFO. Summary of the Invention

[0006] The purpose of the present invention is to provide a joint estimation method for Doppler frequency offset and time-varying channels, which is applied to an RIS-assisted OFDM system. This method can fully consider the time-varying channel estimation in the presence of residual Doppler frequency offset in an RIS-assisted communication system.

[0007] To achieve the above purpose, the present invention provides a joint estimation method for Doppler frequency offset and time-varying channels, which is applied to an RIS-assisted OFDM system. The method includes the following steps:

[0008] Step 1: Obtain the Doppler frequency offset estimation value according to the cyclic prefix of the time-domain received signal;

[0009] Step 2: Construct a training sample set using the time-domain received signal;

[0010] Step 3: Based on randomly initialized network parameters, train the CNN network using the training sample set constructed in Step 2;

[0011] Step 4: Obtain a network model with optimal weight parameters and threshold parameters;

[0012] Step 5: Perform online DFO estimation based on the trained network model;

[0013] Step 6: Perform frequency offset compensation on the time-domain received signal using the DFO estimation;

[0014] Step 7: Perform direct link channel estimation. Assume that during the first OFDM symbol period, all RIS units are turned off, and use the LS method and linear interpolation method to obtain the estimation of the direct link channel

[0015] Step 8: Perform cascaded channel estimation. Turn on the RIS units in sequence. When the m-th RIS unit is turned on and other RIS units are turned off, based on the direct link channel estimation in Step 7 and the LS method, obtain the estimation of the pilot sub-channel in the m-th cascaded channel;

[0016] Step 9: Use the linear interpolation method and the pilot sub-channel estimation in Step 8 to obtain the m-th cascaded channel estimation as

[0017] Step 10: Adopt the threshold denoising method to further improve the accuracy of the cascaded channel estimation;

[0018] Step 11: Perform FFT transformation on the denoised time-domain cascaded channel estimation to obtain the frequency-domain cascaded channel estimation;

[0019] Step 12: Obtain the final full-channel information.

[0020] As a further improvement of the present invention, the formula for calculating the Doppler frequency offset estimation value in Step 1 is as follows:

[0021]

[0022] where arg{·} is the angle-taking operation, r(u,n) represents the n-th time-domain discrete sampling signal of the u-th OFDM symbol received by the base station, N cp is the cyclic prefix length of the OFDM symbol, and N is the length of the OFDM symbol.

[0023] As a further improvement of the present invention, the formula for constructing the training sample set using the time-domain received signal in Step 2 is as follows:

[0024] Τ DFO ={(s (1) ,ε (1) ),...,(s (v) ,ε (v) ),...,(s (V) ,ε (V) )}

[0025] where v is the number of training samples, s (v) is the v-th training input sample composed of the received signals, and respectively represent the real part-taking and imaginary part-taking operations,

[0026] r (v) (u,n) is the received signal in the v-th sample, u is the number of OFDM symbols, ε (v) is the v-th output sample composed of the ideal DFO, is the ideal DFO on the u-th OFDM symbol in the v-th sample. To improve the practicality of the method, the DFO estimation is used as the training target of the network, that is,

[0027]

[0028] wherein, is the v-th sample of the DFO estimate component, is the DFO estimate of the u-th OFDM symbol obtained by using Step 1 in the v-th sample.

[0029] As a further improvement of the present invention, the formula for obtaining the estimate of the pilot sub-channel in the cascaded channel in Step 8 is:

[0030]

[0031] wherein, is the received pilot signal without the direct channel signal, is the u-th frequency-domain pilot signal transmitted, is the received signal on the k-th sub-carrier in the u-th OFDM symbol after compensation, is the direct channel estimate on the k-th sub-carrier in the u-th OFDM symbol, Γ tr is the set of pilot positions on the u-th OFDM symbol.

[0032] As a further improvement of the present invention, Step 10 includes:

[0033] First, perform an N-point IFFT transformation on the initial estimate of the cascaded channel to obtain the time-domain cascaded channel estimate, and then use the threshold denoising method to obtain the denoised time-domain channel estimate as

[0034]

[0035] where n = 0,..., N cp -1, when n ≥ N cp when q is the threshold, that is

[0036]

[0037] As a further improvement of the present invention, the formula for obtaining the frequency-domain cascaded channel estimate in Step 11 is as follows:

[0038]

[0039] wherein, F N is an N×N-dimensional normalized discrete Fourier transform matrix, is a 1×(N - N cp )-dimensional all-zero vector.

[0040] As a further improvement of the present invention, the full channel information obtained in Step 12 includes:

[0041]

[0042] Among them,

[0043] The beneficial effects of the present invention are as follows: Compared with the prior art, the present invention discloses a method for jointly estimating Doppler frequency offset and time-varying channel, which mainly considers the time-varying channel estimation in the case of residual Doppler frequency offset in a RIS-assisted communication system. Due to the existence of residual Doppler frequency offset having a serious impact on channel estimation, this method first obtains a high-precision DFO estimation through a convolutional neural network and uses it to compensate the received signal; then, based on the received signal after frequency offset compensation, the direct channel estimation and cascaded channel estimation are performed. In order to further reduce the influence of noise and direct channel estimation error, the present invention adopts a threshold-based noise reduction method to perform noise reduction processing on the initial cascaded channel estimation, thereby obtaining high-precision cascaded channel state information. This method can obtain a more accurate DFO estimation and a more accurate channel estimation value, so it has strong practical value. Description of the Drawings

[0044] Figure 1 is a flowchart of the present invention.

[0045] Figure 2 is a structural diagram of the convolutional neural network adopted by the present invention.

[0046] Figure 3 is the MSE performance of the DFO estimation method in the present invention under different cyclic prefix lengths.

[0047] Figure 4 is the MSE performance of the DFO estimation method in the present invention with different training samples.

[0048] Figure 5 is the MSE performance of the present invention and the prior channel estimation method during ideal and non-ideal DFO synchronization. Detailed Embodiments

[0049] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail below with reference to the drawings and specific embodiments.

[0050] The present invention discloses a method for jointly estimating Doppler frequency offset and time-varying channel, which is applied to a RIS-assisted OFDM system. The method includes the following steps:

[0051] Step 1: Obtain a Doppler frequency offset estimation value according to the cyclic prefix (CP) of the time-domain received signal

[0052]

[0053] where arg{·} is the angle-taking operation, r(u,n) represents the nth time-domain discrete sampling signal of the u-th OFDM symbol received by the base station, N cp is the cyclic prefix length of the OFDM symbol, and N is the length of the OFDM symbol;

[0054] Step 2: Construct a training sample set using the time-domain received signal, that is

[0055] Τ DFO ={(s (1) ,ε (1) ),...,(s (v) ,ε (v) ),...,(s (V) ,ε (V) )}

[0056] where V is the number of training samples, s (v) is the v-th training input sample composed of the received signal, where and represent the real-part taking and imaginary-part taking operations respectively, r (v) (u,n) is the received signal in the v-th sample, U is the number of OFDM symbols, and ε (v) is the v-th output sample composed of the ideal DFO, is the ideal DFO on the u-th OFDM symbol in the v-th sample. Since the ideal DFO is unknown in actual communication, the present invention uses the DFO estimation as the training target of the network to improve the practicability of the method, that is

[0057]

[0058] where, is the v-th sample composed of the DFO estimation, is the DFO estimation of the u-th OFDM symbol obtained in the v-th sample using Step 1.

[0059] Step 3: Based on the randomly initialized network parameters, use the training sample set constructed in Step 2 to train the CNN network;

[0060] Step 4: Obtain a network model with optimal weight parameters and threshold parameters;

[0061] Step 5: Perform online DFO estimation based on the trained network model;

[0062] Step 6: Use the DFO estimation to perform frequency offset compensation on the time-domain received signal;

[0063] Step 7: Perform direct channel estimation. Assume that all RIS units are turned off during the first OFDM symbol, and use the LS (least squares) method and linear interpolation method to obtain the estimation of the direct channel.

[0064] Step 8: Perform cascaded channel estimation. Turn on the RIS units one by one. Assume that the m-th RIS unit is turned on and other RIS units are turned off. Based on the direct channel estimation in Step 7 and the LS method, the estimation of the pilot sub-channel in the m-th cascaded channel is

[0065]

[0066] where, is the received pilot signal without direct channel signal, is the u-th frequency-domain pilot signal transmitted, is the received signal on the k-th subcarrier in the u-th OFDM symbol after compensation, is the direct channel estimation on the k-th subcarrier in the u-th OFDM symbol, Γ tr is the set of pilot positions on the u-th OFDM symbol;

[0067] Step 9: Use the linear interpolation method and the pilot sub-channel estimation to obtain the m-th cascaded channel estimation as

[0068] Step 10: Adopt the threshold denoising method to further improve the accuracy of the cascaded channel estimation. First, perform N-point IFFT transformation on the initial cascaded channel estimation to obtain the time-domain cascaded channel estimation, and then use the threshold denoising method to obtain the denoised time-domain channel estimation as

[0069]

[0070] where n = 0,..., N cp -1, when n ≥ N cp When q is the threshold, that is

[0071]

[0072] Step 11: Perform FFT transformation on the denoised time-domain cascaded channel estimation to obtain the frequency-domain cascaded channel estimation

[0073]

[0074] where, F N is an N×N dimensional normalized discrete Fourier transform matrix, is 1×(N - Ncp ) A zero vector of dimension

[0075] Step 12: Obtain the final full channel information

[0076]

[0077] Wherein,

[0078] For example, in a RIS-assisted uplink OFDM communication system with N subcarriers, where a RIS with M passive reflecting units is deployed to assist in wireless uplink communication from the user to the base station, and there is one antenna at both the user side and the base station side. Thus, there are M + 1 channels between the user and the base station, including a direct channel and M RIS reflection channels. To obtain U OFDM signals in the M + 1 channel signal frames, where the u-th OFDM symbol is X u = diag[X u (0), …, X u (N - 1)], where diag[·] represents the diagonal operation, and X u (k) is the frequency-domain transmission signal on the k-th subcarrier of the u-th OFDM symbol.

[0079] Generally, the BS and the RIS are fixed. Therefore, the channel between the BS and the RIS is a quasi-static Rice fading channel with L1 paths. The channel between the m-th RIS unit and the BS can be written as

[0080]

[0081] Wherein, c m and are respectively the LOS component and the scattering component of the channel between the m-th RIS unit and the base station, follows the Rayleigh distribution, L1 is the number of multipaths, is the normalized delay of the l1-th path.

[0082] Due to the mobility of the user, the channels of UE - BS (direct channel) and UE - RIS are considered as time-varying Rice fading channels. Here, the Rice component is usually a constant, which is the DFO caused by the Doppler frequency shift, and the scattering component is a complex fading affected by the random Doppler spectrum and follows the Rayleigh distribution. Therefore, the channel between the user and the m-th RIS unit during the u-th OFDM symbol is

[0083]

[0084] Wherein, and are respectively the LOS component and the scattering component, and ε uis the normalized DFO by the subcarrier spacing during the duration of the u-th OFDM symbol, and L2 is the number of multipaths of the channel between the user and the RIS. is the normalized delay of the l2-th path. N s = N cp + N, where N cp is the length of the cyclic prefix. The channels from the UE to the RIS and from the RIS to the BS are combined into a cascaded channel, i.e.,

[0085]

[0086] In addition, the direct channel from the UE to the BS within the duration of the u-th OFDM symbol is

[0087]

[0088] where c0 and are the LOS and scattered components of the direct channel respectively, L0 is the number of multipaths of the direct channel, is the normalized delay of the l0-th path.

[0089] Therefore, the n-th time-domain discrete sampling signal received at the receiver during the u-th OFDM symbol can be written as

[0090]

[0091] where, is the convolution operation, θ u (m) is the reflection coefficient of the m-th RIS reflection unit during the u-th OFDM symbol, here β u,m and represent the reflection amplitude and phase shift of the m-th reflection unit during the duration of the u-th OFDM symbol respectively, and β u,m ∈(0,1), x(u,n) is the time-domain transmission signal corresponding to the u-th OFDM symbol, and w(u,n) is the additive white Gaussian noise during the u-th OFDM symbol. The frequency-domain received signal after removing the CP and performing the N-point IFFT operation is

[0092]

[0093] where, W(u,l) is the frequency-domain noise, and I(u,l) is the interference caused by the DFO and the time-varying channel, which can be expressed as

[0094]

[0095] where, Γ is the subcarrier index set, and I1(u,l) and I2(u,l) are the interferences caused by the DFO and the time-varying channel respectively, which can be expressed as

[0096]

[0097]

[0098] After receiving all subcarrier signals, the u-th OFDM symbol can be written as

[0099]

[0100] where

[0101]

[0102] I u =[I(u,0),...,I(u,l),...,I(u,N - 1)] T

[0103] W u =[W(u,0),...,W(u,l),...,W(u,N - 1)] T

[0104] This section will simulate and analyze the technology of the present invention to verify its performance. The present invention considers an OFDM system with an FFT length of 128, a cyclic prefix length of 16, a comb - shaped pilot structure with 32 pilot numbers, and 4 RIS units. It is assumed that the normalized Doppler shift of the channel is 0.044, the channel is a 5 - path Rice channel with a Rice factor of 5. The carrier frequency is considered to be 2.3 GHz, and the sub - carrier spacing is 15 kHz. In terms of the convolutional neural network, the convolutional layer of the neural network has 64 convolutional kernels of size 3*3. To compare the performance of the technology of the present invention, the simulation also gives the performance of the ON / OFF RIS channel estimation method and the MMSE channel estimation method.

[0105] Figure 3 The MSE performance of the DFO estimation method in the present invention with different cyclic prefix lengths is given. Among them, the DFO estimation uses a multi - layer CNN, and its structure diagram is as Figure 2 shown. It can be seen from Figure 3 that as the cyclic prefix length increases, the accuracy of the DFO estimation also increases. This is mainly because the longer the cyclic prefix length, the better the performance of estimating DFO using the cyclic - prefix - based method. Using a highly accurate DFO estimation as the training target of the network can make the trained network model better, so as to obtain a DFO with higher estimation accuracy. However, a large cyclic prefix will also lead to a decrease in the system transmission efficiency. Therefore, in actual use, a trade - off should be made between the estimation performance and the system efficiency to determine the length of the cyclic prefix.

[0106] Figure 4 shows the MSE performance of the DFO estimation described in the technology of the present invention under different training samples. In the simulation, the cyclic prefix length is 16. From Figure 4 it can be seen that when the number of training samples increases, the accuracy of the DFO estimation also improves, but too many training samples will cause longer training time. Therefore, in order to balance performance and complexity, an appropriate number of training samples should be selected in practical applications.

[0107] Figure 5 shows the MSE performance of the technology of the present invention and other methods. In the simulation, the number of training samples adopted by the technology of the present invention is 2000 and the cyclic prefix is 16. The MMSE method adopts accurate channel statistical information. In the figure, "fine synchronization" means that there is no influence of DFO in the system; "CP-based DFO estimation" means that there is an influence of DFO on the received signal in the system, and the system uses the CP-based DFO estimation method to estimate and compensate the received signal for frequency offset, and the channel is estimated based on this compensated signal; "DFO estimation of the present invention" means that there is an influence of DFO on the received signal in the system, and the system uses the DFO estimation method given by the present invention to estimate and compensate the received signal for frequency offset, and the channel is estimated based on this compensated signal. From Figure 5 it can be seen that in the case of fine synchronization, the existing methods have good estimation performance, but their performance degrades in the case of non-fine synchronization, and the ON / OFF method has the worst performance. In the case of non-fine synchronization, the performance of the existing technology using the DFO estimation method given by the present invention is better than that using the traditional DFO estimation method, because the DFO estimation method given by the present invention is based on CNN and has higher estimation accuracy than the traditional method. Moreover, in the case of non-fine synchronization, the technology of the present invention has better performance, mainly because this method takes into account the influence of the residual frequency offset. In the case of using the DFO estimation method given by the present invention, although the performance of the MMSE method is slightly better than that of the technology of the present invention at low signal-to-noise ratios, it is inferior to the performance of the technology of the present invention at high signal-to-noise ratios, mainly because the present invention uses channel noise reduction processing. Moreover, the performance of the MMSE method given in the simulation is all carried out with accurate channel statistical information, which is unknown in practice and needs to be obtained through estimation and statistical methods, which will reduce the estimation performance of this method. In addition, the MMSE method involves the inverse operation of matrices, and this method will have greater computational complexity in the case of large-dimensional RIS units, all of which limit the practical application of the MMSE method.

[0108] In summary, the present invention discloses a method for joint estimation of Doppler frequency offset and time-varying channel. This method mainly considers the time-varying channel estimation in the case of residual Doppler frequency offset in an RIS-assisted communication system. Due to the existence of residual Doppler frequency offset having a serious impact on channel estimation, this method first obtains a high-precision DFO estimation through a convolutional neural network and uses it to compensate the received signal; then, based on the received signal after frequency offset compensation, direct channel estimation and cascaded channel estimation are performed. In order to further reduce the influence of noise and direct channel estimation error, the present invention adopts a threshold-based noise reduction method to perform noise reduction processing on the initial cascaded channel estimation, thereby obtaining high-precision cascaded channel state information. This method can obtain more accurate DFO estimation and more accurate channel estimation values, so it has strong practical value.

Claims

1. A joint estimation method for Doppler frequency offset and time-varying channel, which is applied to an RIS-assisted OFDM system, is characterized in that The joint estimation method of Doppler frequency offset and time-varying channel includes the following steps: Step 1: Obtain the Doppler frequency offset estimation value according to the cyclic prefix of the time-domain received signal; Step 2: Construct a training sample set using the time-domain received signal; Step 3: Based on the randomly initialized network parameters, train the CNN network using the training sample set constructed in Step 2; Step 4: Obtain a network model with optimal weight parameters and threshold parameters; Step 5: Perform online DFO estimation based on the trained network model; Step 6: Perform frequency offset compensation on the time-domain received signal using the DFO estimation; Step 7: Perform direct channel estimation. During the first OFDM symbol, turn off all RIS units and use the LS method and linear interpolation method to obtain the estimation of the direct channel Step 8: Perform cascaded channel estimation. Sequentially turn on the RIS units, turn on the m-th RIS unit and turn off other RIS units, and obtain the estimation of the pilot sub-channel in the m-th cascaded channel based on the direct channel estimation and the LS method in Step 7; Step 9: Using the linear interpolation method and the pilot sub-channel estimation in Step 8, the m-th cascaded channel estimation is obtained as Step 10: Further improve the accuracy of the cascaded channel estimation using the threshold denoising method; Step 11: Perform FFT transformation on the denoised time-domain cascaded channel estimation to obtain the frequency-domain cascaded channel estimation; Step 12: Obtain the final full-channel information.

2. The joint estimation method of Doppler frequency offset and time-varying channel according to claim 1, characterized in that The formula for calculating the Doppler frequency offset estimation value in Step 1 is as follows: where arg{·} is the angle-taking operation, r(u,n) represents the n-th time-domain discrete sampling signal of the u-th OFDM symbol received by the base station, and N cp is the cyclic prefix length of the OFDM symbol, and N is the length of the OFDM symbol.

3. The method for jointly estimating Doppler frequency offset and time-varying channel according to claim 2, wherein The formula for constructing the training sample set using the time-domain received signal in Step 2 is as follows: T DFO = {(s (1) , ε (1) ),...,(s (v) , ε (v) ),...,(s (V) , ε (V) )} where \(V\) is the number of training samples, \(s\) (v) is the \(v\)-th training input sample composed of received signals, and represent the real part taking and imaginary part taking operations respectively, \(r\) (v) \((u,n)\) is the received signal in the \(v\)-th sample, \(u\) is the number of OFDM symbols, \(\varepsilon\) (v) is the \(v\)-th output sample composed of the ideal DFO, is the ideal DFO on the \(u\)-th OFDM symbol in the \(v\)-th sample. To improve the practicality of the method, the DFO estimation is used as the training objective of the network, that is Among them, is the v-th sample for DFO estimation, is the DFO estimation of the u-th OFDM symbol obtained by using Step 1 in the v-th sample.

4. The method for jointly estimating Doppler frequency offset and time-varying channel according to claim 3, wherein The formula for obtaining the estimation of the pilot sub-channel in the cascaded channel in Step 8 is as follows: Among them, is the received pilot signal without a direct connection channel signal, is the u-th frequency-domain pilot signal transmitted, is the received signal on the k-th subcarrier in the u-th OFDM symbol after compensation, is the direct connection channel estimation on the k-th subcarrier in the u-th OFDM symbol, Γ tr is the set of pilot positions on the u-th OFDM symbol.

5. The Doppler frequency offset and time-varying channel joint estimation method according to claim 4, wherein Step 10 includes: First, perform an initial estimation of the cascaded channel Perform an N-point IFFT transformation to obtain the time-domain cascaded channel estimation. Then, the denoised time-domain channel estimation can be obtained using a threshold denoising method as where n = 0,..., N cp -1, when n ≥ N cp when q is a threshold value, i.e., 6. The method for jointly estimating Doppler frequency offset and time-varying channel according to claim 5, wherein The formula for obtaining the frequency-domain cascaded channel estimation in Step 11 is as follows: Among them, F N is an N×N dimensional normalized discrete Fourier transform matrix, is a 1×(N - N cp ) dimensional all-zero vector.

7. The method for jointly estimating Doppler frequency offset and time-varying channel according to claim 6, characterized in that The formula for the full-channel information obtained in Step 12 is as follows: Among them,

Images