Millimeter wave frequency division duplex large-scale MIMO system channel estimation method based on compressed sensing and deep learning
By combining compression perception and deep learning AS-JOMP and DCNN-LSTM methods, the problem of large channel estimation overhead in frequency division duplex large-scale MIMO systems is solved, the estimation accuracy of time-varying channels is improved, and the non-priority test is achieved Efficient channel recovery in the case of channel information.
Patent Information
- Application Number
- CN202411021391.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-29
AI Technical Summary
In frequency division duplex large-scale MIMO systems, channel estimation overhead is large and the estimation accuracy of time-varying channels is insufficient. Especially in the absence of prior channel information, traditional methods are difficult to effectively reduce noise and improve channel recovery accuracy.
Adaptive structured orthogonal matching tracking (AS-JOMP) algorithm based on compression perception combined with deep learning convolutional neural network and long-term memory network (DCNN-LSTM) method, adaptive reconstruction of channels is used to utilize the structural sparsity of channels. , and remove noise through the DCNN-LSTM network to improve the final estimation accuracy of CSI.
Reduces the overhead of downlink training and uplink feedback, improves the channel estimation accuracy of time-varying channels, and can adaptively reconstruct the channel without knowing the number of non-zero elements in the CSI, making full use of the effective information of time-varying channels. .
Smart Images

Figure CN118972207A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of wireless communication technology, and in particular, relates to a method for channel estimation using compressed sensing and deep learning in a massive multiple-input multiple-output (MIMO) system in a millimeter-wave frequency division duplex (FDD) mode. The present invention aims to reduce the overhead of channel estimation and improve the estimation accuracy in time-varying channels. Background Art
[0002] Massive Multiple-input Multiple-output (MIMO) technology has been regarded as a key technology for the fifth generation of wireless communication systems due to its huge performance advantages in terms of capacity and reliability. [1] In order to achieve these performance gains of massive MIMO, it is necessary to obtain downlink channel state information (CSI) at the base station (BS). [2] In time division duplex systems, channel reciprocity is usually used to obtain downlink CSI, but this requires high precision calibration of hardware circuits and is costly. [3] In a frequency-division duplexing (FDD) system, the reciprocity of the uplink and downlink channels does not hold, so the downlink channel CSI must be estimated first and then fed back to the BS. [4] . Therefore, it does not require complex hardware calibration. At the same time, traditional cellular systems mainly work in FDD mode. Therefore, FDD massive MIMO system is currently a very hot topic in academia and industry. However, since massive antennas are used on antenna arrays, FDD massive MIMO will lead to huge channel acquisition overhead and channel time-varying fading will further increase the complexity of channel acquisition. For traditional pilot algorithms, such as least squares or minimum mean square error algorithms, the number of required pilots is always linearly related to the number of antennas, which will generate huge computational complexity. [5] To solve this problem, Compressed Sensing (CS) is widely used in large-scale MIMO channel estimation to recover sparse channel state information from the reduced received pilot signals. [6][7]. However, considering the channel estimation and feedback costs, they cannot play a greater advantage in time-varying channels. Therefore, it is necessary to study the channel training and feedback schemes for FDD massive time-varying MIMO systems. Although the channel is time-varying, its statistical characteristics are often stable, so the information of the previous estimation can be used to track the time-varying characteristics of the channel in the current estimation. In particular, the stable statistical characteristics will lead to the sparsity of the channel in some areas, so compressed sensing technology can be used to estimate the channel. Reference [8] proposed a CS recovery method for estimating time-varying channels using a priori support channels, and showed that the training signal can be further reduced by using time correlation. Reference [9] further utilized the time correlation of the time-varying channel and proposed a differential-based structured compressed sampling matching pursuit (S-CoSaMP) algorithm to obtain CSI. The above method relies on the prior knowledge of the channel, that is, the correlation and sparsity of the channel. Reference
[10] proposed a distributed compressed sensing-assisted channel estimation method, which makes full use of the slow variation of channel statistics in consecutive time slots and the spatial common sparsity of multiple subchannels in the frequency domain. Reference
[11] proposed a feasible downlink training sequence design method based on partial CSI estimation. However, the channel estimation schemes in the above-mentioned references have strict requirements on prior channel information. In references
[12] -
[16] , several channel estimation schemes without prior channel information are proposed. References
[12] and
[13] proposed adaptive step-size sparsity adaptive matching pursuit (AS-SAMP) and unknown sparsity orthogonal matching pursuit algorithm (OMP-US) for sparse channel estimation when the number of non-zero coefficients is unknown. Reference
[14] proposed a low-overhead adaptive channel estimation and feedback algorithm for OFDM systems, which can adaptively adjust the training overhead and pilot design to achieve reliable CSI estimation. Compared with the method in reference [8], reference
[15] proposed an adaptive M-SP method, which can adaptively adjust the prior channel support quality parameter to a suitable value in the case of model mismatch. In
[16] , an approximate message passing algorithm based on expectation maximization and Gaussian mixture distribution is proposed, which can simultaneously learn signal distribution and recover signals. However, all CS-based methods have a disadvantage that the noise added on non-zero elements is difficult to eliminate. In recent years, deep learning (DL) methods have been successfully applied to wireless communications, especially in channel estimation and feedback. In
[17] and
[18] , two RNN networks were proposed to improve the feedback accuracy of CSI in massive MIMO systems. In
[19] , a method based on a convolutional long-short-term memory (LSTM) network was proposed to predict downlink CSI from uplink CSI.By utilizing the correlation of time-varying channels, the convolutional neural network (CNN) and recurrent neural network (RNN) in
[20] extract spatial features and inter-frame correlations respectively. In
[21] , the CS-based and DL-based methods are combined to form a framework called CS-ReNet, which compresses the perfect CSI on the user side and then reconstructs the CSI on the BS side using a deep neural network. From the perspective of bit-level quantization performance, the joint convolutional residual network in
[22] is beneficial for the extraction and recovery of MIMO channel features. For high mobility environments, a channel estimation network consisting of a CNN that simulates the frequency domain interpolation process and a bidirectional LSTM network for time domain channel prediction is developed in
[23] . In addition to the data-driven methods mentioned above, there are also some model-driven deep learning methods to obtain CSI. In
[24] , the fully convolutional denoising network (FCDNet) is combined with the learning approximate message passing network (LAMP) to propose a fully convolutional denoising approximate message passing (FCDAMP) algorithm.
[0003] [1] Z.Qin, J.Fan, Y.Liu, Y.Gao, and GYLi, "Sparse representation for wireless communications: Acompressive sensing approach," IEEE SignalProcess.Mag., vol.35, no.3, pp.40–58, May2018.
[0004] [2] Papazafeiropoulos A, Kourtessis P, Renzo MD, et al. "PerformanceAnalysis of Cell-Free Massive MIMO Systems: A Stochastic Geometry Approach," IEEE Transactions on VehicularTechnology, 2020, 69(4): 3523-3537.
[0005] [3] M.Arnold,S. S. Cammerer, S. Yan, J. Hoydis, and STBrink, “Enabling FDD massive MIMO through deep learning-based channel prediction,” 2019, arXiv:1901.03664.
[0006] [4]X.Rao and V.K.N.Lau,“Distributed compressive CSIT estimation andfeedback for FDD multi-user massive MIMO systems,”IEEE Trans.Signal Process.,vol.62,no.12,pp.3261–3271,Jun.2014.
[0007] [5]M.K.Ozdemir and H.Arslan,“Channel estimation for wireless OFDMsystems,”IEEE Commun.Surv.Tut.,vol.9,no.2,pp.18–48,Apr.–Jun.2007.
[0008] [6]X.Zhu,L.Dai,G.Gui,W.Dai,Z.Wang,and F.Adachi,“Structured matchingpursuit for reconstruction of dynamic sparse channels,”in Proc.IEEEGlob.Commun.Conf.,2015,pp.1–5.
[0009] [7]Y.Han,P.Zhao,L.Sui,and Z.Fan,“Time-varying channel estimationbased on dynamic compressive sensing for OFDM systems,”in Proc.IEEEInt.Symp.Broadband Multimedia Syst.Broadcast.,2014,pp.1–5.
[0010] [8]H.Yang,Y.Fan,D.Liu,Z.Zheng,and S.Lin,“Compressive sensing andprior support based adaptive channel estimation in massive MIMO,”in Proc.2ndIEEE Int.Conf.Comput.Commun.,2016,pp.1618–1622.
[0011] [9]W.Shen,L.Dai,Y.Shi,B.Shim,and Z.Wang,“Joint channel training andfeedback for FDD massive MIMO systems,”IEEE Trans.Veh.Technol.,vol.65,no.10,pp.8762–8767,Oct.2016.
[0012]
[10] R.Zhang,H.Zhao,and J.Zhang,“Distributed compressed sensing aidedsparse channel estimation in FDD massive MIMO system,”IEEE Access,vol.6,pp.18383–18397,2018.
[0013]
[11] M.A.Naser,M.Q.Alsabah,and M.A.Taher,“A partial CSI estimationapproach for downlink FDD massive-MIMO system with different base transceiverstation topologies,”Wireless Pers.Commun.,vol.119,pp.3609–3630,2021.
[0014]
[12] Y.Zhang,R.Venkatesan,O.A.Dobre,and C.Li,“An adaptive matchingpursuit algorithm for sparse channel estimation,”in Proc.IEEE WirelessCommun.Netw.Conf.,2015,pp.626–630.
[0015]
[13] M.J.Azizipour and K.Mohamed-Pour,“Compressed channel estimationfor FDD massive MIMO systems without prior knowledge of sparse channelmodel,”IET Commun.,vol.13,no.6,pp.657–663,Apr.2019.
[0016]
[14] Z.Gao,L.Dai,Z.Wang,and S.Chen,“Spatially common sparsity basedadaptive channel estimation and feedback for FDD massive MIMO,”IEEETrans.Signal Process.,vol.63,no.23,pp.6169–6183,Dec.2015.
[0017]
[15] X.Bi,J.Zhao,G.Wang,Y.Lu,L.Zhou,and D.Li,“Modified CS-baseddownlink channelestimation with temporal correlation in FDD massive MIMOsystems,”in Proc.Wireless Telecommun.Symp.,2018,pp.1–7.
[0018]
[16] J.P.Vila andP.Schniter,“Expectation-maximizationGaussian-mixtureapproximatemessage passing,”IEEE Trans.SignalProcess.,vol.61,no.19,pp.4658–4672,Oct.2013.
[0019]
[17] C.Lu,W.Xu,H.Shen,J.Zhu,and K.Wang,“Mimo channel informationfeedback using deeprecurrentnetwork,”IEEE Commun.Lett.,vol.23,no.1,pp.188–191,Jan.2019.
[0020]
[18] X.Li and H.Wu,“Spatio-temporal representation with deep neuralrecurrent network in MIMO CSI feedback,”IEEEWirelessCommun.Lett.,vol.9,no.5,pp.653–657,May2020.
[0021]
[19] J.Wang,Y.Ding,S.Bian,Y.Peng,M.Liu,and G.Gui,“UL-CSI data drivendeep learning for predicting DL-CSI in cellular FDD systems,”IEEE Access,vol.7,pp.96105–96112,2019.
[0022]
[20] T.Wang,C.Wen,S.Jin,and G.Y.Li,“Deep learning-based CSI feedbackapproach for time-varying massive MIMO channels,”IEEE Wireless Commun.Lett.,vol.8,no.2,pp.416–419,Apr.2019.
[0023]
[21] P.Liang,J.Fan,W.Shen,Z.Qin,and G.Y.Li,“Deep learning andcompressive sensing-based CSI feedback in FDD massive MIMO systems,”IEEETrans.Veh.Technol.,vol.69,no.8,pp.9217–9222,Aug.2020.
[0024]
[22] C.Lu,W.Xu,S.Jin,and K.Wang,“Bit-level optimized neural networkfor multi-antenna channelquantization,”IEEEWireless Commun.Lett.,vol.9,no.1,pp.87–90,Jan.2020.
[0025]
[23] Y.Liao,Y.Hua,and Y.Cai,“Deep learning based channel estimationalgorithm for fast time-varying MIMO-OFDM systems,”IEEE Commun.Lett.,vol.24,no.3,pp.572–576,Mar.2020.
[0026]
[24] H.He, C.-K.Wen, S.Jin, and GYLi, "Deep learning-based channel estimation for beamspace mmwave massive MIMO systems," IEEE WirelessCommun.Lett., vol.7, no.5, pp.852–855, Oct.2018. Summary of the invention
[0027] In order to solve the problems existing in the above-mentioned prior art, the present invention proposes a channel estimation scheme for an FDD massive MIMO system based on CS and DL, which reduces the overhead of downlink training and uplink feedback, and improves the channel estimation accuracy in a time-varying channel; the scheme first directly feeds back the pilot received by the user to the base station, and then adopts an adaptive structured orthogonal matching pursuit (AS-JOMP) method, which can utilize the structural sparsity of the channel to adaptively reconstruct the channel without knowing the number of non-zero elements in the CSI; this structural sparsity is caused by the close arrangement of antennas on the base station so that the signals experience similar paths; after obtaining the initial estimate of the CSI, the initial CSI cannot effectively eliminate the noise addition at the position of the non-zero elements, and does not fully utilize the effective information between the time-varying channels. The present invention develops a DL-based DCNN-LSTM method to reconstruct the channel matrix, further improving the accuracy of the final CSI estimation.
[0028] To achieve the above object, the present invention provides the following technical solutions:
[0029] A channel estimation method for a millimeter-wave frequency division duplex massive MIMO system based on compressed sensing and deep learning, the steps are as follows:
[0030] S1. Compressed sensing preliminary estimation: The received pilot signal is processed using the compressed sensing method to extract the preliminary CSI matrix;
[0031] S2, deep learning network structure: using a combination of convolutional neural network and long short-term memory network, extracting spatial features through convolutional neural network and capturing temporal correlation through long short-term memory network;
[0032] S3, ConvLSTM network processing: The initially estimated CSI matrix is input into the ConvLSTM layer in chronological order. The ConvLSTM network integrates time-related information to improve the accuracy of CSI estimation;
[0033] S4, CSI reconstruction: Use the same padding, ReLU activation function and appropriately sized filters to obtain the same size as the input data; obtain the final CSI estimation result through dimensionality transformation and denormalization.
[0034] Furthermore, it is performed through a large-scale MIMO-OFDM system in millimeter-wave frequency division duplex mode. In the large-scale MIMO-OFDM system, communication is performed between a base station and multiple single-antenna users, a comb pilot method is adopted, the pilot interval within each OFDM symbol is equal, and the pilot symbols are randomly selected.
[0035] Furthermore, in the massive MIMO-OFDM system, the following steps are performed:
[0036] D1. Pilot symbol generation: randomly select the pilot symbol at each antenna from the OFDM symbol, and obtain the transmission symbol in the time domain through inverse discrete Fourier transform;
[0037] D2. Signal reception: At the receiving end, the received signal is represented in the frequency domain through Fourier transform;
[0038] D3. Channel state information extraction: The user extracts pilot symbols from the received signal and recovers the sparse CSI matrix through compressed sensing method.
[0039] Furthermore, the pilot frequency C of the i-th transmitting antenna of the k-th user ik ∈C P×1 , where P is the number of pilots, and the pilot symbol at the i-th antenna is C i ∈C P×1 , where C i The elements in are from OFDM symbols Randomly selected from; through inverse discrete Fourier transform, the transmission symbol in the time domain is obtained as:
[0040]
[0041] in is the DFT matrix, (.) H is the complex conjugate transpose, N c is the number of subcarriers; the i-th antenna at the BS transmits the time domain signal X i ; Then the signal y received in the time domain is expressed as follows:
[0042]
[0043] where h i ∈C L×1 is the channel state information from the ith antenna of the BS to the user, L is the channel length, represents independent and identically distributed (iid) additive white complex Gaussian noise, * represents the convolution operation, also expressed as matrix multiplication;
[0044] If the OFDM system has a cyclic prefix of appropriate length, the linear convolution in equation (2) becomes a circular convolution after the cyclic prefix is removed at the receiving end; i Using zero padding, the convolution operation will be expressed as follows:
[0045]
[0046] in From L to N c The elements are all 0, and the convolution in formula (3) can also be expressed as the following matrix multiplication:
[0047]
[0048] in is a Toeplitze matrix; as follows:
[0049]
[0050] In formula (4), From y o To N c The cumulative sum of the sequence; and the circular convolution in formula (3) is represented by matrix convolution;
[0051] Through DFT transformation, the symbol received in the frequency domain can be expressed as
[0052]
[0053] in The diagonal elements are Matrix of
[0054] In order to estimate the channel state information h i , the user must receive the signal y f Extract pilot symbols The received signal can be written as follows:
[0055]
[0056] in is from the index set Ω Selected from F Ω,L ∈C P×L The corresponding is a partial Fourier matrix, the rows are selected from the pilot indices, and the columns are selected from the first L columns of F; obviously, Equal to c i;therefore
[0057]
[0058] in is the CSI from the ith base station antenna to the user, It means that the mean is zero and the variance is Additive Gaussian white noise; if the pilot intervals within the OFDM symbol are equal and the pilot symbols are randomly selected comb pilots, the pilot received from a certain user is
[0059]
[0060] in
[0061] Formula (10) is regarded as a compressed sensing problem, Φ is the sensing matrix, and h is the sparse signal; therefore, the reconstruction algorithm based on compressed sensing is used to obtain the value of Φ and y c In this channel model, the uplink channel and the downlink channel have the same signal-to-noise ratio, and the noise parameter represents the total noise power of the downlink and uplink channels; for block fading time-varying MIMO channels, the CSI varies from one time slot to another, but remains unchanged within a time slot; the dynamic channel is modeled by the change of CSI support and the evolution of the amplitude of non-zero elements:
[0062]
[0063] Among them, h t is the CSI at time t, s t (l)∈{0,1} represents s t Whether the lth support index of is zero, g t (l)∈C is g t The amplitude at the lth support index of Represents the Hadamard product; specifically, a first-order Markov process is used to model the change in support, that is, two transition probabilities are defined as:
[0064]
[0065] Among them, In order to achieve a steady-state sparsity rate p∈(0,1), the amplitude adopts a first-order autoregressive model, which is expressed as follows:
[0066]
[0067] The correlation coefficient ρ = J 0 (2πf d τ) is given by the zero-order Bessel function of the first kind, f dis the maximum Doppler frequency, τ is the duration, and the parameters is a complex Gaussian variable iid.
[0068] Furthermore, the method further includes the following steps: using the AS-JOMP algorithm based on compressed sensing to recover the CSI from the received signal; the AS-JOMP algorithm combines the sensing matrix Φ with the remaining observation vector r according to the structural sparsity. i The cross-correlation vector of the product is divided into N BS The vectors are summed and then the maximum value indicator like OMP is selected; in the second observation, two threshold parameters are used to stop the iteration under high SNR and low SNR.
[0069] Furthermore, the method further comprises the following steps: using the estimated CSI as the input of DCNN-LSTM to remove noise and learn the characteristics between time-varying channels; the DCNN-LSTM is a joint network based on CNN and LSTM, in which the CSI channel group As input, where T is the number of CSI in the channel group i = 1, 2, ..., N sample , where N sample is the number of samples; the output of the network is expressed as Obtained by AS-JOMP algorithm Is a size of LN BS ×1 complex vector; therefore, separate the real and imaginary parts of the complex vector into two LN BS ×1 real vector, and then transform it into two vectors of size The real matrix of Therefore, the data of each layer of the network is a four-dimensional tensor. The sum of the products on the left side of each layer represents the size of the CSI channel group; the number above is the number of feature maps.
[0070] Furthermore, the DCNN-LSTM includes a channel feature extraction module, a time-varying channel correlation extraction module and a CSI reconstruction module.
[0071] Furthermore, the channel feature extraction module is completed by the CNN network in the first layer. In this process, the matrix of each time dimension can be convolved separately to extract features. The convolution layer contains 64 convolution kernels of size 3×3 and a sliding step of 1. The same zeros are used to fill the input during convolution so that the feature map output by each layer of the network remains the same size as the original channel matrix, and ReLU is used as the activation function.
[0072] Furthermore, the time-varying channel correlation extraction module: the extraction of time correlation is mainly completed by the second layer LSTM network; the ConvLSTM network consists of an input gate i tand an output gate O t The input gate is used to record the information of the current state, and the output gate is used to control how much information of the current state can be seen by the external network; the forget gate f t Control the number of historical state information flows to the current state after allowing entry, memory unit C t The information of the previous moment is stored; as shown in formula 16, where W xi , W hi , W ci , W ho , W xo , W co , W xf represents the weight matrix, b i , b o , b f , b c Represents bias, and the number in the range [0, 1] represents the proportion of information transmission, 0 means no transmission, and 1 means full transmission;
[0073] The initially estimated CSI matrix is used as the input of the ConvLSTM layer in chronological order; at each time step, the ConvLSTM network can fuse the time-related information learned at the previous time point into the input of the current time step, and the relevant information is updated with the time step;
[0074]
[0075] Furthermore, the CSI reconstruction module: uses the same padding, ReLU activation function and 3×3×2 filter size to obtain the same size as the input data; obtains the final CSI estimation result through dimensionality transformation and denormalization
[0076] Beneficial effects of the present invention:
[0077] Compared with the prior art, the millimeter wave frequency division duplex large-scale MIMO system channel estimation method based on compressed sensing and deep learning described in the present invention uses compressed sensing and deep learning to perform channel estimation, aiming to reduce the overhead of channel estimation and improve the estimation accuracy in time-varying channels, reduce the overhead of downlink training and uplink feedback, and improve the channel estimation accuracy in time-varying channels; the scheme first directly feeds back the pilot received by the user to the base station, and then adopts an adaptive structured orthogonal matching pursuit (AS-JOMP) method, which can utilize the structural sparsity of the channel to adaptively reconstruct the channel without knowing the number of non-zero elements in the CSI; this structural sparsity is caused by the close arrangement of antennas on the base station so that the signal experiences similar paths; after obtaining the initial estimate of the CSI, the initial CSI cannot effectively eliminate the noise addition at the position of the non-zero element, and does not fully utilize the effective information between time-varying channels. The present invention develops a DL-based DCNN-LSTM method to reconstruct the channel matrix, further improving the accuracy of the final CSI estimation. BRIEF DESCRIPTION OF THE DRAWINGS
[0078] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:
[0079] Figure 1 is a comb-shaped pilot pattern of the present invention;
[0080] Figure 2 For the present invention D CNN -LSTM network architecture diagram;
[0081] Figure 3 This is a basic unit structure diagram of the ConvLSTM network of the present invention. DETAILED DESCRIPTION
[0082] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Figure 1-3 The channel estimation method of millimeter-wave frequency division duplex massive MIMO system based on compressed sensing and deep learning is further explained.
[0083] Example 1
[0084] The present invention provides a channel estimation method for a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning. The specific implementation of the present invention will be described in detail below so that those skilled in the art can implement the present invention without creative work.
[0085] 1. System Model
[0086] The method of the present invention is applicable to a large-scale MIMO-OFDM system in a millimeter wave frequency division duplex (FDD) mode. In the system, a base station (BS) communicates with multiple single-antenna users. The present invention adopts a comb pilot scheme, such as Figure 1 In this scheme, the pilot intervals in each OFDM symbol are equal and the pilot symbols are randomly selected.
[0087] -Pilot symbol generation: The pilot symbol at each antenna is randomly selected from the OFDM symbol and the transmission symbol in the time domain is obtained through the Inverse Discrete Fourier Transform (IDFT).
[0088] -Signal reception: At the receiving end, considering the cyclic prefix and zero padding technology of the OFDM system,
[0089] The received signal is represented in the frequency domain by Fourier transform.
[0090] -Channel State Information (CSI) extraction: The user extracts pilot symbols from the received signal,
[0091] The sparse CSI matrix is restored through compressed sensing technology.
[0092] 2. Channel Estimation
[0093] The present invention uses a method combining compressed sensing and deep learning to perform channel estimation. The specific steps are as follows:
[0094] a) Compressed sensing preliminary estimation: The received pilot signal is processed using compressed sensing technology to extract a preliminary CSI matrix.
[0095] b) Deep learning network structure: using a combination of convolutional neural network (DCNN) and long short-term memory network (LSTM), such as Figure 2 As shown in Figure 2, DCNN is used to extract spatial features and LSTM is used to capture temporal correlations.
[0096] c) ConvLSTM network processing: The initially estimated CSI matrix is input into the ConvLSTM layer in chronological order. The ConvLSTM network can integrate time-related information to further improve the accuracy of CSI estimation. The basic unit structure of ConvLSTM is as follows: Figure 3 shown.
[0097] d) CSI reconstruction: In order to obtain the same size as the input data, the same padding, ReLU activation function and appropriately sized filters are used. Finally, the final CSI estimation result is obtained through dimension transformation and inverse normalization.
[0098] Example 2
[0099] 1.1 System Model
[0100] The present invention considers a millimeter wave massive MIMO-OFDM system in frequency division duplex mode. In the system, N antennas are installed in the BS and K single antenna users. BS The present invention adopts a comb pilot scheme, such as Figure 1 shown.
[0101] In this scheme, the pilot intervals in each OFDM symbol are equal, and each circle represents a resource unit in the time domain and frequency domain. Users working in non-overlapping frequency bands. The pilot C of the i-th transmitting antenna of the k-th user ik ∈C P×1 , where P is the number of pilots. In order to avoid loss of generality, the present invention omits the subscript k in the following discussion. The pilot symbol at the i-th antenna is C i ∈C P×1 , where C i The elements in are from OFDM symbols Through the inverse discrete Fourier transform (IDFT), the transmission symbol in the time domain can be obtained as
[0102]
[0103] in is the DFT matrix, (·) H is the complex conjugate transpose, N c is the number of subcarriers. The i-th antenna at the BS transmits the time domain signal X i Then the signal y received in the time domain is expressed as follows
[0104]
[0105] where h i ∈C L×1 is the channel state information from the ith antenna of the BS to the user, L is the channel length, represents independent and identically distributed (iid) additive white complex Gaussian noise (AWGN), and * represents the convolution operation, which can also be expressed as matrix multiplication.
[0106] If the OFDM system has a cyclic prefix of appropriate length, the linear convolution in (2) becomes a circular convolution after the cyclic prefix is removed at the receiver.i With proper zero padding (ZP), the convolution operation will be expressed as follows
[0107]
[0108] in From L to N c The elements are all 0, and the convolution in (3) can also be expressed as the following matrix multiplication
[0109]
[0110] in is a Toeplitze matrix. As shown below
[0111]
[0112] In (4), From y o To N c The cumulative sum of the sequence. And the circular convolution in (3) can be represented by matrix convolution.
[0113] Through DFT transformation, the symbol received in the frequency domain can be expressed as
[0114]
[0115] in The diagonal elements are The matrix of .
[0116] In order to estimate the channel state information h i , the user must receive the signal y f Extract pilot symbols The received signal can be written as follows
[0117]
[0118] in is from the index set Ω Selected from F Ω,L ∈C P×L The corresponding is a partial Fourier matrix, where the rows are selected from the indices of the pilots and the columns are selected from the first L columns of F.
[0119] Obviously, Equal to c i .therefore
[0120]
[0121] in is the CSI from the ith base station antenna to the user, It means that the mean is zero and the variance is Additive Gaussian white noise. If we use Figure 1 The comb pilot shown in Figure 2 is the pilot received from a certain user.
[0122]
[0123] in
[0124] Based on a large amount of experimental data and theoretical analysis, it is known that CSI h i The number of non-zero elements of is much smaller than its length, so the channel can be compressed in the time domain. Therefore, (10) can be regarded as a compressed sensing problem, where Φ is the sensing matrix and h is the sparse signal. Therefore, the present invention can use a reconstruction algorithm based on compressed sensing to obtain the value of Φ and y. c In this channel model, the uplink channel and the downlink channel have the same signal-to-noise ratio, and the noise parameter Indicates the total noise power of the downlink and uplink channels
[25] For block fading time-varying MIMO channels, the CSI varies from one slot to another but remains constant within a slot. The dynamic channel can be modeled by the variation of the CSI support and the evolution of the amplitude of the non-zero elements:
[0125]
[0126] Among them, h t is the CSI at time t, s t (l)∈{0,1} represents s t Whether the lth support index of is zero, g t (l)∈C is g t The amplitude at the lth support index of Denotes the Hadamard product. Specifically, a first-order Markov process is used to model the change of support, that is, two transition probabilities are defined as
[0127]
[0128] Among them, To achieve a steady-state sparsity rate p∈(0,1), the amplitude adopts a first-order autoregressive model, which is expressed as follows:
[0129]
[0130] The correlation coefficient ρ = J 0 (2πf d τ) is given by the zero-order Bessel function of the first kind, f dis the maximum Doppler frequency, τ is the duration, and the parameters is a complex Gaussian variable iid. Some classic Cy algorithms require knowing the number of non-zero elements in the channel matrix in advance, which is not suitable for time-varying channel systems because it is difficult to obtain prior knowledge in each time slot. Therefore, the present invention designs a new method that does not require prior knowledge of the channel and is more suitable for millimeter wave FDD massive MIMO systems.
[0131] 1.2 Joint channel training and feedback
[0132] 1.2.1 AS-JOMP algorithm based on compressed sensing
[0133] In this section, a joint training and feedback scheme is proposed, which includes two steps. First, the AS-JOMP algorithm based on compressed sensing is used to recover the CSI from the received signal. Second, the estimated CSI is used as the input of DCNN-LSTM to remove noise and learn the features between time-varying channels.
[0134] For simplicity, the superscript t is omitted in the above equation. For FDD systems, the CSI knowledge at the BS is obtained by the steps of estimating the CSI of the i-th user and then feeding it back to the BS. Using the traditional LS-based CSI estimation technique, the channel can be estimated in the following way:
[0135]
[0136] in, is the Moore-Pensrose pseudo-inverse. However, this LS-based approach requires P ≥ LN BS , when N BS When P is large, this will increase the pilot training and CSI feedback overhead. <LN BS When , formula (7) is an underdetermined problem, which can be solved by using CS technology to exploit the sparseness of CSI. For the general CS model, and Φ are called measurement and measurement matrices, respectively. Obviously, due to the reduction in the number of pilots, the downlink training overhead is also reduced. In addition, the user feeds back the received pilots to the BS, so that the uplink feedback can also reduce LN BS to P.
[0137] Among many compressed sensing reconstruction algorithms, the orthogonal matching pursuit (OMP) algorithm is a classic algorithm that has been used and improved many times.
[26] OMP reconstructs the channel by iteratively identifying a support set containing the indices of the Φ columns that correspond to the metric In each iteration, it selects an index based on the maximum correlation test and subtracts the contribution of the corresponding column from the current metric. The iterative process continues until the indices of all non-zero elements are identified. However, sparsity S is not available in general practical applications. Therefore, the present invention adopts an adaptive structured orthogonal matching pursuit algorithm based on the OMP algorithm, namely the AS-JOMP algorithm. The AS-JOMP algorithm combines the perception matrix Φ with the remaining observation vector r according to the structural sparsity. i The cross-correlation vector of the product is divided into N BS The vector is summed and then the maximum value indicator is selected like OMP. This method can more accurately select the non-zero element indicators of CSI. In the second observation, two threshold parameters are used to stop the iteration under high SNR and low SNR. Therefore, the AS-JOMP algorithm can adaptively recover CSI without any prior sparsity information. Compared with the most advanced CS-based channel estimation schemes, the AS-JOMP algorithm has the following significant features:
[0138] (1) The proposed AS-JOMP algorithm can adaptively obtain the sparsity of the mmWave massive MIMO channel, avoiding the unrealistic assumption of using channel sparsity as a priori information for channel estimation. The proposed stopping criterion enables the AS-JOMP algorithm to achieve good channel estimation performance under both high and low SNR conditions.
[0139] (2) The AS-JOMP algorithm considers the structural sparsity of h and provides a more accurate support update. The CSI recovery performance can be further improved by updating the support of .
[0140] 1.2.2 DCNN-LSTM based on deep learning
[0141] Although the AS-JOMP algorithm can select the non-zero element index of the sparse channel, the noise is not eliminated and the time-varying characteristics of the channel are not well utilized. Based on these requirements, the present invention proposes a network architecture based on deep learning, namely DCNN-LSTM. DCNN-LSTM is a joint network based on CNN and LSTM. The network can not only reduce noise by extracting channel structure information through CNN, but also learn the correlation of time domain channels through LSTM network. In this way, the spatiotemporal characteristics of the channel can be effectively learned, thereby improving the final CSI estimation accuracy. The network architecture is as follows: Figure 2 shown.
[0142] In this network, the CSI channel group As input, where T is the number of CSI in the channel group i = 1, 2, ..., N sample , where N sampleis the number of samples. The output of the network is expressed as For the sake of convenience, the present invention omits the subscript i and and Expressed as and At present, the input data of deep neural networks are all real numbers. Therefore, two real number matrices are used to represent the real and imaginary parts of the input channel matrix. Then all elements in the matrix are normalized to [0, 1], which is equivalent to the image data of two channels, which is convenient for training. Is a size of LN Bs ×1 complex vector. Therefore, the real and imaginary parts of the complex vector are separated into two parts of size LN BS ×1 real vector, and then transform it into two vectors of size The real matrix of Therefore, the data of each layer of the network is a four-dimensional tensor, and the sum of the products on the left side of each layer represents the size of the CSI channel group. The number above is the number of feature maps.
[0143] In fact, the D CNN -LSTM network consists of three layers, mainly including channel feature extraction, time-varying channel correlation extraction and CSI reconstruction. The main components are as follows:
[0144] (1) Channel feature extraction: This is mainly done by the CNN network in the first layer. In this process, each matrix of the time dimension can be convolved separately to extract features. The convolution layer contains 64 convolution kernels of size 3×3 with a sliding step size of l. The same zero padding around the input is used during convolution to keep the feature map output by each layer of the network the same size as the original channel matrix, and ReLU is used as the activation function.
[0145] (2) Time-varying channel correlation extraction: The extraction of time correlation is mainly completed by the second-layer LSTM network. At present, most LSTM networks that predict time states use a fully connected structure, but the input of the fully connected network is one-dimensional and does not consider spatial correlation. Compared with traditional methods, the convolution operation used in the ConvLSTM network can obtain better space-time relations because ConvLSTM, like LSTM, uses the output of the previous layer as the input of the next layer. ConvLSTM network is like Figure 3 As shown, the network consists of an input gate i t and an output gate O t The input gate is used to record the current state information, and the output gate is used to control how much current state information can be seen by the external network. t Control the number of historical state information flows to the current state after allowing entry, memory unit Ct The information of the previous moment is stored. As shown in formula 16, where W xi , W hi , W ci , W ho , W xo , W co , W xf represents the weight matrix, b i , b o , b f , b c Represents bias, and the number in the range [0, 1] represents the proportion of information transmission, 0 means no transmission, and 1 means all transmission.
[0146] The initially estimated CSI matrix is used as the input of the ConvLSTM layer in chronological order. At each time step, the ConvLSTM network can fuse the time-related information learned at the previous time point into the input of the current time step, and the relevant information is updated with the time step.
[0147]
[0148] (3) CSI reconstruction: In order to obtain the same size as the input data, consider using the same padding, ReLU activation function and 3×3×2 filter size. Finally, the final CSI estimation result is obtained through dimension transformation and denormalization.
[0149] Through the detailed description of the above specific implementation methods, it can be seen that the present invention can effectively perform channel estimation for millimeter wave FDD massive MIMO systems. Those skilled in the art can adjust and optimize the above steps and structures in combination with practical applications to achieve all or part of the technical effects of the present invention.
[0150] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A channel estimation method for millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning, characterized in that: Here are the steps: S1. Compressed sensing preliminary estimation: The received pilot signal is processed using the compressed sensing method to extract the preliminary CSI matrix; S2, deep learning network structure: using a combination of convolutional neural network and long short-term memory network, extracting spatial features through convolutional neural network and capturing temporal correlation through long short-term memory network; S3, ConvLSTM network processing: The initially estimated CSI matrix is input into the ConvLSTM layer in chronological order. The ConvLSTM network integrates time-related information to improve the accuracy of CSI estimation; S4, CSI reconstruction: Use the same padding, ReLU activation function and appropriately sized filters to obtain the same size as the input data; obtain the final CSI estimation result through dimensionality transformation and denormalization.
2. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 1, characterized in that: It is executed through a large-scale MIMO-OFDM system in millimeter-wave frequency division duplex mode. In the large-scale MIMO-OFDM system, communication is carried out between a base station and multiple single-antenna users, a comb pilot method is adopted, the pilot interval within each OFDM symbol is equal, and the pilot symbols are randomly selected.
3. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 2, characterized in that: In the massive MIMO-OFDM system, the following steps are performed: D1. Pilot symbol generation: randomly select the pilot symbol at each antenna from the OFDM symbol, and obtain the transmission symbol in the time domain through inverse discrete Fourier transform; D2. Signal reception: At the receiving end, the received signal is represented in the frequency domain through Fourier transform; D3. Channel state information extraction: The user extracts pilot symbols from the received signal and recovers the sparse CSI matrix through compressed sensing method.
4. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 2, characterized in that: The pilot C of the i-th transmitting antenna of the k-th user ik ∈C P×1 , where P is the number of pilots, and the pilot symbol at the i-th antenna is C i ∈C P×1 , where C i The elements in are from OFDM symbols Randomly selected from; through inverse discrete Fourier transform, the transmission symbol in the time domain is obtained as: in is the DFT matrix, (·) H is the complex conjugate transpose, N c is the number of subcarriers; the i-th antenna at the BS transmits the time domain signal X i ; Then the signal y received in the time domain is expressed as follows: where h i ∈C L×1 is the channel state information from the ith antenna of the BS to the user, L is the channel length, represents independent and identically distributed (iid) additive white complex Gaussian noise, * represents the convolution operation, also expressed as matrix multiplication; If the OFDM system has a cyclic prefix of appropriate length, the linear convolution in equation (2) becomes a circular convolution after the cyclic prefix is removed at the receiving end; i Using zero padding, the convolution operation will be expressed as follows: in From L to N c The elements are all 0, and the convolution in formula (3) can also be expressed as the following matrix multiplication: in is a Toeplitze matrix; as follows: In formula (4), From y o To N c The cumulative sum of the sequence; and the circular convolution in formula (3) is represented by matrix convolution; Through DFT transformation, the symbol received in the frequency domain can be expressed as in The diagonal elements are Matrix of In order to estimate the channel state information h i , the user must receive the signal y f Extract pilot symbols The received signal can be written as follows: in is from the index set Ω Selected from F Ω,L ∈C P×L The corresponding is a partial Fourier matrix, the rows are selected from the pilot indices, and the columns are selected from the first L columns of F; obviously, Equal to c i ;therefore in is the CSI from the ith base station antenna to the user, It means that the mean is zero and the variance is Additive Gaussian white noise; if the pilot intervals within the OFDM symbol are equal and the pilot symbols are randomly selected comb pilots, the pilot received from a certain user is in Formula (10) is regarded as a compressed sensing problem, Φ is the sensing matrix, and h is the sparse signal; therefore, the reconstruction algorithm based on compressed sensing is used to obtain the value of Φ and y c In this channel model, the uplink channel and the downlink channel have the same signal-to-noise ratio, and the noise parameter represents the total noise power of the downlink and uplink channels; for block fading time-varying MIMO channels, the CSI varies from one time slot to another, but remains unchanged within a time slot; the dynamic channel is modeled by the change of CSI support and the evolution of the amplitude of non-zero elements: Among them, h t Yes, the CSI at the moment, s t (l)∈{0,1} represents s t Whether the lth support index of is zero, g t (l)∈C is g t The amplitude at the lth support index of Represents the Hadamard product; specifically, a first-order Markov process is used to model the change in support, that is, two transition probabilities are defined as: Among them, In order to achieve a steady-state sparsity rate p∈(0,1), the amplitude adopts a first-order autoregressive model, which is expressed as follows: The correlation coefficient ρ = J0(2πf d τ) is given by the zero-order Bessel function of the first kind, f d is the maximum Doppler frequency, τ is the duration, and the parameters is a complex Gaussian variable iid.
5. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 1, characterized in that: The following steps are also included: The AS-JOMP algorithm based on compressed sensing is used to recover CSI from the received signal. According to the structural sparsity, the AS-JOMP algorithm combines the sensing matrix Φ with the remaining observation vector r i The cross-correlation vector of the product is divided into N BS The vectors are summed and then the maximum value indicator like OMP is selected; in the second observation, two threshold parameters are used to stop the iteration under high SNR and low SNR.
6. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 1, characterized in that: The following steps are also included: The estimated CSI is used as the input of DCNN-LSTM to remove noise and learn the features between time-varying channels; the DCNN-LSTM is a joint network based on CNN and LSTM, in which the CSI channel group As input, where T is the number of CSI in the channel group i = 1, 2, ..., N sample , where N sample is the number of samples; the output of the network is expressed as Obtained by AS-JOMP algorithm Is a size of LN BS ×1 complex vector; therefore, separate the real and imaginary parts of the complex vector into two LN BS ×1 real vector, and then transform it into two vectors of size The real matrix of Therefore, the data of each layer of the network is a four-dimensional tensor. The sum of the products on the left side of each layer represents the size of the CSI channel group; the number above is the number of feature maps.
7. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 6, characterized in that: The DCNN-LSTM includes a channel feature extraction module, a time-varying channel correlation extraction module and a CSI reconstruction module.
8. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 7, characterized in that: The channel feature extraction module is completed by the CNN network in the first layer. In this process, each matrix of the time dimension can be convolved to extract features separately. The convolution layer contains 64 3×3 The convolution kernel is of size , and the sliding step is 1; the same zero padding around the input is used during convolution, so that the feature map output by each layer of the network remains the same size as the original channel matrix, and ReLU is used as the activation function.
9. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 6, characterized in that: The time-varying channel correlation extraction module: The extraction of time correlation is mainly completed by the second layer LSTM network; the ConvLSTM network consists of an input gate i t and an output gate O t The input gate is used to record the information of the current state, and the output gate is used to control how much information of the current state can be seen by the external network; the forget gate f t Control the number of historical state information flows to the current state after allowing entry, memory unit C t The information of the moment before the book is saved; As shown in formula 16, where W xi , W hi , W ci , W ho , W xo , W co , W xf represents the weight matrix, b i , b o , b f , b c Represents bias, and the number in the range [0, 1] represents the proportion of information transmission, 0 means no transmission, and 1 means full transmission; The initially estimated CSI matrix is used as the input of the ConvLSTM layer in chronological order; at each time step, the ConvLSTM network can fuse the time-related information learned at the previous time point into the input of the current time step, and the relevant information is updated with the time step; 10. The method for channel estimation of a millimeter wave frequency division duplex massive MIMO system based on compressed sensing and deep learning as claimed in claim 6, characterized in that: The CSI reconstruction module: uses the same padding, ReLU activation function and 3×3×2 filter size to obtain the same size as the input data; obtains the final CSI estimation result through dimensionality transformation and denormalization
Citation Information
Patent Citations
Channel estimation method based on compressed sensing and deep learning, medium and equipment
CN112615801A
Compressed sensing FDD large-scale MIMO system sparse channel estimation method and system based on deep learning
CN113595941A
SAR three-dimensional rotating target rotating speed estimation method based on CV-ConvLSTM
CN114114263A
Compressed sensing and deep learning assisted MIMO beam space channel estimation method
CN117640302A
CSI compression feedback method based on deep learning
WO2024045689A1
Cited By
Channel estimation with varying numbers of transmit layers
US12706780B2
