A method for channel estimation enhancement in a millimeter wave MIMO-OFDM system

Abstract

The present application relates to a kind of millimeter wave MIMO-OFDM system channel estimation enhancement method, focus on solving the prior information of stable path that existing method cannot effectively utilize, leading to unnecessary calculation overhead and precision loss problem in estimation process.Its implementation steps are:1) for each subcarrier, base station (BS) is equipped with different beamforming vector in continuous time frame, user (UE) independently detects transmission signal in each time subframe;2) separate known path information from received signal, obtain the tensor slice of prior and unknown information joint, based on parallel factor (PARAFAC) model it is constructed as three-order tensor;3) using prior information optimization alternating least squares (ALS) algorithm, estimate three factor matrices, then use maximum likelihood (ML) estimation to obtain unknown channel parameters.The channel estimation enhancement method proposed in the present application can optimize ALS algorithm using the prior information of stable path, can reduce estimation error, improve system performance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and particularly to a channel estimation enhancement method for a millimeter wave MIMO-OFDM system BACKGROUND

[0002] Millimeter wave communication is an advanced communication technology that utilizes the 30GHz-300GHz frequency band for wireless transmission. Due to the short wavelength of millimeter waves, multiple-input multiple-output (MIMO) antenna arrays can be deployed in limited space, significantly enhancing signal quality through beamforming technology. This frequency band provides abundant spectrum resources, but also faces many challenges, including high path loss, complex multipath effects, and rapid channel changes in dynamic environments, etc. Therefore, accurate channel state information (CSI) estimation is the key to its successful application.

[0003] In millimeter wave communication, the channel may change rapidly due to mobility, environmental changes, etc., causing traditional channel estimation methods to fail. In a multi-path scenario, the reflection path information from fixed objects (such as ceilings, walls, etc.) is usually relatively stable, and therefore can be used as prior information to provide a stable reference framework for channel estimation. In rapidly changing environments, prior information of known paths can help the system better track channel changes, and through joint estimation of known and unknown paths, channel estimation error can be reduced and overall system performance can be improved.

[0004] At present, there are more literatures on the estimation of CSI of millimeter wave channel. In the paper of Z. Gao, C. Hu, L. Dai and Z. Wang (Channel Estimation for Millimeter-Wave Massive MIMO With Hybrid Precoding Over Frequency-Selective Fading Channels [J]. IEEE Commun. Lett., vol. 20, no. 6, pp. 1259-1262, June 2016.), for the millimeter wave system running on the wideband channel with frequency selectivity, the method based on distributed compressed sensing is proposed by using the angular domain structure sparsity of millimeter wave wideband frequency selective fading channel. In the paper of Z. Zhou, J. Fang, L. Yang, H. Li, Z. Chen and R. S. Blum (Low-Rank Tensor Decomposition-Aided Channel Estimation for Millimeter Wave MIMO-OFDM Systems [J]. IEEE J. Sel. Areas Commun., vol. 35, no. 7, pp. 1524-1538, July 2017.), a channel parameter estimation method based on parallel factor (PARAFAC) decomposition is proposed by using the sparse scattering characteristics of millimeter wave channel. However, these studies do not fully consider the difference in path stability, and usually regard all paths as unknown paths for estimation, which fails to effectively utilize the prior information of stable paths, resulting in unnecessary calculation overhead and precision loss in the estimation process. SUMMARY

[0005] The present application aims at the defects of the prior art, and provides a millimeter wave MIMO orthogonal frequency division multiplexing (OFDM) system channel estimation enhancement method to obtain higher precision CSI.

[0006] Technical scheme: the millimeter wave MIMO-OFDM system channel estimation enhancement method comprises:

[0007] For each subcarrier, the base station (BS) is equipped with different beamforming vectors in the continuous time frame, and the user (UE) independently detects the transmission signal in each time subframe;

[0008] Separate the known path information from the received signal to obtain the tensor slice of the prior and unknown information joint, and construct it into a three-order tensor based on the PARAFAC model;

[0009] The three factor matrices are estimated by using prior information to optimize an alternating least squares (ALS) algorithm, and unknown channel parameters are estimated by using maximum likelihood (ML) estimation;

[0010] Further, the BS is equipped with different beamforming vectors in consecutive time frames for each subcarrier, and the UE independently detects the transmission signal in each time subframe, and specifically includes:

[0011] Assuming that the BS is equipped with antennas and RF chains, the UE is equipped with antennas and RF chains, the BS and the UE both use uniform linear arrays, and there are , Since channel estimation is independently performed for each UE in a downlink scenario, only one UE is considered, and each UE is equipped with one RF chain. The total number of OFDM subcarriers is set to , and the first subcarriers are assumed to be used for training. In order to simplify the problem, the interference between adjacent cells due to frequency reuse is ignored. The beamforming vector expression of the time frame on the subcarrier is:

[0012]

[0013] wherein is the pilot symbol vector on the subcarrier, is a digital precoding matrix, is a general RF precoding matrix. A geometric broadband millimeter wave channel model is considered, that is, there are scattering points between the BS and the UE, each scattering point corresponds to a path with a corresponding departure angle , an arrival angle , and a time delay , and each path has a complex path gain , so the channel matrix in the time delay domain can be written as:

[0014]

[0015] wherein is a delta function, and are the steering vectors at the UE and the BS, respectively, denotes transposition.

[0016] For a uniform linear array, the steering vector is:

[0017]

[0018] in, It is the angle of arrival or the angle of departure. It refers to the number of antennas. It is the signal wavelength. This refers to the spacing between adjacent antenna elements. It is assumed that different scattering points correspond to different departure angles, arrival angles, and time delays, i.e., when... hour, , , .

[0019] Based on the time-delay domain channel model, we can obtain the first... Frequency domain channel matrix on each subcarrier for:

[0020]

[0021] in, The sampling frequency.

[0022] Assuming in Among the paths, there are One path represents the route of the transmitted signal to the UE via a fixed scattering point (such as a building), and its channel parameter information is known; the rest... The path represents the route from the first path to the UE via an unstable scattering point (such as a vehicle). Its channel parameter information is unknown, requiring channel estimation. Since the path order does not affect channel estimation, it is assumed that the path from the first path to the second path... If the path is a known path, then:

[0023]

[0024]

[0025]

[0026] in, This is the known part of the channel matrix. For known path channel parameter index, This represents the unknown part of the channel matrix. This is an index of channel parameters for unknown paths.

[0027] Within each time frame, the UE uses sequentially A number of radio frequency (RF) combination vectors are used to detect the transmitted signal, and these RF combination vectors are universal across all subcarriers. Within each time subframe, the received signal is combined in the RF domain, and after removing the cyclic prefix, it is transformed back to the frequency domain using a discrete Fourier transform. After processing, the... On the subcarrier The received signal of the kth time sub-frame can be represented as:

[0028]

[0029] wherein represents the combined vector of the kth time sub-frame, represents the additive Gaussian noise.

[0030] Further, separating the known path information from the received signal, a tensor slice of the joint prior and unknown information is obtained, which is constructed into a third-order tensor based on the PARAFAC model, specifically including:

[0031] Arranging the received signals within each time frame, the following is obtained:

[0032]

[0033] wherein , , , and the entries in , are uniformly distributed. For and , the unknown part of the channel matrix must be estimated from the received signal . Assuming that the digital precoding matrix and the pilot matrix remain consistent on each subcarrier, i.e., for , , , this facilitates channel parameter estimation based on tensor decomposition with a minimal amount of measurement cost. Defining , , , , we have:

[0034]

[0035] The UE can receive signals from multiple subcarriers, so the received signal can be represented by a third-order tensor , and the kth entry of this tensor is . Let , , , , the slice of the tensor is obtained:

[0036]

[0037] ​​​Stacking along the subcarriers, the received signal is in the form of a third-order tensor:

[0038]

[0039] where the symbol denotes the outer product, , is the third-order tensor form of the Gaussian noise.

[0040] Further, the ALS algorithm is optimized by using the prior information to estimate the three factor matrices, and then the ML estimation is used to estimate the unknown channel parameters, which specifically includes:

[0041] Due to the sparse scattering characteristics of the millimeter wave channel, the number of paths is usually much smaller than the dimension of the tensor. Therefore, the received signal tensor has an inherent low-rank structure, which ensures that the scale ambiguity and permutation ambiguity of the PARAFAC decomposition of the received tensor are unique. Therefore, by performing PARAFAC decomposition on the received signal , the estimated value of the unknown channel parameter can be obtained.

[0042] The known part of the noise-free received signal has factor matrices in three modes respectively:

[0043]

[0044]

[0045]

[0046] The unknown part of the noise-free received signal has factor matrices in three modes respectively:

[0047]

[0048]

[0049]

[0050] Let the factor matrix to be estimated be , and its corresponding estimated value is respectively .

[0051] In order to realize the PARAFAC decomposition of the unknown path part in the received signal tensor, the problem:

[0052]

[0053] where is the true value of the unknown path part of the received signal tensor.

[0054] ALS algorithm is used to solve the above problem. Specifically, by fixing two factor matrices to minimize the fitting error of the third factor matrix, and alternating the three factor matrices to achieve effective optimization. The formula is as follows:

[0055]

[0056]

[0057]

[0058] Where, , , are the expansion forms of the three modes, respectively. The subscript represents the number of iterations of the ALS algorithm.

[0059] There is a fuzzy relationship between the estimated factor matrix and the real factor matrix as follows:

[0060]

[0061]

[0062]

[0063] Where, the unknown matrix is a non-singular diagonal matrix satisfying , represents the unit matrix, and the matrix means that the size of each column in the estimated factor matrix is scaled by its real value; the permutation matrix is also unknown, which means that the permutation order of the columns in the estimated factor matrix is different from that in the real value; and , and are the estimation errors of the three factor matrices.

[0064] Since each column of the factor matrix is related to the angle of arrival of the corresponding path, the angle of arrival can be obtained by ML estimation, that is:

[0065]

[0066] Where, represents the column of . Similar to the solution of the angle of arrival, the departure angle and time delay estimates are:

[0067]

[0068]

[0069] wherein, denotes the first column of the first column of is .

[0070] The factor matrixes and are regenerated by using the estimated and , and the matrix is constructed by using the estimated . Since , and are obtained based on the factor matrix estimation values , and , the , and have the same column vector arrangement order as , and . It is known that the arrangement matrix in the fuzzy relationship is universal and can be ignored, thus the non-singular diagonal matrixes and are obtained, and can be calculated by combining the condition . By using the constructed and the factor matrix estimation values , the channel gain can be estimated by inverse operation.

[0071] Finally, the unknown part of the channel matrix can be recovered according to the estimated , and the estimation value of the channel matrix can be obtained by combining the known path prior information.

[0072] Beneficial effects: Compared with the prior art, the main advantages of the present application are as follows: the present application can effectively utilize the prior information of stable paths, reduce the calculation cost and precision loss in the estimation process; by using the method, the CSI estimation can be enhanced, and the system performance can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1A channel estimation enhancement method for a millimeter wave MIMO-OFDM system of the present application;

[0074] Figure 2 A millimeter wave MIMO-OFDM system structure diagram of the present application is shown in

[0075] Figure 3 The performance diagrams of the channel parameter mean square error (MSE) and the corresponding Cramer-Rao bound (CRB) with respect to the signal-to-noise ratio under different known path numbers and unknown path numbers are shown in

[0076] Figure 4 The performance diagrams of the channel normalized mean square error (NMSE) with respect to the signal-to-noise ratio under different known path numbers and unknown path numbers are shown in

[0077] Figure 5 The performance diagrams of the channel parameter MSE and the corresponding CRB with respect to the number of subcarriers used for training under different known path numbers and unknown path numbers are shown in

[0078] Figure 6 The performance diagrams of the channel NMSE with respect to the number of subcarriers used for training under different known path numbers and unknown path numbers are shown in DETAILED DESCRIPTION

[0079] In order to make the features and advantages of the present application more obvious and easy to understand, the present application will be described in detail below with reference to the accompanying drawings.

[0080] Figure 2 A millimeter wave MIMO-OFDM system structure diagram of the present application is shown in Figure 2 The downlink communication system shown in The BS and the UE end both adopt a uniform linear array, the BS end is equipped with antennas, and the UE end is equipped with antennas. Due to the adverse propagation conditions, it is assumed that the direct path between the BS and the UE is highly attenuated, therefore, for the proposed millimeter wave downlink communication system, the direct path is not considered, the transmission signal is reflected by a plurality of scattering points, transmitted to the UE through the known path and the unknown path for reception.

[0081]

[0082] Please refer toFigure 3 and Figure 4 , which are performance figures of the application under different known path numbers and unknown path numbers , channel parameter MSE and corresponding CRB and channel NMSE performance figures with respect to signal-to-noise ratio. System parameters are: , , , It can be found that, under the condition of constant total path number, the more the known path number , the better the channel parameter MSE performance and channel NMSE performance. And the CRB of channel parameter decreases exponentially with the increase of signal-to-noise ratio, except that the MSE of channel parameter is very close to the corresponding CRB under low signal-to-noise ratio. In addition, under the condition of constant total path number, the curve of the more known path number is always below the curve of the less known path number , which also means that the channel estimation accuracy is higher with more prior information, and the application effectively utilizes prior information to enhance the CSI estimation.

[0083] Embodiment 2

[0084] Please refer to Figure 5 and Figure 6 , which are performance figures of the application under different known path numbers and unknown path numbers , channel parameter MSE and corresponding CRB and channel NMSE performance figures with respect to the number of subcarriers used for training. System parameters are: , , , , and the signal-to-noise ratio is set to 15dB. It can be found that, the more the known path number , the less the unknown path number , the better the channel parameter MSE performance and channel NMSE performance. And except that the MSE of channel parameter is very close to the corresponding CRB under a small number of training subcarriers. In addition, under the condition of constant total path number, the more information can be used for estimation with the increase of the number of subcarriers used for training, and the curve of the more known path number is always below the curve of the less known path number The less the curve under the area, the higher the channel estimation accuracy of the prior information, and the application effectively utilizes the prior information to enhance the CSI estimation.

[0085] In conclusion, the application considers a kind of channel estimation enhancement method of millimeter wave MIMO-OFDM system, pilot is sent by BS, is received by UE end after reflection by a plurality of scattering points, and with the aid of optimized ALS fitting algorithm, effectively utilize prior information to enhance the CSI estimation.

[0086] The above embodiments are only for helping to understand the method and the main idea of the application. The content of the specification cannot limit the scope of protection of the application, therefore, the protection scope of the application should be subject to the appended claims.

Claims

1. A method for channel estimation enhancement in a millimeter wave MIMO-OFDM system, characterized in that: wherein f0 is the sampling frequency, it is assumed that in L paths, n paths are the paths of the transmission signal reaching the UE through fixed scattering points, the channel parameter information of which is known, and the remaining L-n paths are the paths of the transmission signal reaching the UE through unstable scattering points, the channel parameter information of which is unknown and needs to be estimated, and since the order of the paths does not affect the channel estimation, it is assumed that the first to the nth paths are known paths, so that For each subcarrier, the base station BS is equipped with different beamforming vectors in successive time frames, and the user UE independently detects the transmission signal in each time subframe, which specifically includes: assuming that the BS is equipped with N BS root antennas and R BS strip radio frequency chains, the UE is equipped with N UE root antennas and R UE strip radio frequency chains, the BS and the UE both use uniform linear arrays, and N BS >R BS , N UE >R UE , in the downlink scenario, the channel estimation is independently performed for each UE, so only one UE is considered, and each UE is equipped with one radio frequency chain, and the total number of orthogonal frequency division multiplexing OFDM subcarriers is set to Assuming that the first K subcarriers are selected for training, the beamforming vector expression of the t-th time frame on the k-th subcarrier is wherein, is the pilot symbol vector on the kth subcarrier, is the digital precoding matrix, is the general radio frequency precoding matrix, considering the geometric wideband millimeter wave channel model, i.e. there are L scattering points between the BS and the UE, each of which has a corresponding path with a corresponding departure angle φ l ∈ [0, 2π], an angle of arrival θ l ∈ [0, 2π] and a time delay τ l , and each path has a complex path gain g l , so the channel matrix in the time delay domain is where δ(·) is the delta function, κ UE (θ l ) and κ BS (φ l ) are the steering vectors at the UE and BS, respectively, (·) T denotes the transpose, and Hk(θ, φ) is the frequency domain channel matrix at the kth subcarrier, based on the delay domain channel model, assuming that the departure angle, the arrival angle and the time delay corresponding to different scattering points are different. is The known path information is separated from the received signal to obtain a tensor slice of the joint prior and unknown information, which is constructed into a third-order tensor based on a parallel factor (PARAFAC) model, specifically including: arranging S received signals in each time frame to obtain wherein, is the known part of the channel matrix, is the unknown part of the channel matrix, in each time frame, the UE uses S radio combining vectors in sequence to detect the transmitted signal, and these radio combining vectors are common for all subcarriers, the received signal in the s-th time subframe on the k-th subcarrier is denoted as wherein denotes the combined vector of the s-th time subframe, n k,s (t) denotes an additive Gaussian noise; The alternating least squares (ALS) algorithm is optimized using the prior information to estimate three factor matrices, and the unknown channel parameters are estimated using maximum likelihood (ML) estimation, specifically including: the factor matrices of the known part of the noise-free received signal in three modes are respectively wherein For and The unknown part of the channel matrix has to be estimated from the received signal y k (t) in the presence of noise n It is assumed that the digital precoding matrix and the pilot matrix remain constant on each subcarrier, i.e. for there is R k (t) = R(t), z k (t) = z(t), which facilitates channel parameter estimation based on tensor decomposition with a minimal amount of measurement cost, respectively defining then there is The signals from the plurality of subcarriers are received at the UE, so the received signals are represented by a third order tensor , and the (s, t, k)th entry of this tensor is y k,s (t), let The slice of the tensor is obtained Y k Stacking over subcarriers, resulting in a third-order tensor form of the received signal where the symbol denotes the outer product, is a third order tensor form of the Gaussian noise; The factor matrices of the unknown part of the noise-free received signal in three modes are respectively ​ Let the factor matrix to be estimated be Ω1= [η1... η L-n ], Ω2= [ρ1... ρ L-n ], and their corresponding estimates be To achieve the PARAFAC decomposition of the unknown path part of the received signal tensor, the problem where, To receive the signal tensor unknown path part of the true value, the ALS algorithm is used to solve the above problem, the formula is as follows where Y (1) , Y (2) , Y (3) are The unfolding of the three modes, the superscript j denotes the ALS algorithm iteration number, and the estimated factor matrix and the true factor matrix have the following fuzzy relationship where unknown matrix {Θ1, Θ2, Θ3} is a non-singular diagonal matrix satisfying Θ1Θ2Θ3=I, I represents a unit matrix, permutation matrix P is unknown, ζ1, ζ2, and ζ3 are estimation errors of the three factor matrices, since each column of factor matrix Ω1 corresponds to the arrival angle θ v Therefore, the arrival angle θ v is obtained by ML estimation. wherein, denotes the (v - n)th column of the matrix similar to the angle of arrival estimation, the angle of departure and time delay estimates are wherein, the (v-n)th column of the (v-n)th column of the (v-n)th column of is reconstructs the factor matrix and reconstructs the factor matrix and reconstructs the factor matrix and Since and are obtained based on the factor matrix estimates and , so and have the same column vector arrangement order as and , and since the arrangement matrix P in the ambiguity relation is universal and is uniformly ignored, the non-singular diagonal matrices Θ1 and Θ2 are obtained, and Θ3 is calculated in combination with the condition Θ1Θ2Θ3 = I, and the constructed and the factor matrix estimates are used to estimate the channel gain by inverse operation Finally, the unknown part of the channel matrix is recovered according to the estimated , and the channel matrix {H k} is obtained in combination with the known path prior information.

Images