阅读说明：本技术 一种用于毫米波mimo混合预编码系统的信道估计方法 (Channel estimation method for millimeter wave MIMO hybrid precoding system ) 是由 方舒 王子豪 何志强 陈权 于 2021-07-30 设计创作，主要内容包括：本发明属于通信技术领域,具体来说涉及一种用于毫米波MIMO混合预编码系统的信道估计方法。本发明针对在混合预编码系统架构下由于基站天线数量远大于RF链数量从而使信道矩阵收到压缩这一限制而无法采用常规方法进行信道估计这一问题,提出了利用毫米波MIMO信道稀疏特性的一种信道估计方法。本发明采用压缩感知的方法对经压缩的信道矩阵进行重构；经仿真验证提出的信道估计算法性能较好,估计得到的信道矩阵能够用于混合预编码。(The invention belongs to the technical field of communication, and particularly relates to a channel estimation method for a millimeter wave MIMO hybrid precoding system. The invention provides a channel estimation method utilizing the sparse characteristic of a millimeter wave MIMO channel, aiming at the problem that the channel estimation can not be carried out by a conventional method because the number of base station antennas is far greater than the number of RF chains under the framework of a hybrid precoding system and the channel matrix is limited by compression. The invention adopts a compressed sensing method to reconstruct a compressed channel matrix; the channel estimation algorithm provided by simulation verification has good performance, and the estimated channel matrix can be used for hybrid precoding.)

1. A channel estimation method for millimeter wave MIMO mixed pre-coding system, the base station in the millimeter wave MIMO mixed pre-coding system has N_tRoot antenna, base station configuration N_RFA root RF chain, and has N_t＞N_RFThe mobile station has N_rThe mobile station transmits uplink sounding signals to carry out channel estimation; obtaining the uplink channel matrix estimated value by the uplink receiving signal yThe channel estimation method comprises the following steps:

s1, performing channel estimation by using the uplink sounding signal, and establishing a model as follows:

y＝H_upx_p+n

wherein H_upIs an uplink channel matrix with dimension N_t×N_r；x_pIs an uplink sounding signal with dimension N_rX 1; n is additive Gaussian noise; y is the signal received at the base station antenna and has a dimension N_t×1；

S2, defining a compression matrix G with the dimension of N_RF×N_tThe uplink channel estimate is expressed as:

Gy＝GH_upx_p+Gn

s3, transmitting the compressed received signal Gy and the uplink sounding signal x_pEstimating to obtain the estimated value of the compressed uplink channel matrix by using LS criterionWhereinIs defined as:

estimation value of uplink channel matrix by utilizing millimeter wave channel sparsityExpressed as:

wherein U is a sparse transform matrix and U is,for the estimation value of the beam domain channel matrix, the channel estimation problem is converted into:

s4, solving the sparse channel estimation problem of step S3, specifically: defining a sensing matrix a ═ GU,the estimation value of the compressed uplink channel matrix is obtained, and the signal sparsity of the estimation value is K; the iteration number is t and t is initialized to 1, and the_tIndex value set merging initialization representing t previous iterationsλ_tIndex value, A, representing the qualifying condition found in the t-th iteration_tSet of column vectors representing matrix A selected from the set of index values for the first t iterations and initializationInitializationThe following iterative process is performed:

S41、

S42、

S43、

S44、

s45, if t is equal to t +1, returning to the step S42 if t is equal to or less than K, otherwise, stopping iteration and entering the step S46;

s46, after the iteration is finished, the reconstruction is obtained

S5 passing beam domain channel matrixAn estimated value of the uplink channel matrix is obtained and is expressed as:

Technical Field

The invention belongs to the technical field of communication, and particularly relates to a channel estimation method for a millimeter wave MIMO hybrid precoding system.

Background

Signal processing in millimeter wave MIMO systems is often constrained by a number of practical issues, as conventional all-digital MIMO systems typically perform all-digital precoding at the base station, which needs to freely control the amplitude and phase of the signal, so all-digital precoding systems require the number of transmit days and RF chains of the base station to be equal. Because a millimeter wave MIMO system usually employs a very large number of antennas at a base station, in order to save the number of RF chains and reduce the hardware complexity of the system, a hybrid precoding structure is usually employed, that is, precoding is implemented by dividing precoding into digital precoding and analog precoding. The hybrid precoding technology in the millimeter wave MIMO system can effectively reduce the configuration quantity of RF chains and reduce certain cost on the premise that the performance loss can be accepted, and is a compromise scheme of system performance and hardware overhead.

In order to perform precoding operation efficiently, the base station should ideally acquire complete channel information. In a conventional all-digital precoding system, a base station can acquire uplink channel information by receiving a sounding signal sent by a mobile station through an uplink channel, and obtain downlink channel information according to channel reciprocity, thereby performing precoding operation.

The structure of the all-digital precoding system is briefly analyzed according to fig. 1. N generated by preamble link_sAfter being processed by a digital precoder, the data stream goes to an RF chain for carrier frequency up and the like, and is transmitted into space from an antenna connected with the RF chain in the form of electromagnetic waves. It should be noted that, in the all-digital precoding structure, a method is adopted in which one RF chain is represented, and obviously, this time:

N_RF＝N_t

wherein N is_tIndicates the number of base station antennas, N_RFRepresenting the number of RF chains configured by the base station. Considering that the system adopts a mode of performing channel estimation and downlink precoding (utilizing channel reciprocity), the system needs a part of or all antennas of a mobile station to transmit uplink sounding signals, so that the method comprises the following steps:

N_r≥N_sounding

wherein N is_rIndicating the number of mobile station antennas, N_soundingIndicating the number of upstream sounding signal streams. The above equation is true if and only if the equal sign is true when all antennas of the mobile station transmit the uplink sounding signal, and the base station can acquire complete channel information at this time. Meanwhile, the base station needs to be able to correctly process these signals, and needs to satisfy the following conditions:

N_RF≥N_sounding

as for N_soundingAnd N_sDue to the base station needing to transmit N_sStreaming data, so at least N is required_sThe subchannel, and each stream sounding signal can only be used to estimate subchannel information between the mobile station antenna transmitting the signal and all antennas of the base station:

N_sounding≥N_s

wherein N is_sIndicating the number of data streams transmitted by the base station. In summary, in the all-digital precoding system, the above-mentioned system configuration parameters should satisfy:

N_t＝N_RF≥N_r≥N_sounding≥N_s

the biggest difference from the all-digital precoding structure is that, in order to save the RF chain and reduce the hardware complexity, the hybrid precoding system adopts a combination of digital precoding and analog precoding, and thus:

N_t＞＞N_RF

wherein N is_tIndicates the number of base station antennas, N_RFRepresenting the number of RF chains configured by the base station. To perfectly estimate the actual channel, each antenna of the user is required to transmit a sounding signal (which has been explained in the analysis of the structure of the all-digital precoding system), and each antenna of the base station receives the sounding signal and processes the sounding signal through the RF chain. However, since the number of base station antennas is much larger than the number of RF chains in the hybrid precoding structure, and the sounding signal received by the base station antennas is necessarily compressed, the channel information obtained in this way is reduced in dimension, and it is difficult to perform hybrid precoding based on the channel matrix.

Disclosure of Invention

Aiming at the problems, the invention provides a channel estimation method for a millimeter wave MIMO hybrid precoding system, which solves the problem that the channel estimation can not be carried out due to the limitation of an RF chain because an uplink sounding signal is used for carrying out the channel estimation.

For ease of understanding, the techniques employed by the present invention are described below:

millimeter wave channel model:

millimeter wave frequency band has become one of the hot problems of academic research, millimeter wave frequency is high, band bandwidth is abundant, its wavelength is short, communication equipment is small, makes the small-size encapsulation of large-scale antenna array become possible. However, the transmission process is susceptible to air molecules, dust, smoke, etc. to cause severe path loss, and a large-scale antenna array configuration is required to obtain sufficient gain to combat the multipath effect, and the large-scale antenna configuration results in a large antenna correlation, so that the number of paths is much smaller than the number of antennas. Meanwhile, the millimeter wave channel is sparse, only the line-of-sight channel can obtain better energy embodiment, and the performance of other scattering paths approaches zero. According to the sparse characteristic of a millimeter wave channel space, a Saleh-Valencuela channel model (S-V channel model) with a narrow-band cluster channel based on angle expansion is more matched with millimeter wave communication relative to Rayleigh distribution of a traditional MIMO channel.

In the millimeter wave MIMO system, modeling is performed according to an S-V model, and a channel matrix H can be expressed as:

wherein N is_tIndicates the number of base station antennas, N_rIndicating the number of mobile station antennas, N_clRepresenting the number of scattering clusters, N_rayRepresenting the number of paths, alpha, in each scattering cluster_ilA complex gain factor representing the 1 st transmission path in the ith scattering cluster subject to a mean of 0 and a variance ofThe complex gaussian distribution of (a) is,andrespectively representing receiving-end antennasAnd an array response vector of the transmitting end antenna,andangle of arrival (AOA) and angle of departure (AOD) respectively representing azimuth and declination angle ()^HRepresenting the conjugate transpose of the matrix.

Compressed sensing principle:

the compressed sensing theory mainly relates to three aspects, namely sparse representation of signals, design of a measurement matrix and construction of a reconstruction algorithm. Only a few elements of the sparse signal are non-zero or only a few elements of the signal are non-zero in a certain transform domain. The space required to store this space can be reduced for compression purposes if only these non-zero data are retained and the other coefficients are discarded, so that the original signal can be reconstructed with a high probability by these coefficients. If the signal x is sparse in a certain transform domain, it can be expressed as: can be represented by a set of orthogonal basis linear combinations:

where psi ═ psi₁，ψ₂，…ψ_NDenotes an orthogonal base, s_iIs a projection sparseness (known from sparsity of signal s) corresponding to an orthogonal basis_iContaining only a few non-zero numbers). The perceptual signal y can thus be expressed as:

y＝φx＝φΨs

phi is an observation matrix, psi is a sparse representation matrix, and commonly used measurement matrices include a Gaussian random matrix and the like.

Regarding the construction of the reconstruction algorithm, since it is an underdetermined problem to recover the original signal with a small number of signal components, the most optimized problem is generally adopted for solving.

The technical scheme of the invention is as follows:

1. mixing for millimeter wave MIMOChannel estimation method of precoding system, base station in millimeter wave MIMO mixed precoding system having N_tRoot antenna, base station configuration N_RFA root RF chain, and has N_t＞N_RFThe mobile station has N_rThe mobile station transmits uplink sounding signals to carry out channel estimation; obtaining the estimated value of the uplink channel matrix by the uplink receiving signal yThe channel estimation method comprises the following steps:

s1, performing channel estimation by using the uplink sounding signal, and establishing a model as follows:

y＝H_upx_p+n

wherein H_upIs an uplink channel matrix with dimension N_t×N_r；x_pThe uplink sounding signal is an uplink sounding signal, and the dimensionality of the uplink sounding signal is Nr multiplied by 1; n is additive Gaussian noise; y is the signal received at the base station antenna and has a dimension N_t×1；

S2, defining a compression matrix G with the dimension of N_RF×N_tThe uplink channel estimate is expressed as:

Gy＝GH_upx_p+Gn

s3, transmitting the compressed received signal Gy and the uplink sounding signal x_pEstimating to obtain the estimated value of the compressed uplink channel matrix by using LS criterionWhereinIs defined as:

estimation value of uplink channel matrix by utilizing millimeter wave channel sparsityExpressed as:

wherein U is a sparse transform matrix and U is,for the estimation value of the beam domain channel matrix, the channel estimation problem is converted into:

s4, solving the sparse channel estimation problem of step S3, specifically: defining a sensing matrix a ═ GU,the estimation value of the compressed uplink channel matrix is obtained, and the signal sparsity of the estimation value is K; the iteration number is t and t is initialized to 1, and the_tIndex value set merging initialization representing t previous iterationsλ_tIndex value, A, representing the qualifying condition found in the t-th iteration_tSet of column vectors representing matrix A selected from the set of index values for the first t iterations and initializationInitializationThe following iterative process is performed:

S41、

S42、Λ_t＝Λ_t-1∪λ_t，

S43、

S44、

s45, if t is equal to t +1, returning to the step S42 if t is equal to or less than K, otherwise, stopping iteration and entering the step S46;

s46, after the iteration is finished, the reconstruction is obtained

S5 passing beam domain channel matrixAn estimated value of the uplink channel matrix is obtained and is expressed as:

the invention has the beneficial effects that aiming at the problem that the channel estimation can not be carried out by a conventional method because the number of the base station antennas is far more than the number of the RF chains under the framework of a hybrid precoding system, the channel matrix is limited by compression, and the channel estimation method utilizing the sparse characteristic of the millimeter wave MIMO channel is provided. The invention adopts a compressed sensing method to reconstruct a compressed channel matrix; the performance of the channel estimation algorithm provided by simulation verification is good, and the performance of a system for performing hybrid precoding by using the estimated channel matrix is closer to the performance of a system for performing hybrid precoding by using a perfect channel matrix.

Drawings

FIG. 1 is a schematic diagram of an all-digital precoding structure;

FIG. 2 is a schematic diagram of a hybrid precoding structure;

FIG. 3 is a schematic diagram of compressed sensing theory;

FIG. 4 shows channel estimation based on input K values (non-actual sparsity);

FIG. 5 shows the variation of the residual error between the estimated value and the actual value of the channel matrix with the number of RF chains under different K values;

FIG. 6 is an achievable rate for hybrid precoding under different channel estimation schemes;

fig. 7 shows the performance of precoding error rates under different channel estimation schemes.

Detailed Description

The technical scheme of the invention is described in detail in the following with the accompanying drawings:

the problem of channel estimation by using uplink sounding signals under a millimeter wave MIMO mixed precoding structure is modeled as follows:

y＝H_upx_p+n

wherein H_upIs an uplink channel matrix with dimension N_t×N_r；x_pIs an uplink sounding signal with dimension N_rX 1; n is additive Gaussian noise; y is the signal received at the base station antenna and has a dimension N_t×1。

A compression matrix G with dimension N is additionally provided_RF×N_t. According to the analysis of the hybrid pre-coding system, N is caused by the base station adopting the hybrid architecture_t＞N_RFWhen receiving sounding signals, a weight is added to each antenna path for combination and compression, and then the signals are transmitted to an RF chain for processing, so that the compression and combination relationship of each stream of sounding signals are stored in a compression matrix G. According to the actual hybrid precoding structure, the specific implementation manner of the compression matrix G is realized by a phase shifter network connected with the base station antenna, so that this limitation should be considered when designing the compression matrix G, which is embodied as that the compression matrix can only adjust the signal phase but not the amplitude.

Up to now, the whole uplink channel estimation can be expressed as:

Gy＝GH_upx_p+Gn

it can be seen that due to pressureExistence of reduced matrix, N received by base station antenna_tThe signal is compressed into N after passing through a phase shifter network_RFA signal. The base station end can receive the signal Gy and the known pilot frequency sequence x_pObtaining an estimate of a compressed uplink channel matrix by LS criterion estimationHere, the Representing an estimated value of an uplink channel matrix; the problem now turns into how to go through what is knownSolving for unknowns by the sum matrix G

The millimeter wave channel can exhibit sparsity in a beam domain (angle domain). The estimated value of the uplink channel matrix is then determinedThe sparse representation is:

wherein U is a transformation matrix, and U is,is an estimate of the beam domain channel matrix. According to the compressed sensing principle, the signal estimation problem is converted into a sparse signal reconstruction problem: the observations being compressed upstream channel matricesThe observation matrix being designedCompression matrix G, sparse representation of the signal asThe channel estimation problem is expressed as:

the preparation of signal reconstruction is completed, and then sparse signal reconstruction is performed according to the above formula and a specific algorithm.

The channel estimation problem is converted into a sparse signal reconstruction problem, and symbols are firstly explained as follows:

(1)for the compressed upstream channel matrix, defined as the observation of the reconstruction algorithm

(2)Is an estimated value of a wave beam domain channel matrix and is defined as a to-be-recovered quantity of a reconstruction algorithm

(3)Is K sparse; (typically K is unknown at the time of channel estimation);

(4) g is an observation matrix (compression matrix) with dimension N_RF×N_t；

(5) U is a sparse transform matrix and has a dimension of N_t×N_tAccording to the sparse characteristic of the millimeter wave MIMO channel matrix wave beam domain, the method takesWherein U is₀Is N_tOrder DFT matrix;

(6) a is a sensing matrix, A is defined as GU, and the dimension is N_RF×N_t；

The orthogonal matching pursuit sparse signal reconstruction algorithm is as follows:

inputting: a sensing matrix A is GU; compressed uplink channel matrixSignal sparsity K;

and (3) outputting: estimation of a beam domain channel matrix(K sparse); residual r_K；

The algorithm flow is as follows:

in the following flow, t denotes the number of iterations, Λ t denotes the set of index values of the previous t iterations, λ_tIndex values representing the found matching conditions in the t-th iteration; a is_iRepresents the ith column of matrix A; a. the_tRepresenting a set of column vectors of the matrix A selected by the previous t iterations according to the index value set;<，>the inner product of the vector is represented as,indicating an empty set.

Initialization: t is 1;

(1)

(2)Λ_t＝Λ_t-1∪λ_t，

(3)

(4)

(5) if t is equal to or less than K, returning to the step (2), otherwise, stopping iteration and entering the step (6);

(6) is reconstructed to obtain

According to the compressed sensing principle, the channel estimation process can be expressed asThe downlink channel is represented by the channel reciprocity as:

the estimation value of the downlink channel matrix can be obtainedAnd performs a hybrid precoding operation based thereon.

The practical application of the present invention is demonstrated by the following simulation example

In the orthogonal matching pursuit algorithm, K should be the original sparsity of the signal; however, as the millimeter wave channel inevitably has components with smaller power in a certain number of angles during generation and is not strictly sparse, the actually input K value in the simulation is a value smaller than the actual sparsity; according to the orthogonal matching pursuit principle, the sparsity of the output estimation signal is the same as the input K value, so that the obtained signal is a K sparse signal on the premise that the input K is smaller than the actual sparsity, and fig. 4 shows the process. In fig. 4, the left graph represents the gain value of the actual channel, and the right graph represents the gain value (K-sparse) of the estimated channel restored after the K value is artificially input.

Fig. 5 verifies the accuracy of the channel estimation algorithm itself by simulation. Fig. 5 shows the variation of the channel matrix estimation value and the actual value residual with the number of RF chains in the case where the input value K is {10, 20, 30, 40 }. The residual value K10 is the smallest value compared to several other values K in the case of a small number of RF chains, but the actual channel estimation performance is still poor because the residual value itself is still large. With the gradual increase of the number of the RF chains, the channel estimation scheme with the larger K value can show advantages; taking 80 RF chains as an example, taking the K value of 20 can minimize the residual value. The K value selection in the subsequent simulation is carried out according to the simulation, namely the K value which enables the residual error between the channel matrix estimation value and the actual value to be minimum under the current RF chain number is selected for simulation.

On the basis of fig. 5, K is 20, and fig. 6 simulates the OMP precoding and N precoding with the hybrid precoding structure in the case of perfect channel estimation and full-digital OMP precoding in the case of perfect channel estimation_RFThe achievable rates of the five schemes of OMP precoding under the mixed structure after channel estimation by the proposed algorithm are {40, 80, 120 }; the number of transmitting antennas is 256, the number of users is 1, the number of receiving antennas per user is 4, and the number of streams per user is 4.

Fig. 7 shows the bit error rate of the link-level system, which is obtained by performing bit error rate simulation using different channel estimation schemes and precoding schemes. The transmitting days are 256, the number of users is 1, the number of receiving antennas of each user is 4, the number of streams of each user is 4, the modulation mode is 256QAM, and the equalization mode is MMSE-IRC equalization. According to fig. 7, compared with the perfect channel estimation scenario, at the RF chain number of 40, the error rate of the system is poor; when the number of the RF chains is 80, the error rate performance of the system is obviously improved and is relatively close to the error rate under the condition of adopting perfect channel estimation, and the configuration of the number of the RF chains can meet the normal mixed precoding operation. At a number of RF chains of 120, there is some improvement in bit error rate performance, but the improvement is not large.