阅读说明：本技术 一种基于变分贝叶斯推断的大规模mimo系统的下行链路信道估计方法 (Downlink channel estimation method of large-scale MIMO system based on variational Bayesian inference ) 是由 周磊 曹政 戴继生 于 2019-08-19 设计创作，主要内容包括：本发明公开了一种冲击噪声环境下基于变分贝叶斯推断的大规模MIMO系统的下行链路信道估计方法,包括步骤1：基站采用具有N根天线的均匀线性阵列,下行链路中的移动用户采用单天线,设T个时刻内,基站发送导频信号矩阵X,则存在冲击噪声的情况下,移动用户接收到的信号为y＝Φ(β)s+e+w；2：建立q(s),q(e),q(α),q(ν),q(γ)和β的数据模型并初始化参数；3：设置迭代次数计数变量k＝1；4：固定q(e),q(α),q(ν),q(γ),β,更新q(s)；5：固定q(s),q(α),q(ν),q(γ),β,更新q(e)；6：固定q(s),q(e),q(ν),q(γ),β,更新q(α)；7：固定q(s),q(e),q(α),q(γ),β,更新q(ν)；8：固定q(s),q(e),q(α),q(ν),β,更新q(γ)；9：固定q(s),q(e),q(α),q(ν),q(γ),更新β；10：判断迭代计数变量k是否达到上限K或ν是否收敛,如果都不满足,则k＝k+1,并返回步骤4；11：估计最终的信道。本发明能有效改善信道估计的性能。(The invention discloses a downlink channel estimation method of a large-scale MIMO system based on variational Bayesian inference in an impulsive noise environment, which comprises the following steps of 1: a base station adopts a uniform linear array with N antennas, a mobile user in a downlink adopts a single antenna, and the base station sends a pilot signal matrix X within T moments, so that under the condition of impact noise, a signal received by the mobile user is y phi (beta) s + e + w; 2: establishing data models of q(s), q (e), q (alpha), q (nu), q (gamma) and beta and initializing parameters; 3: setting an iteration number counting variable k as 1; 4: fixing q (e), q (alpha), q (nu), q (gamma), beta, updating q(s); 5: fixing q(s), q (alpha), q (nu), q (gamma), beta, updating q (e); 6: fixing q(s), q (e), q (v), q (gamma), beta, updating q (alpha); 7: fixing q(s), q (e), q (alpha), q (gamma), beta, updating q (nu); 8: fixing q(s), q (e), q (alpha), q (nu), beta, updating q (gamma); 9: fixing q(s), q (e), q (alpha), q (nu), q (gamma), updating beta; 10: judging whether the iteration counting variable K reaches the upper limit K or the upper limit v is converged, if not, determining that K is K +1, and returning to the step 4; 11: the final channel is estimated. The invention can effectively improve the performance of channel estimation.)

1. A downlink channel estimation method of a large-scale MIMO system based on variational Bayesian inference is characterized by comprising the following steps:

step 1: a base station adopts a uniform linear array with N antennas, a mobile user in a downlink adopts a single antenna, and the base station sends a pilot signal matrix X within T moments, so that under the condition of impact noise, a signal received by the mobile user is y phi (beta) s + e + w;

step 2: establishing data models of q(s), q (e), q (alpha), q (nu), q (gamma) and beta and initializing parameters;

and step 3: setting an iteration number counting variable k as 1;

and 4, step 4: fixing q (e), q (alpha), q (nu), q (gamma), beta, updating q(s);

and 5: fixing q(s), q (alpha), q (nu), q (gamma), beta, updating q (e);

step 6: fixing q(s), q (e), q (v), q (gamma), beta, updating q (alpha);

and 7: fixing q(s), q (e), q (alpha), q (gamma), beta, updating q (nu);

and 8: fixing q(s), q (e), q (alpha), q (nu), beta, updating q (gamma);

and step 9: fixing q(s), q (e), q (alpha), q (nu), q (gamma), updating beta;

step 10: judging whether the iteration counting variable K reaches the upper limit K or the upper limit v is converged, if not, determining that K is K +1, and returning to the step 4;

step 11: the final channel is estimated.

2. The method for downlink channel estimation of massive MIMO system based on variational Bayesian inference as claimed in claim 1, wherein in step 1,

Φ (β) ═ XA (β) is a measurement matrix,

A(β)＝[a(θ₁+β₁),a(θ₂+β₂),...,a(θ_L+β_L)]showing the flow pattern matrix of the array,

a guide vector is represented by a guide vector,

λ represents the operating wavelength of the electromagnetic wave, d represents the spacing between adjacent antenna elements,

s is a vector of sparse representation of the L-dimensional channel over the measurement matrix phi (beta),

e is an impulse noise vector of dimension T,

w is a gaussian white noise vector with an average of 0 in dimension T and an accuracy of α.

3. The method for downlink channel estimation of massive MIMO system according to claim 1, wherein the step 2 is implemented as follows:

q(s), q (e), q (alpha), q (v), q (gamma) respectively represent approximate posterior distribution functions of s, e, alpha, v, gamma,

q(α,s,e,ν)＝q(s)q(e)q(α)q(ν)q(γ)，

μ_s＝0_L，Σ_s＝I_L，

0_Ldenotes a 0 vector, I, of dimension Lx 1_LRepresenting an identity matrix of dimension L x L,

μ_e＝0_T，Σ_e＝I_T，

Γ (· | a, b) represents a gamma distribution with a shape parameter a, a rate parameter b,

a＝b＝0.0001，

v denotes the precision vector of s,

ρ γ represents the precision vector of e,

ρ＝0.0001。

4. the method for downlink channel estimation of massive MIMO system according to claim 1, wherein the method for updating q(s) in step 4 is as follows:

wherein:

μ_s＝αΣ_sΦ^H(y-μ_e)，Σ_s＝(αΦ^HΦ+diag(ν))^-1，

(·)^Hwhich represents the transpose of the conjugate,

diag (·) denotes a diagonal operation matrix.

5. The method for downlink channel estimation of massive MIMO system according to claim 1, wherein the method for updating q (e) in step 5 is as follows:

wherein:

μ_e＝αΣ_e(y-Φμ_s)，Σ_e＝(αI_T+ρ·diag(γ))^-1。

6. the method for downlink channel estimation of massive MIMO system according to claim 1, wherein the method for updating q (α) in step 6 is as follows:

q(α)＝Γ(α|a+T,b^α),

wherein:

||·||₂represents the 2 norm of the matrix and tr (-) represents the traces of the matrix.

7. The method of claim 1, wherein the step 7 comprises updating q (v) as follows:

wherein:

[·]_i,ithe ith diagonal element of the matrix is represented.

8. The method for downlink channel estimation of massive MIMO system according to claim 1, wherein the method for updating q (γ) in step 8 is as follows:

wherein:

9. the method as claimed in claim 1, wherein the method for updating β in step 9 is as follows:

wherein:

sign () denotes a sign operation,

ζ＝[ζ(β₁),ζ(β₂),…ζ(β_L)]^T，

ζ(β_l)＝2Re(a'(θl_l+β_l)^HX^HXa(θ_l+β_l)c₁+a'(θ_l+β_l)^HX^Hc₂)，

re (-) represents the operation of the real part,

y_-l＝y-X∑_j≠lμ_ja(θ_j+β_j)，

μ_jrepresents μ_sThe jth element of (1) < x >_jlRepresentation sigma_sThe (j, l) -th element of (a),

a'(θ_l+β_l) Denotes a (theta)_l+β_l) At theta_l+β_lThe derivative of (c).

10. The method as claimed in claim 1, wherein in step 11, the estimated value of the channel is: h ═ a (β) μ_s。

Technical Field

The invention belongs to the field of wireless communication, and relates to a channel estimation method of a Multi-input Multi-output (MIMO) system, in particular to a downlink channel estimation method of a large-scale MIMO system based on Variable Bayesian Inference (VBI) in an impulse noise environment.

Background

Massive MIMO systems are receiving much attention due to their ultra-high spectral efficiency. In a massive MIMO system, a base station is configured with a large number of antennas, and the number of mobile users served by the base station is much smaller than the number of base station antennas. Compared with the existing MIMO system, the large-scale MIMO system can obviously improve the frequency spectrum efficiency, the energy efficiency and the robust performance of the system. Currently, massive MIMO technology has become one of the key technologies of 5G wireless networks.

The channel estimation is the basis of signal detection and adaptive transmission, and plays an important role in influencing the performance of large-scale MIMO wireless transmission. A major limiting factor in massive MIMO systems is the accuracy of the instantaneous Channel State Information (CSI) at the base station. In the existing method, background noise is mostly assumed as white gaussian noise when channel estimation is performed, and the background noise often generates non-gaussian noise in the practical process, so that accurate CSI acquisition becomes extremely difficult. At present, many effective methods have been proposed to solve the problem of large-scale MIMO Channel Estimation under gaussian noise environment, for example, a Channel Estimation method of a large-scale MIMO system based on off-network sparse bayes learning is proposed in documents j.dai, a.liu and v.k.n.lau, FDD Massive MIMO Channel Estimation with architecture 2D-Array Geometry, IEEE Transactions on Signal Processing, vol.66, No.10, pp.2584-2599,15May, 2018.

Disclosure of Invention

Aiming at the defects of the existing method, the invention provides a downlink channel estimation method of a large-scale MIMO system based on VBI under the impact noise environment.

The technical solution for implementing the invention comprises the following steps:

step 1: the base station adopts a uniform linear array with N antennas, the mobile users in the downlink adopt a single antenna, and the base station transmits a pilot signal matrix X within T moments, so that the signals received by the mobile users are y phi (beta) s + e + w under the condition of impact noise.

Step 2: establishing data models of q(s), q (e), q (alpha), q (nu), q (gamma) and beta and initializing parameters.

And step 3: the iteration count variable k is set to 1.

And 4, step 4: q (e), q (α), q (v), q (γ), β, update q(s).

And 5: q(s), q (α), q (v), q (γ), β, update q (e).

Step 6: q(s), q (e), q (v), q (γ), β, and q (α) are fixed and updated.

And 7: q(s), q (e), q (α), q (γ), β, update q (ν) are fixed.

And 8: q(s), q (e), q (α), q (ν), β, and q (γ) are fixed and updated.

And step 9: fixing q9s), q (e), q (alpha), q (nu), q (gamma), updating beta.

Step 10: and judging whether the iteration counting variable K reaches the upper limit K or the upper limit v converges, if not, determining that K is K +1, and returning to the step 4.

Step 11: the final channel is estimated.

The invention has the beneficial effects that:

by using the VBI method, the invention obtains a method for iteratively updating q(s), q (e), q (alpha), q (nu), q (gamma) and beta to carry out channel estimation. Compared with the prior art, the method can effectively improve the performance of channel estimation.

Drawings

FIG. 1 is a flow chart of an embodiment of the present invention.

FIG. 2 is a graph showing the normalized root mean square error (NMSE) of the channel estimated by the invention compared with the off-grid sparse Bayesian learning method when the pilot time T varies from 50 to 110 at a SNR of 10dB in 200 Monte Carlo experiments.

Detailed Description

The invention will be further explained with reference to the drawings.

As shown in fig. 1, the implementation of the present invention comprises the following steps:

(1) the base station adopts a uniform linear array with N antennas, the mobile users in the downlink adopt a single antenna, the base station sends a pilot signal matrix X within T moments, and under the condition that impulse noise exists, the signals received by the mobile users are y phi (beta) s + e + w, wherein:

Φ (β) ═ XA (β) is called the measurement matrix,

A(β)＝[a(θ₁+β₁),a(θ₂+β₂),...,a(θ_L+β_L)]showing the flow pattern matrix of the array,

a guide vector is represented by a guide vector,

λ represents the operating wavelength of the electromagnetic wave, d represents the spacing between adjacent antenna elements,

representing a uniform divisionL grid points, i.e.

Beta of (5)_iDenotes theta_iThe angular deviation of the upper part of the shaft,

s is a vector of sparse representation of the L-dimensional channel over the measurement matrix phi (beta),

e is an impulse noise vector of dimension T,

w is a gaussian white noise vector with an average of 0 in dimension T and an accuracy of α.

(2) Establishing a data model and initializing parameters:

q(α)＝Γ(α|a,b)，

simultaneously initializing each element in beta to be 0, wherein:

q(s), q (e), q (alpha), q (v), q (gamma) respectively represent approximate posterior distribution functions of s, e, alpha, v, gamma,

q(α,s,e,ν)＝q(s)q(e)q(α)q(ν)q(γ)，

complex Gaussian with mean μ and variance ΣThe distribution of the water content is carried out,

μ_s＝0_L，Σ_s＝I_L，

0_Ldenotes a 0 vector, I, of dimension Lx 1_LRepresenting an identity matrix of dimension L x L,

μ_e＝0_T，Σ_e＝I_T，

Γ (· | a, b) represents a gamma distribution with a shape parameter a, a rate parameter b,

a＝b＝0.0001，

v denotes the precision vector of s,

ρ γ represents the precision vector of e,

ρ＝0.0001。

(3) the iteration count variable k is set to 1.

(4) Fixing q (e), q (α), q (ν), q (γ), β, updating q(s):

wherein:

μ_s＝αΣ_sΦ^H(y-μ_e)，Σ_s＝(αΦ^HΦ+diag(ν))^-1，

(·)^Hwhich represents the transpose of the conjugate,

diag (·) denotes a diagonal operation matrix.

(5) Fixing q(s), q (α), q (ν), q (γ), β, updating q (e):

wherein:

μ_e＝αΣ_e(y-Φμ_s)，Σ_e＝(αI_T+ρ·diag(γ))^-1。

(6) fixing q(s), q (e), q (v), q (γ), β, update q (α):

q(α)＝Γ(α|a+T,b^α),

wherein:

||·||₂represents the 2 norm of the matrix and tr (-) represents the traces of the matrix.

(7) Fixing q(s), q (e), q (α), q (γ), β, update q (ν):

wherein:

[·]_i,ithe ith diagonal element of the matrix is represented.

(8) Fixing q(s), q (e), q (α), q (ν), β, update q (γ):

wherein:

(9) fixing q(s), q (e), q (α), q (ν), q (γ), updating β:

wherein:

sign () denotes a sign operation,

ζ＝[ζ(β₁),ζ(β₂),…ζ(β_L)]^T，

ζ(β_l)＝2Re(a'(θ_l+β_l)^HX^HXa(θ_l+β_l)c₁+a'(θ_l+β_l)^HX^Hc₂)，l＝1,2,3…..L；

re (-) represents the operation of the real part,

y_-l＝y-X∑_j≠lμ_ja(θ_j+β_j)，

μ_jrepresents μ_sThe jth element of (1) < x >_jlRepresentation sigma_sThe (j, l) -th element of (a),

a'(θ_l+β_l) Denotes a (theta)_l+β_l) At theta_l+β_lThe derivative of (c).

(10) And (4) judging whether the iteration counting variable K reaches the upper limit K of 100 or v converges (namely whether the updating result of the time is equal to the updating result of the last time), if not, judging that K is K +1, and returning to (4).

(11) Estimating the final channel: h ═ a (β) μ_s。

The effect of the present invention will be further explained with the simulation experiment.

In order to evaluate the performance of the method, it is assumed that a base station adopts a uniform linear array with N150 antennas, the operating frequency of a downlink is 2170MHz, a wireless channel is randomly generated by a 3GPP Spatial Channel Model (SCM) model, each element of a pilot signal matrix X transmitted by the base station obeys an independent Gaussian distribution with zero mean unit variance, and background noise is assumed to be a Compound Gaussian Model (CGM).

Conditions of the experiment

When the signal-to-noise ratio is 10dB and the pilot time T is changed from 50 to 110, the channel is estimated for 200 times, the grid number is 150, and the simulation result is shown in figure 2.

Analysis of experiments

As can be seen from fig. 2, the present invention can accurately estimate downlink channel information of a massive MIMO system, and its NMSE performance is significantly better than that of the existing method.

The above-listed detailed description is only a specific description of a possible embodiment of the present invention, and they are not intended to limit the scope of the present invention, and equivalent embodiments or modifications made without departing from the technical spirit of the present invention should be included in the scope of the present invention.