CN108259397B - Large-scale MIMO system channel estimation method based on adaptive regularization subspace tracking compressed sensing algorithm - Google Patents

Abstract

The invention discloses a channel estimation of massive MIMO system based on adaptive regularization subspace tracking compressed sensing algorithm. Aiming at the sparseness of massive MIMO channels in the time domain, a channel estimation algorithm based on compressed sensing is designed, which has the following features: Steps: N _t antennas at the base station send information, N _r antennas at the user end receive the information, and the received pilot signal is the measurement vector y; construct a perception matrix A according to the sent pilot information; Adaptive regularization subspace tracking algorithm Estimate the sparse signal h. The method adopted in the present invention is studied under the premise of channel sparsity, and is improved on the basis of the subspace tracking algorithm. When the step size is selected for the first time, the selection is made adaptively, and the regularity is added between the two selection atoms. In the process of optimization, a group of atoms with the largest energy can be selected, and a relatively accurate estimation effect can be obtained with a small number of pilots. The effect is better than the traditional channel estimation method, and it has certain practical value.

Description

Massive MIMO system based on adaptive regularized subspace tracking compressed sensing algorithm channel estimation method

technical field

The invention belongs to the field of communication signal processing, and in particular relates to a channel estimation method for a massive MIMO system based on an adaptive regularization subspace tracking compressed sensing algorithm.

Background technique

In recent years, the wireless technology of mobile communication has developed rapidly. In the current fourth-generation mobile communication, MIMO technology utilizes the characteristics of spatial multiplexing and transmission diversity, which greatly improves the rate and reliability of information transmission. In the next decade or so, the transmission rate demand of wireless communication is expected to be a thousand times that of the current system. Therefore, 4G is still difficult to meet the application requirements of future mobile communication. Many countries have already focused on the research of fifth-generation mobile communication technology. Massive MIMO technology is a new technology developed on the basis of traditional MIMO technology. Its core idea is to equip the base station with dozens or even hundreds of antennas to form an antenna array to serve multiple users at the same time to improve spectrum utilization and improve The information transmission rate has become one of the key technologies of 5G. Accurate channel state information is required for channel equalization and correlation detection for massive MIMO technology, so it is necessary to perform channel estimation for massive MIMO systems.

At present, in the field of channel estimation of massive MIMO systems, most of the research on massive MIMO channel estimation at this stage is aimed at the TDD transmission mode. The transposition is the downlink channel state information, thereby avoiding the pilot pollution problem caused by too many antennas at the base station in the massive MIMO system. However, this channel reciprocity is not real-time, and the information of the uplink channel may be outdated and inaccurate for the downlink channel. Moreover, FDD is still the mainstream of the current cell cellular system, so it is necessary to study the downlink channel estimation in the frequency division duplex FDD transmission mode. Due to the finite amount of scattering and delay spread in the signal propagation space, and the spatial correlation of the antenna at the base station, the energy of the channel is only concentrated on a few main paths, and the energy on other paths is very small and can be ignored. , so we treat the channel as sparse in the time domain. In the research of FDD downlink channel estimation in massive MIMO systems, traditional channel estimation methods, such as minimum mean square error (MMSE) algorithm and least squares (LS) method, do not take advantage of the sparse characteristics of the channel and require more pilot signals , which wastes frequency band resources and has poor anti-noise capability.

SUMMARY OF THE INVENTION

Aiming at the shortcomings of the existing channel estimation methods, the present invention proposes a channel estimation method for massive MIMO systems based on an adaptive regularization subspace tracking compressed sensing algorithm. In the actual transmission space, due to the existence of a limited number of scattering and delay expansion, and the arrangement of the antennas at the base station is relatively compact and there is spatial correlation, the present invention is carried out on the premise that the channel time domain is sparse, and the technical steps are as follows :

A channel estimation method for massive MIMO systems based on adaptive regularization subspace tracking compressed sensing algorithm, the channel is regarded as sparse in the time domain, and has the following steps:

S1. N _t antennas at the base station send information, and N _r antennas at the user end receive the information, and the received pilot signal is a measurement vector y; construct a perception matrix A according to the sent pilot information;

S2. The adaptive regularized subspace tracking algorithm estimates the sparse signal h.

The specific steps of the step S1 are as follows:

The i-th antenna at the transmitting end sends an OFDM symbol with U subcarriers, and performs IFFT transformation to realize OFDM modulation. A cyclic prefix CP is added before each output OFDM symbol to reduce the influence of channel delay spread. These processed After digital-to-analog conversion, the OFDM signal is transmitted to the antenna of each user end in the wireless channel, and the cyclic prefix CP and FFT transformation are performed at the jth receiving antenna. The pilot signal received by the receiving end is the measurement vector y _j , y = _yj ;

其中p_n为选取的N个位置的导频信息，F_L为U点离散傅里叶变换矩阵F中对应的导频所在位置的N行和信道长度的前L列，则

Randomly select N positions on U subcarriers to place pilot symbols, and construct a sensing matrix

where p _n is the pilot information of the selected N positions, and _FL is the N rows of the corresponding pilot positions in the U-point discrete Fourier transform matrix F and the first L columns of the channel length, then

It can be obtained that the channel impulse response can be solved by the compressive sensing model.

Then use the compressed sensing reconstruction algorithm to reconstruct the channel impulse response. The reconstruction algorithm here uses an improved greedy algorithm, that is, the adaptive regularization subspace tracking algorithm:

The specific steps of the step S2 are as follows:

感知矩阵支持集：

稀疏度K；S21, perform reconstruction initialization: initial residual: r ₀ =y, number of iterations: i=1, initial step size: s=1, stage step size: stage=1, column serial number index set:

Perceptual matrix support set:

sparsity K;

S22. According to the formula u={u _j |u _j =|<r _j ,A _j >|,j=1,2,...,N}, find the inner product of the residual r and each column of the perception matrix A, and select one of them The first s maximum values are recorded in the index set J _i ;

S23. Carry out the regularization energy grading process in J _i according to the formula |u _i |≤2|u _j |,i,j∈J _i , store the index value obtained after regularization into J _x , and according to J _x Update the support set A _Jx ;

的伪逆得到

S24, ask

The pseudo-inverse of

中绝对值最大的K个元素对应的索引值记入J_t，并根据索引值J_t更新支撑集

S25. According to the retrospective thinking, select

The index values corresponding to the K elements with the largest absolute value are recorded in J _t , and the support set is updated according to the index value J _t

的伪逆得到

S26, ask

The pseudo-inverse of

S27, update residual

S28, i=i+1, if i≤K, continue, otherwise stop iterating and go to step S210;

S29, compare residual values:

If ||r _new || ₂ ≥||r _n-1 || ₂ , then stage=stage+1, s=s·stage, go to step S22;

If ||r _new || ₂ <||r _n-1 || ₂ , s=s, go to step S22;

在J_t处有非零值，其值为最后一次迭代得到的

S210. Reconstructed

There is a non-zero value at J _t whose value is obtained from the last iteration

S211. Perform the same processing on the information received by other receiving antennas, and take a union to obtain the final estimated h.

The traditional channel estimation technology does not consider the channel sparse characteristics, and will use an excessive number of pilots to obtain a certain estimation effect, which greatly reduces the frequency band utilization rate and wastes resources. The method adopted in the present invention is studied under the premise of channel sparsity, and is improved on the basis of the subspace tracking algorithm. When the step size is selected for the first time, the selection is made adaptively, and the regularity is added between the two selection atoms. In the process of optimization, a group of atoms with the largest energy can be selected, and a relatively accurate estimation effect can be obtained with a small number of pilots. The effect is better than the traditional channel estimation method, and it has certain practical value.

Based on the above reasons, the present invention can be widely promoted in the fields of communication signal processing and the like.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

FIG. 1 is a flowchart of a channel estimation method for a massive MIMO system based on an adaptive regularization subspace tracking compressed sensing algorithm in a specific embodiment of the present invention.

FIG. 2 is a transmission flow of a massive MIMO system in a specific embodiment of the present invention.

FIG. 3 is a flowchart of the estimation of the adaptive regularization subspace tracking algorithm in the specific embodiment of the present invention.

Detailed ways

In order to make the purposes, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

As shown in Fig. 1-Fig. 3, a channel estimation method for massive MIMO system based on adaptive regularization subspace tracking compressed sensing algorithm is used for downlink channel of single cell FDD transmission mode. N _t antennas are configured at the base station, and there are N _r single-antenna user terminals. In this specific embodiment, N _t =16 and N _r =8 are used for specific description.

The channel is regarded as sparse in the time domain, and the estimation has the following steps:

h_i为路径增益，τ_i为路径延迟，信道长度为L，h_i中非零个数为K个，K＜＜L。我们取信道长度为256，信道中非零个数K为6，即信道稀疏度为6；S1. N _t antennas at the base station transmit information, and N _r antennas at the user end receive the information. The received pilot signal is the measurement vector y: the channel impulse response between the i-th transmitting antenna and the j-th receiving antenna is:

hi is the path gain, τ _i is the path delay, the channel length is L, the number of non- _zeros in _hi is K, and K<<L. We take the channel length as 256, and the non-zero number K in the channel as 6, that is, the channel sparsity is 6;

The i-th antenna at the transmitting end sends an OFDM symbol with 4096 sub-carriers, and performs IFFT transformation to realize OFDM modulation. A cyclic prefix CP is added before each output OFDM symbol. In the channel, it is transmitted to the antenna of each user end, and the jth receiving antenna is used to remove the cyclic prefix CP and FFT transformation, and the pilot signal received by the receiving end is the measurement vector y _j , y=y _j , considering the in-channel noise n, then the received symbol received by the jth user is

S2. Construct a perception matrix A according to the sent pilot information:

其中p_n为选取的500个位置的导频信息，F_L为U点离散傅里叶变换矩阵F中对应的导频所在位置的500行和信道长度的前256列，则

Randomly select 500 positions on 4096 subcarriers to place pilot symbols to construct a perception matrix

where p _n is the pilot information of the selected 500 positions, F _L is the 500 rows of the corresponding pilot positions in the U-point discrete Fourier transform matrix F and the first 256 columns of the channel length, then

S3. The adaptive regularized subspace tracking algorithm estimates the sparse signal h:

In a massive MIMO system, due to the limited amount of scattering and delay spread in the signal propagation space, and the spatial correlation of the antennas at the base station, the energy of the channel is concentrated only on a few main paths, and the other paths The upper energy is very small and can be ignored, so we regard the channel as sparse in the time domain, and use the compressed sensing method to estimate it according to the characteristics of channel sparsity. Compressed sensing is a new sampling theory, which is different from the traditional Nyquist sampling theorem. It states that as long as the signal is compressible or sparse in some transform domain, then a high-dimensional signal can be projected onto a low-dimensional space with an observation matrix uncorrelated with the transform basis, and then solved by solving an optimization The problem is that the original signal can be reconstructed with high probability from these few projections. Assuming that there is an N×1-dimensional signal x in the R ^N space, x can be sparsely represented by a N×N transform basis Ψ matrix in the R ^N space, that is, x=Ψh;

h is a compressible sparse signal, i.e. only K values in h are non-zero. Compressed sensing theory shows that for a sparse signal h in a certain Ψ domain, an M×N-dimensional measurement matrix Φ uncorrelated with the transform domain Ψ can be used to project h onto y, thereby obtaining a compressed sensing measurement model, Such as formula y=Φx=ΦΨh=Ah.

Among them, y is a measurement vector of N×1 dimension and 500×1, and A is a perception matrix of N×(N _t *L) dimension, that is, 500×4096. The sparse signal h can be reconstructed using the known observation vector y and perception matrix A.

In order to solve the compressive sensing reconstruction problem, the sensing matrix A needs to satisfy the Restricted Isometry Property (RIP), but it is difficult to verify whether the sensing matrix satisfies this property, usually as long as the transformation matrix Ψ is not correlated with the measurement matrix Φ can be solved.

是一个典型的可以用压缩感知方法解决的模型。所以用压缩感知重构算法来估计信道冲激响应。

is a typical model that can be solved with compressed sensing methods. So the compressive sensing reconstruction algorithm is used to estimate the channel impulse response.

S31. Input: N×(N _t *L)-dimensional perception matrix A, N×1-dimensional observation vector y, step size s=1, sparsity K;

感知矩阵支持集：

稀疏度K；Perform reconstruction initialization: initial residual: r ₀ =y, number of iterations: i=1, initial step size: s=1, stage step size: stage=1, column serial number index set:

Perceptual matrix support set:

sparsity K;

S32. According to the formula u={u _j |u _j =|<r _j ,A _j >|,j=1,2,...,N}, find the inner product of the residual r and each column of the perception matrix A, and select one of them The first s maximum values are recorded in the index set J _i ;

S33. Carry out the regularization energy classification process for J _i according to the formula |u _i |≤2|u _j |,i,j∈J _i , store the index value obtained after regularization into J _x , and according to J _x Update the support set A _Jx ;

的伪逆得到

S34, ask

The pseudo-inverse of

中绝对值最大的K个元素对应的索引值记入J_t，并根据索引值J_t更新支撑集

S35. According to the retrospective thinking, select

The index values corresponding to the K elements with the largest absolute value are recorded in J _t , and the support set is updated according to the index value J _t

的伪逆得到

S36, ask

The pseudo-inverse of

S37, update residual

S38, i=i+1, if i≤K, continue, otherwise stop iterating and go to step S310;

S39. Compare residual values:

If ||r _new || ₂ ≥||r _n-1 || ₂ , then stage=stage+1, s=s·stage, go to step S32;

If ||r _new || ₂ <||r _n-1 || ₂ , s=s, go to step S32;

在J_t处有非零值，其值为最后一次迭代得到的

S310. Reconstructed

There is a non-zero value at J _t whose value is obtained from the last iteration

S311 , perform the same processing on the information received by other receiving antennas, and take a union to obtain the final estimated h.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions described in the foregoing embodiments can still be modified, or some or all of the technical features thereof can be equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present invention. scope.

Claims (1)

1. a massive MIMO system channel estimation method based on adaptive regularization subspace tracking compressed sensing algorithm, described channel is regarded as sparse in time domain, it is characterized in that having the following steps: S1. N _t antennas at the base station send information, and N _r antennas at the user end receive the information, and the received pilot signal is a measurement vector y; construct a perception matrix A according to the sent pilot information; S2. The adaptive regularization subspace tracking algorithm estimates the sparse signal h; The specific steps of the step S1 are as follows: The i-th antenna at the transmitting end sends an OFDM symbol with U subcarriers, and performs IFFT transformation to realize OFDM modulation. A cyclic prefix CP is added before each output OFDM symbol. After digital-to-analog conversion, these processed OFDM signals are sent to the wireless In the channel, it is transmitted to the antenna of each user end, and the jth receiving antenna is removed cyclic prefix CP and FFT transformation, and the pilot signal received by the receiving end is the measurement vector y _j , y=y _j ; Randomly select N positions on U subcarriers to place pilot symbols, and construct a sensing matrix A=

where p _n is the pilot information of the selected N positions, and _FL is the N rows of the corresponding pilot positions in the U-point discrete Fourier transform matrix F and the first L columns of the channel length, then

in

is the transmitted pilot time-domain signal, H(i,j) is the channel frequency-domain response matrix, h(i,j) is the channel time-domain sparse signal, and n is the channel noise; The specific steps of the step S2 are as follows: S21, perform reconstruction initialization: initial residual: r ₀ =y, number of iterations: i=1, initial step size: s=1, stage step size: stage=1, column serial number index set:

Perceptual matrix support set:

sparsity K; S22. According to the formula u={u _j |u _j =|<r _j ,A _j >|,j=1,2,...,N}, find the inner product of the residual r and each column of the perception matrix A, and select one of them The first s maximum values are recorded in the index set J _i ; S23. Carry out the regularization energy grading process in J _i according to the formula |u _i |≤2|u _j |,i,j∈J _i , store the index value obtained after regularization into J _x , and according to J _x Update the support set A _Jx ; S24, ask

The pseudo-inverse of

S25. According to the retrospective thinking, select

The index values corresponding to the K elements with the largest absolute value are recorded in J _t , and the support set is updated according to the index value J _t

S26, ask

The pseudo-inverse of

S27, update residual

S28, i=i+1, if i≤K, continue, otherwise stop iterating and go to step S210; S29, compare residual values: If ||r _new || ₂ ≥||r _n-1 || ₂ , then stage=stage+1, s=s·stage, go to step S22; If ||r _new || ₂ <||r _n-1 || ₂ , s=s, go to step S22; S210. Reconstructed

There is a non-zero value at J _t whose value is obtained from the last iteration

S211. Perform the same processing on the information received by other receiving antennas, and take a union to obtain the final estimated h; where r _new is the residual of this iteration, r _n-1 is the residual of the previous iteration, ||r _new || ₂ is the 2-norm of the residual value of this iteration, that is, the residual energy, ||r _n-1 || ₂ is the 2-norm of the residual value of the previous iteration, that is, the residual energy.

Images