Reduced-overhead channel estimation method and system thereof for massive MIMO systems

Abstract

A reduced-overhead channel estimation method and system thereof for massive MIMO systems is provided in present application. The method is applied in a base station device, and includes following steps: first, enabling the base station device to acquire a plurality of channel matrixes between the base station device and one or a plurality of external user devices, then enabling the base station device to label the positions of a non-zero coefficient and a common support coefficient in a plurality of fields of the channel matrixes, and then enabling the base station device to configure the non-zero coefficient and the common support coefficient to have the weights different from the weights of the coefficients in the other fields in the channel matrixes so as to provide estimating channel matrixes.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention generally relates to a reduced-overhead channel estimation method and the system thereof, in particular to a reduced-overhead channel estimation method and the system thereof for massive MIMO systems.

2. Description of the Prior Art

Currently, the implementation of channel feedback under most MIMO systems depends on the precoding of the codebook in order to reduce the load of the feedback. However, for the massive MIMO systems, the size of the codebook will significantly increase due to large numbers of antennas in the systems; on the other hand, the quantified channel status information must be influenced by the error of the quantification, so its precision will be low.

Besides, if user devices need to obtain the status information of downlink channels, the base station device should transmit pilot signals with long length; further, the user devices should perform highly complicated channel estimation, which will result in serious communication overhead.

Therefore, it has become an important technical issue in the technical field to provide a solution in order to solve the aforementioned problems.

SUMMARY OF THE INVENTION

Therefore, it is a primary objective of the present invention to provide a reduced-overhead channel estimation technology for massive MIMO system so as to solve the aforementioned problems.

To achieve the foregoing objective, the present invention provides a reduced-overhead channel estimation method for massive MIMO systems. The method is applied to a base station device, and includes the following steps: first, enabling the base station device to acquire a plurality of channel matrixes between the base station device and one or a plurality of external user devices. Then, enabling the base station device to label the position of a non-zero coefficient and a common support coefficient in a plurality of fields of the channel matrixes. Afterward, enabling the base station device to configure the non-zero coefficient and the common support coefficient to have the weights different from the weights of the coefficients in the other fields in the channel matrixes so as to provide estimating channel matrixes.

To achieve the foregoing objective, the present invention further provides a reduced-overhead channel estimation system for massive MIMO systems. The system includes a MIMO antenna module and a processing module. The MIMO antenna module communicates with one or a plurality of user devices. The processing module connects to the MIMO antenna module, wherein the processing module acquires a plurality of channel matrixes between the base station device and one or a plurality of external user device, labels the positions of a non-zero coefficient and a common support coefficient in a plurality of fields of the channel matrixes, and configures the non-zero coefficient and the common support coefficient to have the weights different from the weights of the coefficients in other fields in the channel matrixes so as to provide estimating channel matrixes.

To sum up, the reduced-overhead channel estimation method and the system thereof for massive MIMO systems in accordance with the present invention can configure the coefficients of different fields in the channel matrixes to have different weights in order to set the estimating channel matrixes, which can effectively reduce the channel information which needs to be returned.

BRIEF DESCRIPTION OF THE DRAWINGS

For a better understanding of the aforementioned embodiments of the invention as well as additional embodiments thereof, reference should be made to the Description of Embodiments below, in conjunction with the following drawings in which like reference numerals refer to corresponding parts throughout the figures.

FIG. 1 is the flow chart of the first embodiment of the reduced-overhead channel estimation method for massive MIMO systems in accordance with the present invention.

FIG. 2 is the block diagram of the second embodiment of the reduced-overhead channel estimation system for massive MIMO systems in accordance with the present invention.

FIG. 3 is the timing diagram of the downlink channel status information acquisition method in accordance with the present invention.

FIG. 4 and FIG. 5 are the schematic views of the channel matrixes in accordance with the present invention.

FIG. 6 is the schematic view of the matrix formula for receiving signals in accordance with the present invention.

FIG. 7 is the flow chart of one embodiment in accordance with the present invention.

FIG. 8 is the schematic view of the simulation result of one embodiment in accordance with the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The following description is about embodiments of the present invention; however it is not intended to limit the scope of the present invention.

Please refer to FIG. 1, which is the flow chart of the first embodiment of the reduced-overhead channel estimation method for massive MIMO systems in accordance with the present invention. The method is applied to a base station device, and includes the following steps:

S101: enabling the base station device to acquire a plurality of channel matrixes between the base station device and one or a plurality of external user devices.

S102: enabling the base station device to label the positions of a non-zero coefficient and a common support coefficient in a plurality of fields of the channel matrixes.

S103: enabling the base station device to configure the non-zero coefficient and the common support coefficient to have the weights different from the weights of the coefficients in the other fields in the channel matrixes so as to provide estimating channel matrixes.

In another embodiment, the position of the common support coefficient of the method is the intersection field of the channel matrixes. In another embodiment, the channel matrixes of the method are sparse matrixes. In another embodiment, the channel matrixes of the method include a training symbol matrix that the base station device transmits to the user device, and includes no error loop (NEL) channel matrixes that the user device returns to the base station device. In another embodiment, the estimating channel matrixes of the method are related to the downlink channel, wherein the channel matrixes are angular space channel matrixes.

Please refer to FIG. 2, which is the block diagram of the second embodiment of the reduced-overhead channel estimation system 1 for massive MIMO systems in accordance with the present invention. The system 1 includes a MIMO antenna module 11 and a processing module 12. The MIMO antenna module 11 communicates with one or a plurality of user devices. The processing module 12 connects to the MIMO antenna module 11, wherein the processing module 12 acquires a plurality of channel matrixes between the base station device and one or a plurality of external user device, labels the positions of a non-zero coefficient and a common support coefficient in a plurality of fields of the channel matrixes, and configures the non-zero coefficient and the common support coefficient to have the weights different from the weights of the coefficients in other fields in the channel matrixes so as to provide estimating channel matrixes.

In another embodiment, the position of the common support coefficient of the system 1 is the intersection field of the channel matrixes. In another embodiment, the channel matrixes of the system 1 are sparse matrixes. In another embodiment, the channel matrixes of the system 1 include a training symbol matrix that the base station device transmits to the user device, and includes no error loop (NEL) channel matrixes that the user device returns to the base station device. In another embodiment, the estimating channel matrixes of the system 1 are related to the downlink channel, wherein the channel matrixes are angular space channel matrixes.

The following content will describe the first embodiment of the reduced-overhead channel estimation method in accordance with the present invention. The method is mainly applied to a FDD multi-user massive MIMO system; the reduced-overhead channel estimation method is developed according to the sparsity of the channel status caused by the height difference between the base station device and the user devices.

Consider a multi-user FDD massive MIMO system, which includes a base station device with M antennas and K user devices; each user device uses N antennas to receive signals (N<<M). For the purpose of estimating the downlink channel, the base station device will transmit T training symbols from each of its antennas; under the assumption of flat fading, the signal received by the i^th user device during the training stage can be expressed by the following equation:Y_i=H_iX+N_i,1≤i≤K  (1)

In the equation, H_i∈C^N×M stands for the i^th downlink channel matrix; H∈C^M×T stands for the training symbol matrix, and the total power is Tr(X^HX)=PT, where P stands for the energy transmitted by each training during one unit time; N_i∈C^N×T is the noise matrix, where each element is the variance σ_n² and the zero-mean additive white Gaussian noise, and these elements are independent from one another.

As the base station device uses a large number of antennas, the number of the CSI coefficients is also large, which is in proportion to the size of the antenna array used by the base station device; directly estimating these coefficients will not only significantly increase the computation complexity, but also will result in a great amount of CSI feedback overhead.

Accordingly, the present invention provides a downlink channel information acquisition method, as shown in FIG. 3. In the method, the base station device broadcasts the training symbol matrix X to all user devices S201; each user device receives the matrix as shown and observed in Equation (1) S202, and then returns a no error loop feedback channel to the base station device S203. The present invention assumes that teach user device transmits the signal to the base station device via a collision-free protocol, such as time division multiple access (TDMA); therefore, the base station device can receive the signal transmitted from each of the user devices and deal with the data Y_i so as to estimate the downlink channel corresponding to each of the user devices S204.

The present invention defines g_l, D_l, θ_l and ϕ_l as the attenuation of the i^th path, the distance between the transmitter and the receiver, the emergent angle and the incident angle respectively; λ_c stands for the wavelength of the carrier eave; L_t and L_r are normalized antenna array lengths of the transmitter and the receiver respectively; d_t and d_r are the antenna spacing intervals of the transmitter and the receiver respectively. Assume the used downlink channel has L paths, and the downlink channel matrix H_i of Equation (1) can be expressed by the following equation:

$\begin{array}{cc}{H}_i=\sum _{l=1}^L{g}_l^b{a}_r\left(\cos {\varphi }_i\right)a_t^H\left(\cos {\theta }_l\right)& \left(2\right)\end{array}$

In the equation, g_l^b, a_r(cos ϕ_l) and a_t(cos θ_l) are the effective attenuation for the l^th path, the array heading vector in the direction of the incident angle ϕ_l for the receiver and the array heading vector in the direction of the emergent angle θ_l for the transmitter, which can be expressed by the following equations:

$\begin{array}{cc}{g}_l^b=g_l\sqrt{\mathrm{MN}}\exp \left(-j2\pi d_l/{\lambda }_c\right)& \left(3\right)\\ a_r\left(\cos {\varphi }_l\right)={\frac{1}{\sqrt{N}}\left[1\exp \left(-j2\pi \frac{d_r}{{\lambda }_c}\cos {\varphi }_l\right)L\exp \left(-j2\pi \left(N-1\right)\frac{d_r}{{\lambda }_c}\cos {\varphi }_l\right)\right]}^T,a_t\left(\cos {\theta }_l\right)={\frac{1}{\sqrt{M}}\left[1\exp \left(-j2\pi \frac{d_t}{{\lambda }_c}\cos {\theta }_l\right)L\exp \left(-j2\pi \left(M-1\right)\frac{d_t}{{\lambda }_c}\cos {\theta }_l\right)\right]}^T.& \left(4\right)\end{array}$

The signal subspaces of the transmitter and the receiver can be generated by the following normalized orthogonal bases respectively:S_r={a_r(0), . . . ,a_r((N−1)/L_r)},S_t={a_t(0), . . . ,a_t((N−1)/L_t)},  (5)

These bases include the array heading vectors of Equation (4); thus, both the training symbol matrix X and the received signal Y_i can be expressed by the following equations:X=A_TX^a Y_i=A_RY_i^a  (6)

In the equations, A_R and A_T stands for the normalized matrixes formed by the vectors respectively from the S_r basis and S_t basis. According to Equation (6), which can be expressed by the following equation:

$\begin{array}{cc}\begin{array}{c}{Y}_i^a=A_R^H{H}_i{A}_T{X}^a+A_R^H{N}_i\\ =H_i^a{X}^a+N_i^a\end{array}& \left(7\right)\end{array}$

In the equation, N_i^a is the expression of the noise matrix N_i of the angle space; H_i^a is the expression of the channel matrix H_i of the angle space, which can be reduced as the following equation:H_i^a=A_R^HH_iA_T.  (8)

According to Equation (2), the (n,m)th element of H_i^a can be expressed by the following equation:

$\begin{array}{cc}\begin{array}{c}{h}_{\mathrm{nm}}^a=a_r^H\left(\left(n-1\right)/L_r\right)H_i{a}_t\left(\left(m-1\right)/L_t\right)\\ =\sum _{l=1}^L{g}_l^b\left[a_r^H\left(\left(n-1\right)/L_r\right)a_r\left(\cos {\varphi }_l\right)\right]\left[a_t^H\left(\cos {\theta }_l\right)a_t\left(\left(m-1\right)/L_t\right)\right]\end{array}& \left(9\right)\end{array}$

Please note that h_nm^a is not zero, and if and only if |(n−1)/L_r−cos ϕ_l|<1/L_r and |(m−1)/L_t−cos θ_l|<1/L_t.

Since the base station device uses a lot of transmitting antennas and the scattering is limited; H_i^a is a sparse matrix. The following content lists some basic assumptions and the descriptions of their principles:

• • 1. The column vectors of H_i^a have the same non-zero position, which is marked by T_i. The reason of which is that the height of the user devices is usually low, so will result in a lot of scattering.
  • 2. All channel matrixes share the same non-zero position T_c∩_1≤i≤KT_i, which is not empty. The reason of which is that the height of the base station device is usually high, so the scattering is limited.
  • 3. Regarding the above two assumed sparsity indicators |T_c| and |T_i|, there is a statistic limit {s_c,s₁,L,s_K}, which results in |T_c|≥s_c and |T_l|≤s_i, 1≤i≤K. In addition, the limit can be estimated by some pre-measurement processes, so can be obtained by the base station device.

FIG. 4 is the schematic view of H_i^a according to the above assumptions. When the quantity of the antennas of the base station device increases, H_i^a will be sparser because being limited by the environment. In this kind of environment, the compressive sensing approach will be an important technology for the estimation of the channel matrixes.

The present invention expresses Equation (1) as:

$\begin{array}{cc}{\stackrel{\_}{Y}}_i=\stackrel{\_}{X}{\stackrel{\_}{H}}_i+{\stackrel{\_}{N}}_i,1\le i\le K,\mathrm{where},& \left(10\right)\\ {\stackrel{\_}{H}}_i={\left(H_i^a\right)}^H,{\stackrel{\_}{Y}}_i=\sqrt{\frac{M}{\mathrm{PT}}}{Y}_i^H{A}_R,\stackrel{\_}{X}=\sqrt{\frac{M}{\mathrm{PT}}}{X}^H{A}_T,{\stackrel{\_}{N}}_i=\sqrt{\frac{M}{\mathrm{PT}}}{N}_i^H{A}_R.& \left(11\right)\end{array}$

Conjugately transposing and vectoring the matrixes at two sides of the equality sign of Equation (10) can obtain the following equation:y_i=Yh_i+n_i,1≤i≤K,  (12)

In the above equation, y_i∈vec (Y_i^H) and n_i∈vec(N_i^H).h_ivec(H_i^H)∈C^MN  (13)

The above equation stands for the unknown channel vector in the angle space, and:Ψ=(X⊗I_N)∈C^TN×MN  (14)

The above equation stands for the effective training symbol matrix, where vec(.) and ⊗ stand for the vectored function and Kronecker product respectively. Here, Equation (12) already transformed the equivalent received signal into the mode which can be dealt with by the conventional compressive sensing approach; the only difference is that h_i is block sparse; in other words, its non-zero position is of block-type expression. Regarding the block sparse characteristic of Equation (12), its schematic view is as shown in FIG. 6 (this is a noise-free version). Here, Φ is set as a unit matrix, so Θ is equal to Ψ.

As the equivalent received signal of Equation (12) has the block sparse characteristic, the performance of channel estimation can be significantly increased by using the compressive sensing approach according to the aforementioned assumption conditions for the non-zero positions of the channel model. Accordingly, the present invention proposes a two-stage weight-based l₁ minimization algorithm to recover the signals of the compressive sensing approach; the algorithm can not only take advantage of the block sparse characteristic, but also can use the l₁ minimization algorithm to substitute the OMP-based algorithm in order to achieve higher precision.

The present invention defines the weight-based block noun of the vector z=[z^T(1) . . . z^T(M)]^T∈C^MN by the following equation:

$\begin{array}{cc}{z}_{1,w,B}=\sum _{j=1}^M{w}_j{z\left(j\right)}_2,& \left(15\right)\end{array}$

In the equation, z(j)∈C^N stands for the sub-vector of the j^th block, and w_j∈[0,1] is the weight corresponding to the sub-vector. According to Equation (15), the flow chart of the method proposed by the present invention is as shown in FIG. 7, and the detailed content of the method is as follows:

Input: receive the pilot symbol signal matrix Y_i, where 1≤i≤K; the pilot symbol matrix X; the statistic upper sparsity limit {s_c,s₁,L,s_K}; the weight parameter 0≤w_c≤w≤1.

Output: the estimated channel matrix Ĥ_i, where 1≤i≤K.

Stage 1 (define the common and individual non-zero positions of the channel vector)

• 1. Under 1≤i≤K, calculate the optimization question of the minimization of the following l₁ without weight; in other words, all weights are defined to be w_j=1, and its solution is expressed by ĝ_i∈C^MN.

$\left(P1\right)\underset{w_j=1,1\le j\le M}{\mathrm{minimize}}{z_i}_{1,w,B},s.t.{y_i-\Psi z_i}_2\le \varepsilon$

• 2. Record the directory of the block sub-vectors, in the vectors ĝ_i, with the top s_i largest l₂ norms, and then express the set formed by the directory as {tilde over (T)}_i; the set has the cardinality |{tilde over (T)}_i|=s_i.
• 3. Regarding all directory sets {tilde over (T)}_i, where 1≤i≤K, search the top s_c block directories most frequently selected by the directory sets {tilde over (T)}_i, and then express which as {circumflex over (T)}_c. It is the estimated value of the common non-zero position T_c.
• 4. For the directories in all directory sets {tilde over (T)}_i, respectively search the directories by excluding the common directory set {circumflex over (T)}_c, and then express the set formed by the directories as T_i. The set has the cardinality |T_i|=s_i−s_c, and has no any intersection with the common directory set {circumflex over (T)}_c.
• 5. Express {circumflex over (T)}_i as {circumflex over (T)}_i=T_i∪{circumflex over (T)}_c, which is the estimated value of the individual non-zero position T_i.

Stage 2 (estimate the channel matrixes):

• 1. Under 1≤i≤K, calculate the optimization question of the minimization of the following weight-based block l₁, and the express its solution ĥ_i∈C^MN as the estimated value of the channel vector under the angle space.

$\left(P2\right)\underset{\underset{\underset{w_j=1,j\notin {\hat{T}}_i}{w_j=w,j\in {\hat{T}}_i\backslash {\hat{T}}_c}}{w_j=w_c,j\in {\hat{T}}_c}}{\mathrm{minimize}}{z_i}_{1,w,B},s.t.{y_i-\Psi z_i}_2\le \varepsilon$

• 2. Use Equation (11) and Equation (13) to transform ĥ_i into the channel matrix estimation value Ĥ_i^a under the angle space; then, the original channel estimation value Ĥ_i can be calculated by the equation Ĥ_i=A_RĤ_i^aA_T^H.

The preferred embodiment of the method according to the present invention is as follows: consider a multi-user FDD massive MIMO system, which includes a base station device with M antennas and K user devices; each user device uses N antennas to receive signals (N<<M). For the purpose of estimating the downlink channel, the base station device will transmit T training symbols from each of its antennas; the base station device broadcasts the training symbol matrix X to all user devices S201 with T symbol durations; each user device returns the received matrix to the base station device via a no error loop feedback channel; then, the base station uses the two-stage weight-based l₁ minimization algorithm to recover the signals of the compressive sensing approach in order to estimate the downlink channel information. The detailed settings of the parameters and the weight values are as shown in Table 1 and Table 2. FIG. 8 shows the simulation result of the normalized mean square error (NMSR) under different SNRs, where the red line is the algorithm proposed by the present invention, the black line is the algorithm without weight, and the blue line is the conventional OMP-based algorithm. As shown in FIG. 8, the algorithm proposed by the present invention can achieve better performance under both high SNR and low SNR.

	TABLE 1
	Parameter
	M	N	K	T	Si	Sc
	Value	100	2	8	45	15	4

	TABLE 2
	SNR(dB)
	(wc, w)	0	5	10	15	20	25	30
	wc	1	1	0.4	0	0	0	0
	w	1	1	0.6	0	0	0	0

The above disclosure is related to the detailed technical contents and inventive features thereof. People skilled in this field may proceed with a variety of modifications and replacements based on the disclosures and suggestions of the invention as described without departing from the characteristics thereof. Nevertheless, although such modifications and replacements are not fully disclosed in the above descriptions, they have substantially been covered in the following claims as appended.

Claims

1. A reduced-overhead channel estimation method for massive multi-input multi-output (MIMO) systems, applied in a base station device, comprising: acquiring at the base station device, a plurality of channel matrixes between the base station device and one or a plurality of user devices;labelling at the base station device, positions of a non-zero coefficient and a common support coefficient in a plurality of fields of the plurality of channel matrixes; andconfiguring at the base station device, the non-zero coefficient and the common support coefficient to have weights different from weights of coefficients in other fields in the plurality of channel matrixes so as to provide estimating channel matrixes.

2. The method as claimed in claim 1, wherein the position of the common support coefficient is an intersection field of each of the plurality of channel matrixes.

3. The method as claimed in claim 1, wherein the plurality of channel matrixes are sparse matrixes.

4. The method as claimed in claim 1, wherein the plurality of channel matrixes comprise a training symbol matrix that the base station device transmits to the one or the plurality of user devices, and comprise no error loop (NEL) channel matrixes that the one or the plurality of user devices returns to the base station device.

5. The method as claimed in claim 1, wherein the estimating channel matrixes are related to a plurality of downlink channels between the base station device and the one or the plurality of user devices, wherein the plurality of channel matrixes are angular space channel matrixes.

6. A reduced-overhead channel estimation system for massive multi-input multi-output (MIMO) systems, comprising: a MIMO antenna module, communicating with one or a plurality of user devices;a processor, connecting to the MIMO antenna module, wherein the processor acquires a plurality of channel matrixes between a base station device and one or a plurality of user devices, labels positions of a non-zero coefficient and a common support coefficient in a plurality of fields of the plurality of channel matrixes, and configures the non-zero coefficient and the common support coefficient to have weights different from weights of coefficients in other fields in the plurality of channel matrixes so as to provide estimating channel matrixes.

7. The system as claimed in claim 6, wherein the position of the common support coefficient is an intersection field of each of the plurality of channel matrixes.

8. The system as claimed in claim 6, wherein the plurality of channel matrixes are sparse matrixes.

9. The system as claimed in claim 6, wherein the plurality of channel matrixes comprise a training symbol matrix that the base station device transmits to the one or the plurality of user devices, and comprise no error loop (NEL) channel matrixes that the one or the plurality of user devices returns to the base station device.

10. The system as claimed in claim 6, wherein the estimating channel matrixes is related to a plurality of downlink channels between the base station device and the one or the plurality of user devices, wherein the plurality of channel matrixes are angular space channel matrixes.