WO2012163125A1

WO2012163125A1 - Method and system for channel information feedback

Info

Publication number: WO2012163125A1
Application number: PCT/CN2012/072146
Authority: WO
Inventors: 姜静; 郁光辉; 朱常青
Original assignee: 中兴通讯股份有限公司
Priority date: 2011-06-02
Filing date: 2012-03-09
Publication date: 2012-12-06
Also published as: CN102811111A

Abstract

Disclosed are a method and system for channel information feedback capable of decomposing a pre-coding matrix to be fed back into a product of a plurality of Givens rotation matrices, each rotation matrix corresponding to a rotation angle; the corresponding rotation angles of the decomposed rotation matrices being fed back. The method and the system of the present invention are a channel compression and quantization feedback solution based on Givens transform and can use continuous Givens rotations to convert a pre-coding matrix to be fed back into a finite angular value, and a base station can use these fed back and quantized angles to re-establish a channel matrix, thereby calculating a pre-coding vector to effectively reduce feedback overheads. Regarding conventional scalar quantization, the present invention is able to effectively reduce the amount of quantization elements, thereby reducing the feedback amount; and regarding vector quantization, as it is only necessary to perform a decomposition and quantization operation, it is possible to effectively reduce system complexities while guaranteeing system performance.

Description

Channel information feedback method and system

The present invention relates to the field of communications, and in particular, to a channel information feedback method and system. Background technique

Cooperative Multipoint Transmission (CoMP) technology can effectively improve the frequency utilization of the system, and can effectively suppress inter-user interference through coordinated processing between cells. However, as the number of antennas increases and the cooperation of multiple cells, the user needs to simultaneously feed back the channel information of the local cell and its coordinated cell, and the amount of feedback is large. Therefore, the user performs quantization and compression on the estimated channel state information (CSI), and then feeds back the quantized bit information to the base station. The base station reconstructs the channel according to the quantized information obtained by the feedback, and designs a precoding matrix. This can effectively suppress inter-user interference through inter-cell cooperation, and improve the edge throughput and average throughput of the cell.

Traditional channel quantization compression schemes are mainly divided into two types: scalar quantization and vector quantization. One method in scalar quantization is to quantize each element in the matrix to a fixed number of bits; another method is to quantize the diagonal elements and upper triangular elements of the channel covariance matrix by a fixed number of bits. Scalar quantization has a lot to do with the number of elements that need to be quantized. When the number of elements to be quantized is large, a large amount of feedback is needed. Vector quantization can effectively reduce the amount of feedback. The idea is: First, select the codeword that reaches the user's maximum energy from the predetermined codebook, and then feed back the codeword index. For relative scalar quantization, vector quantization can effectively reduce the number of quantization bits and improve system performance. However, in the CoMP system, system complexity is greatly increased because of the need for joint optimization search for the codebook of each cell. Summary of the invention

In view of this, the main object of the present invention is to provide a channel information feedback method and system, Effectively reduce system complexity based on system performance.

In order to achieve the above object, the technical solution of the present invention is achieved as follows:

A channel information feedback method includes:

The precoding matrix that needs feedback is decomposed into a product of a plurality of Givens rotation matrices, each rotation matrix being only related to the rotation angle; and the corresponding rotation angle of the rotation matrix that completes the decomposition is fed back.

Value and

^ value.

Where, the value range is [ ⁰ , ² ], and the range of ^ is [ ⁰ , ^/2 ];

Differentiate with different bits.

The rotation angle of the feedback is expressed in the form of a feedback matrix, and the format of the feedback matrix is: ^{V =} [ ^V i NJ;

The channel information feedback matrix of the terminal to the first coordinated base station, v ₂ is a channel information feedback matrix of the terminal to the second coordinated base station, and V _N is a channel information feedback matrix of the terminal to the Nth cooperative base station.

Each channel information feedback matrix adopts the same feedback format as the single cell; or, based on the first feedback matrix, the backward N-1 feedback matrices are sequentially fed back in the form of a difference. A channel information feedback system, comprising a decomposition unit and a feedback unit; wherein

The decomposition unit is configured to decompose a precoding matrix that needs feedback into a product of a plurality of Givens rotation matrices, each rotation matrix being only related to a rotation angle;

The feedback unit is configured to feed back a corresponding rotation angle of the rotation matrix that is decomposed by the decomposition unit. Τ,υ, n ₇ '=1 2^ _{=; +1} 'values and i=i U values.

Where, the value range is [0, ² ], and the range of ^ is [ ⁰ , ^/2 ]; And ^ are quantized with different bits.

The rotation angle fed back by the feedback unit is expressed in the form of a feedback matrix. The format of the feedback matrix is: ^¥ =[ v ₂ ,..., V _N 】;

Each channel information feedback matrix adopts the same feedback format as the single cell; or, based on the first feedback matrix, the backward N-1 feedback matrices are sequentially fed back in the form of a difference. It can be seen that the channel information feedback technique of the present invention is a channel compression quantization feedback scheme based on Givens transform. The M-dimensional vector can be transformed into a vector with only one non-zero element by continuous Given rotation, and the pre-coding matrix that needs feedback can be converted into a finite angle value by continuous Givens rotation, and the base station can reconstruct the quantized angle by using these feedbacks. The channel matrix, and thus the precoding vector, can effectively reduce the feedback overhead. Compared with the traditional scalar quantization, the present invention can effectively reduce the number of quantized elements and thus reduce the amount of feedback; whereas with respect to vector quantization, since only the decomposition quantization operation is required, joint optimization search is not required, and thus the system performance can be guaranteed. Effectively reduce system complexity. DRAWINGS

Figure 1 shows the system model of CoMP;

2 is a schematic diagram of a channel information feedback process according to an embodiment of the present invention;

FIG. 3 is a diagram of a channel information feedback system according to an embodiment of the present invention. detailed description

In practical applications, a channel compression quantization feedback scheme based on Gwyneth Givens rotation can be proposed. The precoding matrix can be decomposed into a series of Givens rotation matrix products by correlation transformation, and each rotation matrix is only related to the rotation angle. This only needs to quantify each rotation angle Degree, the sender can use the feedback information to get the corresponding Givens rotation matrix to reconstruct the channel matrix. It can be seen that the feedback content is the rotation angle associated with the Given rotation matrix, and a finite rotation angle can be quantized with a fixed number of bits. Then, the base station reconstructs the channel using the quantized rotation angle obtained from the feedback link, and calculates a precoding matrix. Compared with the traditional scalar quantization, this scheme can effectively reduce the number of quantized elements and reduce the amount of feedback. Compared with vector quantization, because only the decomposition and quantization operations need to be performed, there is no need to jointly optimize the search, so it can guarantee the performance of the system. Effectively reduce system complexity.

In the system model of the multi-cell multi-user joint transmission scenario, it is assumed that there are two base stations (eNBs) 1 and eNB2 in the system, which jointly perform coordinated transmission for two user equipments (UE) 1 and UE 2, as shown in FIG. . The number of antennas of the base station is M, and the number of UE antennas is N.

The channel of the base station eNBj (j=l, 2) to UEi (i=l, 2) is ^' ^eU . It is assumed here that the UE can estimate the accurate channel matrix information of the two base stations to the UE through the downlink reference signal. In the present system, the linear precoding matrix of UE 1 and UE 2 can be represented by and respectively, and the received signal can be expressed by the following formula:

Ί =H _{11 1} x ₁ + H _{12 2} x ₂ + « _{1 The} ^ and ^ are the data symbols sent to UE 1 and UE 2, respectively, ^η ι and the additive Gaussian zero-mean white noise representing the receiving end (noise variance) For ² ). The client is an MRC (Maximum Ratio Combining) receiver, and the UE receiving channel ratio can be expressed as:

Where {1, , {1, ² } and. The UE rate can be expressed as:

=log ₂ (l + /j) (3)

The sum rate of the system can be obtained by the symmetry of the UE:

C = R!+R ₂ (4) It is assumed that the precoding vectors fed back by UE 1 are 1 and 2, and the precoding vectors fed back by UE 2 are 2 and the precoding vectors respectively correspond to the right singular matrix of the local track matrix and the cooperative channel matrix. In the cooperative beamforming scenario, if zero-forcing precoding is used, the precoding vector in the above equation is:

F = null (ΐ^ 2i ) x s J {ΐ^ 11 x « w/Z (ΐ^ 2i )}

Where is the null space of matrix A, which represents the feature space corresponding to the main singular value of matrix A.

Take the two-dimensional vector ⁼ [ 3⁄4f , and take the angle of x i ^ ^ tan ^ ), then there is a Given matrix:

Then, G^X is equivalent to rotating the X clockwise by an angle of ^ in the _X -y plane and then falling on the X-axis, and the modulus of the vector after rotation remains unchanged, namely:

The Givens matrix can be extended to the rank 2 correction matrix of the identity matrix:

0 0 0 0

0 cos ψ _ι 0 sin3⁄4^ . 0

0 0 I 0 0

0 -sin^ _/ . 0 cos ψΊ j 0

0 0 0 0 ”MxM ( § ) can rotate the vector ^{χ =} [ ι 3⁄4 ... of the first and third elements,

[χ] _{! = Λ} / ₊ [χ], =0, and the remaining position elements of X are unchanged. This allows one element to be erased by a Givens rotation, and multiple elements can be eliminated by multiple rotations. If you take 1, Z, then take ² , ³ , ···, ^Μ , you can transform J into: , „ Givens Rotation for -1 times . , 2 m

x , ⁰ , ⁰ ,···, ⁰ ] ( 9 )

It should be noted that the precoding matrix that needs feedback can be decomposed into the products of several Givens rotation matrices, and the channel information can be reconstructed at the receiving end by simply feeding back the corresponding rotation angle. The specific quantization compression steps are as follows:

(1) Decompose the local channel at UE i ( ^{= 1} , ² ) or do SVD (singular value)

(Under ambiguity, the subscript is omitted in the following analysis:

H = please" (10)

The feedback content is required to be ^=V(:, ^1: )), where ^ corresponds to the number of data streams of the UE.

(2) Change the last row element to a positive real number. Take the matrix:

Where = ngle([V _p ] _M .) ( = 1, 2 ) , get ^ 3⁄4^.

(3) Convert the first column element of ^ to a positive real number, and get the value of ^ where:

D, = diag (e gas e gas..., e "' ¹ , 1) ( ₁₂ )

(4) Use the Givens matrix υ ^^—^υ to erase the 2, 3, ..., and Μ elements of the first column of "„" in order to obtain:

1 0

0 *

₄ (^ _M4 )-..,G ₃₁ (^, ₁ )-G ₂₁ (^ ₂ , ₁ )-£> ₁ § =v ₂

(13) The symbol "*" in the above formula represents an arbitrary element or matrix.

The first element of the second column of the transformed ^ is 0, and the value is determined by the column orthogonality of ^. (5) Repeat steps (3) and (4) to process the other columns of ^. Among them, for the i-th column (i = 2,3,...,min{ -l,N} ), there is.

0 ··· 0 0

0 e ^] 0 0 0

O = : 0 "·. 0 0

0 e 0

0 0 0 1

(14)

The required Givens matrix is in turn ^G ' ₊ "(U ^G ' ₊₂ .'(U, (^/- . Each column requires Z values to make the column element a real number, then ∑'='·" ^Ζ A value of ^ to do Givens rotation.

(6) After the above transformation, ^ is transformed into a unit matrix of MxN: phase shift by D ₍

VP, Givens rotation by G _h ψ _ι )

(15)

The ^D i and ^Gli used in the above transformation are all 酉 matrices, so the inverse transform of the above transform can be used.

It can be seen that the UE only needs to feedback and can use these ^ and ^ reconstructions at the base station end.

Similarly, the right singular matrix of the cooperative base station to the UE channel matrix can be fed back.

In practical applications, the above range of values is [ ⁰ , ² ], and the range of ^ is [ ⁰ , ^/2 ], which can be quantized by different bits, for example:

1 k

Ψ = π\ — + - , t = 0,l,...,2 ⁶ -l In the formula, ό is the number of quantization bits of ^, and + ² is the number of quantization bits.

3.4 feedback format

When making feedback, the specific format of the feedback matrix is as follows:

ν = [1⁄4, ν ₂ ,..., ν _Ν 】 ( ₁₈ ) where 1⁄4 is the channel information feedback matrix of the terminal to the first cooperative base station, and v ₂ is the channel information feedback matrix of the terminal to the second cooperative base station, V _N is a channel information feedback matrix of the terminal to the Nth cooperative base station. Each channel information feedback matrix may adopt the same feedback format as the single cell, or may be based on the first feedback matrix, and the subsequent N-1 feedback matrices are sequentially inverted in the form of a difference.

Embodiment 1:

The specific format of the feedback matrix is as follows:

v ₂ ,..., v _N 】 _U9 ) where 1⁄4 is the channel information feedback matrix of the terminal to the first cooperative base station, v ₂ is the channel information feedback matrix of the terminal to the second cooperative base station, and v _N is the terminal to the Channel information feedback matrix of N cooperative base stations. Each channel information feedback matrix can adopt the same feedback format as the single cell, that is, the same quantization method as that of equation (17).

Embodiment 2:

The specific format of the feedback matrix is as follows:

V ^, V ₂ ,..., V _N 】 ( ₂₀ ) where 1⁄4 is the channel information feedback matrix of the terminal to the first cooperative base station, and v ₂ is the channel information feedback matrix of the terminal to the second cooperative base station, V _N It is a channel information feedback matrix from the terminal to the Nth cooperative base station. The first feedback matrix of each channel information feedback matrix is used as a reference, and the last N-1 feedback matrices are sequentially fed back in the form of differences. That is, the first matrix 1⁄4 is formed by the same quantization method as (17), where is the number of quantization bits of ^, and the number of quantization bits of + ² is; the feedback value of the second matrix is Φ2-Φ1, Φ 2- Φ 1 , the differential value Φ2-Φ 1 is fed back with b-2 bits, Feedback with b bits. The feedback value of the Nth matrix is ΦΝ-Φ Ι , Φ Ν- Φ 1 , and the differential value ΦΝ-Φ 1 is fed back with b-2 bits, and Ψ Ν- Ψ 1 is fed back with b bits.

As shown in the above description, the operation of channel information feedback in the present invention may represent the process shown in FIG. 2, and the process includes the following steps:

Step 210: Decompose the precoding matrix that needs feedback into a product of multiple Givens rotation matrices, and each rotation matrix is only related to the rotation angle.

Step 220: Feedback the corresponding rotation angle of the rotation matrix that completes the decomposition.

To ensure that the above technical description and operation ideas can be implemented smoothly, the settings shown in Figure 3 can be performed. Referring to FIG. 3, FIG. 3 is a diagram of a channel information feedback system according to an embodiment of the present invention, where the system includes a connected decomposition unit and a feedback unit.

In practical applications, the decomposition unit can decompose the precoding matrix that needs feedback into a product of multiple Givens rotation matrices, each rotation matrix is only related to the rotation angle; the feedback unit can feed back the rotation matrix that is decomposed by the decomposition unit. The corresponding rotation angle.

In summary, the channel information feedback technique of the present invention is a channel compression quantization feedback scheme based on Givens transform, whether it is a method or a system. The M-dimensional vector can be transformed into a vector with only one non-zero element by continuous Given rotation, and then the pre-coding matrix that needs feedback can be converted into a finite angle value by using successive Givens rotation, and the base station can reconstruct the channel by using these feedbackd quantization angles. The matrix, and thus the precoding vector, can effectively reduce the feedback overhead. Compared with the traditional scalar quantization, the present invention can effectively reduce the number of quantized elements and thus reduce the amount of feedback; whereas with respect to vector quantization, since only the decomposition quantization operation is required, joint optimization search is not required, and thus the system performance can be guaranteed. Effectively reduce system complexity.

The above is only the preferred embodiment of the present invention and is not intended to limit the scope of the present invention.

Claims

Claim

1. A channel information feedback method, comprising:

The precoding matrix that needs feedback is decomposed into products of a plurality of Givens rotation matrices, each of which is only related to the angle of rotation; the corresponding rotation angle of the rotation matrix that completes the decomposition is fed back.

2. The method according to claim 1, wherein the angle of rotation of the feedback is i = l l = i 0 values and ^ i = l l = i values.

3. The method according to claim 2, wherein the value ranges from [ ⁰ , ² ], and the range of ^ is [ ^Q , ^{/ 2} ];

Differentiate with different bits.

The method according to any one of claims 1 to 3, wherein the rotation angle of the feedback is expressed in the form of a feedback matrix, and the format of the feedback matrix is: ^{V =} [ , ···, ^V 】;

5. The method according to claim 4, wherein

Each channel information feedback matrix adopts the same feedback format as the single cell; or, based on the first feedback matrix, the latter N-1 feedback matrices are sequentially fed back in differential form.

A channel information feedback system, comprising: a decomposing unit and a feedback unit; wherein the decomposing unit is configured to decompose a precoding matrix that needs feedback into a product of a plurality of Givens rotation matrices, each rotation matrix only having a rotation angle Related

The feedback unit is configured to feed back a corresponding rotation angle of the rotation matrix that is decomposed by the decomposition unit.

7. The system of claim 6, wherein the rotation of the feedback unit feedback

/ _T w ^i=l U values and _!=1 U^ values.

8. The system according to claim 7, wherein the value ranges from [ ⁰ , ² ], and the range of ^ is [ ^Q , ^/2 ];

数量 Quantify with different bits.

The system according to any one of claims 6 to 8, wherein the rotation angle fed back by the feedback unit is expressed in the form of a feedback matrix, and the format of the feedback matrix is:

VV = Γ LV Vl, V ^ν 2'·"' V ^V N 1J ·

10. The system according to claim 9, wherein