CN103873197A - Space correlation and clustering-combined 3D (three dimensional) MIMO (multiple input multiple output) limiting feedback spending reducing method - Google Patents

Abstract

The invention relates to the technical field of wireless communication, and discloses a space correlation and clustering-combined 3D (three dimensional) MIMO (multiple input multiple output) limiting feedback spending reducing method. In a 3D MIMO limiting feedback system, a base station end adopts an N*M uniform antenna array, a user end adopts a linear array, and based on the space correlation of a channel and the co-phase characteristic of a 2D (two dimensional) antenna vertical channel, the user end performs feedback optimal precoding matrix index (PMI) on the horizontal dimension and the vertical dimension. When each vertical dimension channel has the same horizontal phase, the horizontal dimension only feeds back one PMI, the vertical dimension is clustered and each cluster feeds back one PMI, one assistant bit is added for the PMI of the horizontal dimension and the vertical dimension during feedback, and the base station finds a corresponding vertical dimension and horizontal dimension precoding matrix to perform expansion and point multiplication on the horizontal dimension and vertical dimension precoding matrix, so the 3D precoding matrix is obtained. According to the method, the feedback spending can be effectively reduced, the complexity is low and the method is easy to realize.

Description

The 3D MIMO Limited Feedback expense reduction method that spatial coherence combines with sub-clustering

Technical field

The present invention relates to wireless communication technology field, relate to mobile communication Long Term Evolution (LTE-Advanced) technical field.

Background technology

In wireless mobile communications, need 4G communication system of future generation that higher data rate and better service quality can be provided.In order to improve power system capacity and spectrum efficiency, the companies such as Nuo Xi, A Lang, DOCOMO, high pass and Waduven have set up ARTIST4G working group, specialize in advanced wireless interfacing (the Advanced Radio Interface TechnologIes for4G SysTems of 4G system, ARTIST4G), wherein, 3D multi-antenna technology is the key technology that improves power system capacity and spectrum efficiency, has caused that the degree of depth of industry is paid attention to.3D MIMO compares with traditional 2D MIMO, and 3D MIMO has increased one dimension in vertical dimension can, for the dimension of utilizing, adopt 2D array antenna structure at base station end.Like this, effectively raise the spectrum efficiency of horizontal dimension and vertical dimension.

In 3D MIMO Limited Feedback system, horizontal dimension and the vertical dimension code book known altogether by constructing a transmitting terminal and receiving terminal, receiving terminal utilizes horizontal dimension and vertical dimension channel condition information (CSI), in pre-designed code book, select optimum pre-coding matrix according to certain Optimality Criteria, then respectively horizontal dimension and vertical dimension PMI are fed back to transmitting terminal, transmitting terminal is according to the horizontal dimension and the vertical dimension PMI that receive, carry out certain calculation process, the final 3D pre-coding matrix that forms, the Limited Feedback basic principle based on code book in Here it is 3D mimo system.

In Limited Feedback system, because uplink feedback link bandwidth is limited, and 3D MIMO in vertical dimension, increase one dimension can be for the dimension of utilizing, like this, feedback overhead will increase to some extent than 2D MIMO, therefore, become for the feedback overhead that how effectively to reduce 3D MIMO the key that can 3D MIMO technology commercialization.

Summary of the invention

In view of above-mentioned technical problem, the present invention proposes a kind of 3D MIMO Limited Feedback expense reduction method that spatial coherence combines with sub-clustering, as long as in coherence distance, vertical dimension channel is carried out to sub-clustering, like this, not only can effectively reduce feedback overhead, and performance loss is little.

In the multidiameter fading channel of space, spatial coherence is very large to the performance impact of system, and the power that affects spatial coherence mainly contains two factors: the size of antenna spacing and angle spread (AS).When angle spread is fixed, antenna spacing is larger, and correlation is less, and antenna spacing is less, and correlation is larger.Generally, available coherence distance D _cweigh the correlation between antenna, it is the value of 0.5 o'clock that coherence distance is generally got coefficient correlation, and antenna spacing is less than coherence distance, thinks relevant between antenna, and antenna spacing is greater than coherence distance, thinks uncorrelated between antenna.For this reason, can utilize spatial coherence to improve the spectrum efficiency of 3D mimo system.

A kind of 3D MIMO Limited Feedback expense reduction method that spatial coherence combines with sub-clustering, its main performing step is as follows:

Step 1: base station end adopts N*M aerial array, receiving terminal adopts even linear array, and the horizontal dimension to 3D MIMO and vertical dimension are fed back respectively.Because vertical dimension channel has identical horizontal phase, therefore, horizontal dimension channel maximizes criterion according to signal to noise ratio, and optimum pre-coding matrix when horizontal dimension is selected maximum signal to noise ratio, records corresponding optimum pre-coding matrix index PMI(horizontal dimension PMI);

Step 2: in coherence distance, vertical dimension channel is carried out to sub-clustering, each bunch according to minimum bit-error rate criterion, vertical dimension is selected an error rate optimum pre-coding matrix hour, records corresponding optimum pre-coding matrix index PMI(vertical dimension PMI);

Step 3: horizontal dimension PMI and vertical dimension PMI are increased respectively to an overhead bit, and fed back to base station end;

Step 4: base station end, according to the PMI of horizontal dimension and vertical dimension feedback, recovers the optimum pre-coding matrix of horizontal dimension and vertical dimension, and then respectively it expanded, and then the matrix after expansion is adopted to point multiplication operation, obtains 3D pre-coding matrix;

The vertical dimension of 3D MIMO and horizontal dimension all adopt independently Limited Feedback mechanism.In the time of feedback PMI, need to increase overhead bit, to distinguish vertical dimension and horizontal dimension.Base station end, according to receiving horizontal dimension and vertical dimension PMI, by decoding, is found out corresponding horizontal dimension and vertical dimension pre-coding matrix, and is expanded in codebook set.

According to formula: $w_k^v=\underset{w_s\&Element;\mathrm{\&Omega;}}{\arg \min }\frac{1}{2}\underset{i=m(k-1)+1}{\overset{\mathrm{mk}}{\mathrm{\&Sigma;}}}\frac{e^{\frac{-p}{2{\mathrm{<figure-callout\; id="0"\; label="N"\; filenames="HDA0000475442720000031.png,HDA0000475442720000032.png"\; state="\{\{state\}\}">N</figure-callout>}}_0}{\left|\right|H_i^v*w_s\left|\right|}^2}}{\sqrt{2\mathrm{\&pi;}}\left|\right|H_i^v*w_s\left|\right|}$ Select the optimum pre-coding matrix of k bunch of vertical dimension channel, wherein

represent i vertical dimension channel information, Ω={ w ₁, w ₂... w _s... w _srepresenting codebook set, p represents transmitting antenna gross power, N ₀represent noise power.Before horizontal dimension PMI, increase bit 0, fed back to base station end.M vertical dimension channel is divided into cluster, before vertical dimension PMI, increases bit 1, fed back to base station end.When the aerial array of base station end employing 8*8, receiving terminal adopts 2 antennas, and the coherence distance of base station end is 2.2 λ, 4 vertical dimension channels can be divided into one bunch.

In traditional mimo system, precoding technique is only for horizontal dimension direction, the present invention fully utilize electromagnetic wave in the horizontal direction with vertical direction on channel information.Like this, can utilize horizontal dimension and vertical dimension to gain simultaneously carry out elevator system effectiveness.In vertical dimension direction, due to antenna spacing hour, between channel response, there is stronger correlation, therefore the present invention proposes a kind of 3D MIMO Limited Feedback expense reduction method that spatial coherence combines with sub-clustering, in coherence distance, vertical dimension channel is carried out to sub-clustering, and then a PMI of each bunch of feedback, can reduce feedback overhead so effectively.

Accompanying drawing explanation

The limited feedback method system block diagram of Fig. 1 3D MIMO of the present invention based on code book;

Fig. 2 Limited Feedback expense of the present invention reduction method schematic diagram;

Base station end spaces correlation emulation under the indoor NLOS environment of Fig. 3 the present invention;

Fig. 4 number of bits of feedback comparison diagram of the present invention;

Fig. 5 simulation comparison diagram of the present invention.

Embodiment

In 3D mimo system Limited Feedback scheme, the optimum precoding index of the each self feed back of horizontal dimension and vertical dimension, base station end is according to the index of feedback, from codebook set Ω={ w ₁, w ₂... w _s... w _sin find corresponding optimum pre-coding matrix, through expansion and dot product, obtain 3D pre-coding matrix, can effectively reduce like this overhead.

Figure 1 shows that 3D MIMO that the present invention the proposes limited feedback method system block diagram based on code book.Suppose that base station end adopts N*M array antenna, wherein N represents the line number of aerial array, and M represents the columns of aerial array, and receiving terminal antenna number is N _r, the data flow of input, through QPSK modulation and layer mapping, is divided into L parallel subflow x, and for convenience of calculation, we suppose L=1.Then through precoding, L parallel subflow matched on the end array antenna of base station and sent.Output signal can be expressed as:

$y=G_{\mathrm{ZF}}^H(\sqrt{p}{\mathrm{HW}}_{3D}x+n_0)$

Wherein G _zFrepresent that ZF detects matrix, p represents transmitting antenna gross power, and H represents 3D channel matrix, W _3Drepresent corresponding 3D pre-coding matrix, x represents parallel data flow, n ₀obeying average is zero, and variance is N ₀additive white Gaussian noise.

Receiving terminal estimates to obtain channel information H to channel:

$H={\left[\begin{array}{ccc}{\mathrm{<figure-callout\; id="1"\; label="h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_{r_{}}<figure-callout\; id="1"\; label="...\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">.</figure-callout>..h_{r_{}}{{}_1,h_N,{\mathrm{<figure-callout\; id="1"\; label="v"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">v</figure-callout>}}_1}_{}& ......& h_{r_1,{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_1,v_M}<figure-callout\; id="1"\; label="...\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">.</figure-callout>..h_{r_{}}{{}_1,h_N,v_{\mathrm{<figure-callout\; id="2"\; label="M\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">M</figure-callout>}}}_{}& \\ h_{r_{}}{{}_2,{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_1,{\mathrm{<figure-callout\; id="1"\; label="v"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">v</figure-callout>}}_1}_{}<figure-callout\; id="2"\; label="...\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">.</figure-callout>..h_{r_{}}{{}_2,h_N,{\mathrm{<figure-callout\; id="1"\; label="v"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">v</figure-callout>}}_1}_{}& ......& h_{r_2,{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_1,v_M}<figure-callout\; id="2"\; label="...\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">.</figure-callout>..h_{r_{}}{{}_2,h_N,v_M}_{}\\ ......& ......& ......\\ h_{r_{N_r},{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_1,v_1}...h_{r_{N_r},h_N,v_1}& ......& h_{r_{N_r},{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_1,v_M}...h_{r_{N_r},h_N,v_M}\end{array}\right]}_{N_r*N*M}$

Further, H is decomposed, obtains j horizontal dimension channel information:

$H_j^h=[{\mathrm{<figure-callout\; id="1"\; label="h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_{r_{}}<figure-callout\; id="1"\; label="...\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">.</figure-callout>..h_{r_{}}{{}_1...r_{\mathrm{Nr}},h_N,v_J}_{}]\&Element;C^{N_r*N}(j=\mathrm{1,2}...N)$

With i vertical dimension channel information:

$H_i^v=[{\mathrm{<figure-callout\; id="1"\; label="h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_{r_{}}<figure-callout\; id="1"\; label="...\; h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">.</figure-callout>..h_{r_{}}{{}_1...r_{\mathrm{Nr}},h_i,v_M}_{}]\&Element;C^{N_r*M}(i=\mathrm{1,2}...M)$

Receiving terminal receives after signal, adopts ZF detection algorithm, and corresponding ZF detects the matrix of a linear transformation and is: $G_{\mathrm{ZF}}={\left(W_{3D}^H{H}^H{\mathrm{HW}}_{3D}\right)}^{-1}{W}_{3D}^H{H}^H$

Corresponding received signal to noise ratio can be expressed as:

$\mathrm{SNR}=\frac{{\left|G_{\mathrm{ZF}}{\mathrm{HW}}_{3D}\right|}^2}{{\left|G_{\mathrm{ZF}}\right|}^2{\mathrm{<figure-callout\; id="0"\; label="N"\; filenames="HDA0000475442720000031.png,HDA0000475442720000032.png"\; state="\{\{state\}\}">N</figure-callout>}}_0}$

In Limited Feedback system, in codebook set Ω, altogether comprise S code word { w ₁, w ₂... w _s... w _s; receiving terminal utilizes signal to noise ratio to maximize criterion according to horizontal dimension CSI and in codebook set, chooses optimum pre-coding matrix; utilize minimum bit-error rate criterion according to vertical dimension CSI; from codebook set, choose optimum pre-coding matrix for each bunch; then the PMI of horizontal dimension and vertical dimension is fed back to base station end after additional overhead bit; base station end is according to the PMI of horizontal dimension and vertical dimension feedback; recover the optimum pre-coding matrix of horizontal dimension and vertical dimension; and then respectively it is expanded; then the matrix after expansion is adopted to point multiplication operation, obtain 3D pre-coding matrix W _3D.

Be illustrated in figure 2 the Limited Feedback expense reduction method schematic diagram that the present invention proposes, its concrete steps are as follows:

(1) according to horizontal dimension CSI, at codebook set Ω={ w ₁, w ₂... w _s... w _sin search meet the maximized optimum pre-coding matrix w of signal to noise ratio ^h,

$w^h=\underset{w_s\&Element;\mathrm{\&Omega;}}{\arg \max }{\left|\right|H_j^h*w_s\left|\right|}^2$

Wherein

represent the individual horizontal dimension channel information of j (j=1,2...N).

(2) according to spatial coherence, vertical dimension channel is carried out to sub-clustering, in coherence distance, channel has strong correlation, for this reason, according to spatial coherence, m vertical dimension channel is divided into cluster, m value is larger, the expense of feedback is less, but cost is systemic loss of energy, and m value is less, systematic function is better, but cost is feedback overhead to be increased, so in sub-clustering, between to compromise.

(3) choose optimum pre-coding matrix according to minimum bit-error rate criterion to the k of vertical dimension channel bunch:

$w_k^v=\underset{w_s\&Element;\mathrm{\&Omega;}}{\arg \min }\frac{1}{2}\underset{i=m(k-1)+1}{\overset{\mathrm{mk}}{\mathrm{\&Sigma;}}}\frac{e^{\frac{-p}{2{\mathrm{<figure-callout\; id="0"\; label="N"\; filenames="HDA0000475442720000031.png,HDA0000475442720000032.png"\; state="\{\{state\}\}">N</figure-callout>}}_0}{\left|\right|H_i^v*w_s\left|\right|}^2}}{\sqrt{2\mathrm{\&pi;}}\left|\right|H_i^v*w_s\left|\right|}$

Wherein, represent i vertical dimension channel information, Ω={ w ₁, w ₂... w _s... w _srepresenting codebook set, p represents transmitting antenna gross power, N ₀represent noise power, m is vertical dimension channel number in every gang.

(4) horizontal dimension PMI and vertical dimension PMI are increased respectively to an overhead bit, and fed back to base station end, such as increase bit 0 before horizontal dimension PMI, before vertical dimension PMI, increase bit 1.

(5) base station end, according to the PMI of horizontal dimension and vertical dimension feedback, recovers the optimum pre-coding matrix of horizontal dimension and vertical dimension, and then respectively it is expanded: W _h=[w ^h; w ^h; ... w ^h] _{(N*M) * L}.Wherein, w ^hrepresent horizontal dimension pre-coding matrix.Vertical dimension pre-coding matrix according to bunch in the channel information number that comprises can be extended to following form, in vertical dimension pre-coding matrix, comprise the pre-coding matrix number with each family that in every gang, vertical dimension channel number m equates.That is:

In the time of m=2: $W_v={\left[(w_1^v,w_1^v,w_2^v,...w_{M/2}^v,w_{M/2}^v,\right]}_{(N*M)*L}^{\&prime;}$

In the time of m=4: $W_v={[w_1^v,w_1^v,w_1^v,w_1^v,...w_{M/4}^v,w_{M/4}^v,]}_{(N*M)*L}^{\&prime;}$

Wherein, W _vrepresent vertical dimension pre-coding matrix, represent the pre-coding matrix of 1st bunch of vertical dimension,

represent the pre-coding matrix of 2nd bunch of vertical dimension,

represent the pre-coding matrix of vertical dimension M/2 bunch,

represent the pre-coding matrix of vertical dimension M/4 bunch, L is the subflow number of input data.

(5) horizontal dimension after expansion and the optimum pre-coding matrix of vertical dimension are adopted to point multiplication operation, obtain 3D pre-coding matrix: W _3D=W _h.*W _v

Figure 3 shows that the present invention's spatial coherence emulation under indoor NLOS environment.For ease of analyzing the impact of antenna spacing on spatial coherence, in the present invention, angle spread is made as definite value, i.e. AS=5 °, and the antenna distance of supposing base station end and receiving terminal when emulation is all 0.5 λ, base station end adopts the aerial array of 8*8, adopts 2 reception antennas at receiving terminal.The normalization spatial coherence of base station end between different antennae is expressed as:

$\mathrm{\&rho;}(\mathrm{\&Delta;d},{\mathrm{\&Delta;d}}_n={0,\mathrm{\&Delta;d}}_u=0,\mathrm{\&tau;}=0)=E\left\{\frac{{\mathrm{<figure-callout\; id="1"\; label="h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_{r_{}}*{\mathrm{<figure-callout\; id="2"\; label="h\; r"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_{r_{}}^{{}_2,{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="HDA0000475442720000031.png,HDA0000475442720000041.png"\; state="\{\{state\}\}">h</figure-callout>}}_2,v_2}}{\sqrt{\mathrm{\&sigma;}}\sqrt{\mathrm{\&sigma;}}}\right\}$

Wherein, ρ representation space correlation, Δ d represents the vertical dimension antenna spacing of base station end, Δ d _nrepresent the horizontal dimension antenna spacing of base station end, Δ d _uthe antenna spacing that represents receiving terminal, τ represents the time delay between different antennae,

with represent respectively the channel response of two antennas of receiving terminal, σ represents the standard deviation of channel response.By figure, we can find out, the coherence distance of base station end is approximately 2.2 λ.Therefore, 4 vertical dimension channels can be divided into one bunch, optimum is got m≤4.

Figure 4 shows that number of bits of feedback comparison diagram of the present invention, wherein feeding back frame number is 100.As can be seen from Figure, in coherence distance, the feedback overhead while getting m=4 and when m=2 is all few a lot of than the feedback overhead of vertical dimension ideal feedback, meanwhile, feedback overhead when m=4 is few during again than m=2, so, in coherence distance, m value is larger, and feedback overhead is just less.

Figure 5 shows that simulation comparison diagram of the present invention.As seen from the figure, m value is less, and bit error rate performance more approaches the desirable unity feedback of vertical dimension.In the time of m=4, the slightly inferior properties of the error rate is in the time of m=2, and as can be seen from Figure 4, corresponding feedback overhead but significantly declines, although performance is lost.In practical application, because uplink feedback link bandwidth is limited, so this performance loss is worth.The present invention proposes a kind of 3D MIMO Limited Feedback expense reduction method that channel space correlation combines with sub-clustering, and the method can reduce feedback overhead effectively, and complexity is low, is easier to realize.

Claims (6)

1. the 3D MIMO Limited Feedback expense reduction method that spatial coherence combines with sub-clustering, is characterized in that, comprises step: base station end adopts N*M aerial array, and receiving terminal adopts even linear array, and the horizontal dimension to 3D MIMO and vertical dimension are fed back respectively; The optimum pre-coding matrix of a horizontal dimension while selecting signal to noise ratio maximum, records corresponding horizontal dimension PMI; In coherence distance, vertical dimension channel is carried out to sub-clustering, in each bunch, select optimum pre-coding matrix according to minimum bit-error rate criterion, record corresponding vertical dimension PMI; Horizontal dimension PMI and vertical dimension PMI are increased respectively to an overhead bit, and fed back to base station end; Base station end, according to horizontal dimension and the vertical dimension PMI of feedback, recovers the optimum pre-coding matrix of horizontal dimension and vertical dimension, after respectively it being expanded, carries out point multiplication operation, obtains 3D pre-coding matrix.

2. method according to claim 1, is characterized in that, in horizontal dimension, according to formula:

at codebook set Ω={ w ₁, w ₂... w _s... w _sin search meet the maximized optimum pre-coding matrix w of signal to noise ratio ^h, wherein, represent the individual horizontal dimension channel information of j (j=1,2...N).

3. method according to claim 1, is characterized in that, selects optimum pre-coding matrix to be specially according to minimum bit-error rate criterion: according to formula: $w_k^v=\underset{w_s\&Element;\mathrm{\&Omega;}}{\arg \min }\frac{1}{2}\underset{i=m(k-1)+1}{\overset{\mathrm{mk}}{\mathrm{\&Sigma;}}}\frac{e^{\frac{-p}{2N_0}{\left|\right|H_i^v*w_s\left|\right|}^2}}{\sqrt{2\mathrm{\&pi;}}\left|\right|H_i^v*w_s\left|\right|}$ Select the optimum pre-coding matrix of k bunch of vertical dimension channel, wherein,

represent i vertical dimension channel information, p represents transmitting antenna gross power, N ₀represent noise power, m is vertical dimension channel number in every gang, Ω={ w ₁, w ₂... w _s... w _sexpression codebook set.

4. method according to claim 1, is characterized in that, horizontal dimension PMI and vertical dimension PMI is increased respectively to an overhead bit and be specially: before horizontal dimension PMI, increase bit 0, m vertical dimension channel is divided into cluster, increase bit 1 before vertical dimension PMI.

5. method according to claim 1, is characterized in that, respectively the optimum pre-coding matrix of horizontal dimension and vertical dimension is expanded specifically and is comprised: according to formula W _h=[w ^h; w ^h; ... w ^h] _{(N*M) * L}the optimum pre-coding matrix of horizontal dimension is expanded, wherein, w ^hrepresent horizontal dimension pre-coding matrix; Vertical dimension pre-coding matrix is: in the time of m=2: $W_v={\left[(w_1^v,w_1^v,w_2^v,...w_{M/2}^v,w_{M/2}^v,\right]}_{(N*M)*L}^{\&prime;};,$ In the time of m=4: $W_v={[w_1^v,w_1^v,w_1^v,w_1^v,...w_{M/4}^v,w_{M/4}^v,]}_{(N*M)*L}^{\&prime;},$ Wherein, W _vrepresent vertical dimension pre-coding matrix,

represent the pre-coding matrix of 1st bunch of vertical dimension,

represent the pre-coding matrix of 2nd bunch of vertical dimension, represent the pre-coding matrix of vertical dimension M/2 bunch,

represent the pre-coding matrix of vertical dimension M/4 bunch, m is vertical dimension channel number in every gang.

6. method according to claim 3, is characterized in that, when the aerial array of base station end employing 8*8, receiving terminal adopts 2 antennas, and the coherence distance of base station end is 2.2 λ, 4 vertical dimension channels can be divided into one bunch.

Images