WO2017021774A2

WO2017021774A2 - Method and apparatus for hybrid beamforming

Info

Publication number: WO2017021774A2
Application number: PCT/IB2016/001163
Authority: WO
Inventors: Lu WU; Keying Wu
Original assignee: Alcatel Lucent
Priority date: 2015-08-05
Filing date: 2016-07-04
Publication date: 2017-02-09
Also published as: TW201707394A; CN106452539B; CN106452539A; WO2017021774A3

Abstract

Embodiments of the present disclosure provide a method and apparatus for hybrid beamforming in a base station as well as a method and apparatus for hybrid beamforming in a mobile terminal. The method for hybrid beamforming in a base station comprising: calculating a wideband analog beamforming matrix based on long-term-level estimation of a physical channel; quantizing the wideband analog beamforming matrix to obtain a quantized wideband analog beamforming matrix; applying the quantized wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel; calculating a sub-band digital beamforming matrix based on short-term-level estimation of the effective channel; and performing hybrid beamforming on downlink signals with the sub-band digital beamforming matrix and the quantized wideband analog beamforming matrix. The hybrid beamforming solution according to the embodiments of the present disclosure makes massive MIMO systems more practical and cost-effective to deploy.

Description

METHOD AND APPARATUS FOR HYBRID BEAMFORMING

FIELD OF THE DISCLOSURE

[0001] Embodiments of the present disclosure generally relate to wireless communication techniques, and more particularly, to a method and apparatus for hybrid beamforming.

BACKGROUND OF THE DISCLOSURE

[0002] Massive multiple-input and multiple- output (MIMO) or large scale antenna system has been well recognized as a key enabling technology for 5G communication system, since it offers huge advantages in terms of energy efficiency, spectrum efficiency, robustness and reliability The fundamental premise of massive MIMO is that the number of base station antennas is much larger than the number of single antenna terminals. Theoretically, massive MIMO system with full digital beamforming (BF) can yield the optimal performance. However, the price to pay is the rapid increase of hardware complexity and cost (number of RF chains) as well as the complexity and energy consumption of signal processing. Consequently, when the antenna number is very large, implementing the same number of RF chains may not be feasible. Thus, how to enable massive MIMO with a limited number of RF chains is a key technical problem.

[0003] To reduce the required number of RF chains, dynamic antenna switching techniques may be a choice, but those solutions provide limited array gain and perform poorly in correlated channels since the benefits of the extra antennas are not fully utilized. To resolve this issue, analog BF with active antennas can be considered, which generally controls the signal phase on each antenna via a network of analog phase shifters. The performance of analog BF is in general sub-optimal because of hardware limitations in amplitude and phase control. In order to achieve multiple data stream precoding along with larger BF gains, hybrid analog and digital BF (hereinafter referred to as hybrid BF for short) strategies were investigated recently. Nevertheless, hybrid BF algorithms are still far less mature in many aspects such as beamforming design considering flexible user scheduling, training signal design and channel information acquisition for FDD (Frequency Division Duplex) or TDD (Time Division Duplex) systems with or without precise antenna calibration.

SUMMARY OF THE DISCLOSURE

[0004] The embodiments of the present disclosure provide a method and apparatus for hybrid beamforming in a base station as well as a method and apparatus for hybrid beamforming in a mobile terminal, so as to solve or at least partially alleviate the above-discussed problems in the prior art.

[0005] In a first aspect, the embodiments of the present disclosure provide a method for hybrid beamforming in a base station. The method comprises: calculating a wideband analog beamforming matrix based on long-term-level estimation of a physical channel; quantizing the wideband analog beamforming matrix to obtain a quantized wideband analog beamforming matrix; applying the quantized wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel; calculating a sub-band digital beamforming matrix based on short-term-level estimation of the effective channel; and performing hybrid beamforming on downlink signals with the sub-band digital beamforming matrix and the quantized wideband analog beamforming matrix.

[0006] In one embodiment, quantizing the wideband analog beamforming matrix comprises: normalizing an amplitude of each of non-zero elements in the wideband analog beamforming matrix; and with respect to the each of the non-zero elements, performing phase search one by one element in a predetermined phase set to select a phase that maximizes capacity of the effective channel.

[0007] In one embodiment, the method further comprises performing sub-band user scheduling based on the short-term-level estimation of the effective channel.

[0008] In one embodiment, the method further comprises: sending to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively; receiving from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and constructing, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

[0009] In a second aspect, the embodiments of the present disclosure further propose a method for hybrid beamforming in a base station. The method comprises: calculating a wideband analog beamforming matrix based on long-term-level estimation of a physical channel; applying the wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel; calculating a sub-band digital beamforming matrix and performing sub-band user scheduling, based on short-term-level estimation of the effective channel; and performing hybrid beamforming on downlink signals for a scheduled user with the wideband analog beamforming matrix and the sub-band digital beamforming matrix.

[0010] In one embodiment, the method further comprises: sending to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively; receiving from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and constructing, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

[0011] In a third aspect, the embodiments of the present disclosure provide a method for hybrid beamforming in a mobile terminal. The method comprises: estimating a horizontal sub-channel covariance matrix associated with a horizontal linear array of antennas of a base station based on a first training signal received from the horizontal linear array; estimating a vertical sub-channel covariance matrix associated with a vertical linear array of antennas of the base station based on a second training signal received from the vertical linear array; and providing the base station with feedbacks of estimation of the horizontal sub-channel covariance matrix and estimation of the vertical sub-channel covariance matrix. [0012] In one embodiment, the method further comprises: associating the horizontal sub-channel covariance matrix with a first transmission correlation coefficient and associating the vertical sub-channel covariance matrix with a second transmission correlation coefficient; and wherein providing the base station with feedbacks of estimation of the horizontal sub-channel covariance matrix and estimation of the vertical sub-channel covariance matrix comprises: providing the base station with feedbacks of an amplitude and a phase of the first transmission correlation coefficient as well as an amplitude and a phase of the second transmission correlation coefficient.

[0013] In one embodiment, the method further comprises: estimating an effective channel based on a third training signal received from the base station; and providing the base station with feedback of estimation of the effective channel.

[0014] In a fourth aspect, the embodiments of the present disclosure provide an apparatus for hybrid beamforming in a base station. The apparatus comprises: a long-term-level estimating unit configured to calculate a wideband analog beamforming matrix based on long-term-level estimation of a physical channel; a quantizing unit configured to quantize the wideband analog beamforming matrix to obtain a quantized wideband analog beamforming matrix; an effective channel obtaining unit configured to apply the quantized wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel; a short-term-level estimating unit configured to calculate a sub-band digital beamforming matrix based on short-term-level estimation of the effective channel; and a hybrid beamforming unit configured to perform hybrid beamforming on downlink signals with the sub-band digital beamforming matrix and the quantized wideband analog beamforming matrix.

[0015] In a fifth aspect, the embodiments of the present disclosure provide an apparatus for hybrid beamforming in a base station. The apparatus comprises: a long-term-level estimating unit configured to calculate a wideband analog beamforming matrix based on long-term-level estimation of a physical channel; an effective channel obtaining unit configured to apply the wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel; a short-term-level estimating unit configured to calculate a sub-band digital beamforming matrix and perform sub-band user scheduling, based on short-term-level estimation of the effective channel; and a hybrid beamforming unit configured to perform hybrid beamforming on downlink signals for a scheduled user with the wideband analog beamforming matrix and the sub-band digital beamforming matrix.

[0016] In a sixth aspect, the embodiments of the present disclosure provide an apparatus for hybrid beamforming in a mobile terminal. The apparatus comprises: a first estimating unit configured to estimate a horizontal sub-channel covariance matrix associated with a horizontal linear array of antennas of a base station based on a first training signal received from the horizontal linear array; a second estimating unit configured to estimate a vertical sub-channel covariance matrix associated with a vertical linear array of antennas of the base station based on a second training signal received from the vertical linear array; and a feedback unit configured to provide the base station with feedbacks of estimation of the horizontal sub-channel covariance matrix and estimation of the vertical sub-channel covariance matrix.

[0017] In one embodiment, the apparatus further comprises: an associating unit configured to associate the horizontal sub-channel covariance matrix with a first transmission correlation coefficient and associate the vertical sub-channel covariance matrix with a second transmission correlation coefficient; and the feedback unit is further configured to provide the base station with feedbacks of an amplitude and a phase of the first transmission correlation coefficient as well as an amplitude and a phase of the second transmission correlation coefficient.

[0018] In one embodiment, the apparatus further comprises: an effective channel estimating unit configured to estimate an effective channel based on a third training signal received from the base station; and the feedback unit is further configured to provide the base station with feedback of estimation of the effective channel.

[0019] The hybrid beamforming solution according to the embodiments of the present disclosure makes massive MEVIO systems more practical and cost-effective to deploy. For example, the hybrid beamforming solution according to the embodiments of the present disclosure may achieve at least one of advantageous effects as below: by quantizing a wideband analog beamforming matrix, existing hardware requirements can be better satisfied, reducing the complexity and cost of hardware implementation; by performing sub-band user scheduling based on short-term-level estimation of an effective channel, frequency selective gain can be achieved; by replacing feedback of a wideband channel covariance matrix with feedback of two lower-dimensional covariance matrices, the feedback overhead of covariance matrix as well as the overhead of training signals for channel estimation can be significantly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Through the detailed description of some embodiments of the present disclosure in the accompanying drawings, the features, advantages and other aspects of the present disclosure will become more apparent, wherein several embodiments of the present disclosure are shown for the illustration purpose only, rather than for limiting. In the accompanying drawings:

[0021] Fig. 1 shows a block diagram of a hybrid analog and digital beamforming architecture in which the embodiments of the present disclosure may be implemented;

[0022] Fig. 2 shows a block diagram of another hybrid analog and digital beamforming architecture in which the embodiments of the present disclosure may be implemented;

[0023] Fig. 3 shows a flowchart of a method for hybrid beamforming in a base station according to a first aspect of the embodiments of the present disclosure;

[0024] Fig. 4 shows a schematic view of a uniform planar array of antennas in which the embodiments of the present disclosure may be implemented;

[0025] Fig. 5 shows a flowchart of a method for hybrid beamforming in a base station according to a second aspect of the embodiments of the present disclosure;

[0026] Fig. 6 shows a flowchart of a method for hybrid beamforming in a mobile terminal according to a third aspect of the embodiments of the present disclosure;

[0027] Fig. 7 shows a block diagram of an apparatus for hybrid beamforming in a base station according to a fourth aspect of the embodiments of the present disclosure;

[0028] Fig. 8 shows a block diagram of an apparatus for hybrid beamforming in a base station according to a fifth aspect of the embodiments of the present disclosure; and

[0029] Fig. 9 shows a block diagram of an apparatus for hybrid beamforming in a mobile terminal according to a sixth aspect of the embodiments of the present disclosure.

DETAILED DESCRIPTION OF EMBODIMENTS

[0030] Preferred embodiments of the present disclosure are described below in more detail with reference to the accompanying drawings. It should be understood although the preferred embodiments are depicted in the accompanying drawings, the present disclosure may be implemented in various manners and should not be limited by the embodiments described here. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete and enable the scope of the present disclosure to be completely conveyed to those skilled in the art.

[0031] Fig. 1 shows a block diagram of hybrid analog and digital BF architecture 100 in which the embodiments of the present disclosure may be implemented. As shown in Fig. 1, a base station is equipped with N_t antennas and serves S single antenna users. Each of N(N<N_t) RF chains is connected to all N_t antennas. First of all, S data streams So(t)...Ss-i(t) are beamformed in the digital domain to generate N digitally beamformed data streams. Then, the N data streams are transformed to the time domain from the frequency domain by IDFT (Inverse Digital Fourier Transform) and inputted to the N RF chains so as to be transformed to the analog domain from the digital domain, thereby generating N analog data stream. Next, the N analog data streams are beamformed in the analog domain and N_T streams are generated. Afterwards, each of the Nj streams is mapped to a transmission antenna of the base station and further sent to UE (User Equipment) 1. . .UE S.

[0032] Fig. 2 shows a block diagram of another hybrid analog and digital BF architecture 200 in which the embodiments of the present disclosure may be implemented. The hybrid BF procedure in Fig. 2 is similar to that in Fig. 1. Nevertheless, in architecture 200 of Fig. 2, each of N(N<N_t) RF chains is connected to only part of antennas, i.e. N N antennas. Therefore, architecture 200 has a lower complexity than the architecture 100 in Fig. 1.

[0033] With reference to Fig. 3 to Fig. 8 below, detailed description is presented to a method and apparatus for hybrid beamforming according to the embodiments of the present disclosure.

[0034] In a first aspect, the embodiments of the present disclosure propose a method for hybrid beamforming in a base station. Fig. 3 shows a flowchart of a method 300 for hybrid beamforming in a base station according to the first aspect of the embodiments of the present disclosure.

[0035] Method 300 starts with step S310, in which a wideband analog BF matrix T is calculated based on long-term-level estimation of a physical channel. Next in step S320, quantization is performed on the wideband analog BF matrix T to obtain a quantized wideband analog BF matrix ΐ . In step S330, the quantized wideband analog BF matrix ΐ is applied to the physical channel to obtain an effective channel of the physical channel. In step S340, a subband digital BF matrix W(b) is calculated based on short-term-level estimation of the effective channel. Finally, hybrid BF is performed on downlink signals using the subband digital BF matrix W(b) and the quantized wideband analog BF matrix ΐ .

[0036] Thus, with the hybrid BF solution according to the embodiments of the present disclosure, the downlink BF matrix F(b) on the subband b is the multiplication of the quantized wideband analog BF matrix ΐ and the subband digital BF matrix W(b), i.e.,

F (b) = ¾ (b ) (1) where b =1,2,... ,B, B is the number of subbands. As seen from the equation (1), to obtain the downlink BF matrix F(b) on the subband b, it is necessary to calculate the quantized wideband analog BF matrix i and the subband digital BF matrix W(b). Therefore, detailed discussion is presented below to how to calculate the quantized wideband analog BF matrix ΐ and the subband digital BF matrix W(b).

[0037] As described above, the quantized wideband analog BF matrix ΐ results from quantizing the wideband analog BF matrix T. Therefore, description is presented first to how to obtain the wideband analog BF matrix T.

[0038] The wideband analog BF matrix T is in wideband level and designed based on channel statistic information only such as wideband channel covariance matrix. Suppose K≤Ξ S users are scheduled simultaneously. H(b) denotes the K X N_t transmission channel matrix on the b* subband: H(b) = [h, (b) (b) - b)]^H , l≤k_t≤S where _k (b) is the N_t X 1 downlink channel of the k^th user on the subband b.

[0039] With respect to hybrid BF architecture 100 shown in Fig. 1, considering maximizing the reduced-dimensional effective channel capacity with analog BF, i.e.,

T = arg max C

A

= arg max— det I + ( H(b) A† H(b) A (3) ~ arg max det (i + A" SA) where & is the corresponding N_t X N_t transmit channel covariance matrix obtained by the sum of wideband channel covariance matrices of the K scheduled users, i.e.,

& =— V H^fl (b)H(b)

s r (4)

= R, + R, + - + R_t

with the wideband channel covariance matrix of the user being

^R _t = ^∑ (b)h (b), k = l, 2, ..., S (5)

D b=l

Hence, the analog BF matrix T is proposed by

T = [_Ul u₂ · · · u_N ] ₍₆₎ where a dimension of T is NtxN, U_j u₂ · · · are the eigenvectors of

¾ corresponding to N largest eigenvalues. As seen from equation (6), analog BF should include the beams pointing to all k scheduled users.

[0040] Next, the N X K digital BF matrix W(b) on subband b can be obtained based on effective channel fi(b) = H(b)T using a conventional precoding algorithm such as zero-forcing (ZF) algorithm.

[0041] In addition, with respect to hybrid BF architecture 200 shown in Fig. 2, since each RF chain is only connected to part of the antennas, i.e. N_t/N antennas, T is a block diagonal matrix in the form of

where g_k is a Nt/N x 1 vector. Thus, the architecture 200 can be viewed as a special case of architecture 100 by setting the non-block diagonal elements to zero. In architecture 200, the key is how to design g_k in equation (7). To this end, the present disclosure proposes to obtain these vectors from the corresponding elements of the analog BF matrix in equation (6) for architecture 100. Consequently, equation (8) is as below:

g_* =V||t_t||,

(8)

t_k = u_k ({k - l) N_t / N + l : kN_t / N)

[0042] Further, the digital BF matrix W(b) on subband b may be generated in the same manner as in hybrid architecture 100, i.e., using a conventional precoding algorithm such as zero-forcing (ZF) algorithm based on the effective channel fi(b) = H(b)T .

[0043] As seen from the foregoing analysis, to obtain the analog BF matrix T and the digital BF matrix W(b), the base station requires the knowledge of wideband channel covariance matrices R_t and the reduced-dimensional effective channel vectors k (b) = T^H _k (b) of all users. Therefore, how to estimate them in hybrid architecture is the key issue. Hereinafter, discussion is presented to TDD system with calibrated antennas, FDD system and TDD system without calibrated antennas, respectively.

[0044] In the TDD system, if antennas are calibrated precisely, channel reciprocity can be utilized to estimate the downlink channels from uplink training signals.

[0045] In this case, the channel state information estimation includes two parts. One part includes the estimation of channel vector _k (b) and wideband channel covariance matrix, which can be done in long-term level considering the slow-varying property of wideband channel covariance matrix and reduction of training overhead. Analog BF matrix T is derived based on wideband channel covariance matrix. The other part includes the estimation of effective channel vector k (b) = T _k (b) with the derived analog BF matrix T. The estimation of hk (b) is in short-term and sub-band level to achieve frequency selective gain.

[0046] Following procedures are proposed by the embodiments of the present disclosure to obtain wideband channel covariance matrix and the reduced-dimensional effective channel of all users.

Long-term-level estimation of channel vector and wideband channel covariance matrix

[0047] Step 1 : estimate channel vector _k (b) on all sub-bands from all users based on uplink training signals.

[0048] The embodiments of the present disclosure propose a method of multiple analog BF reception of orthogonal uplink training signals to estimate channel vectors. In detail, training signals of different users are sent on different subcarriers. Suppose the uplink training signal s_k of user k for the b-th subband channel estimation is transmitted on the b(k)-t subcarrier. N_t I N different analog BF matrices are used respectively in N_t / N OFDM symbols for reception. Denote y as the Nxl received signal vector on the k-t subcarrier in the i-th OFDM symbol. The received signals on the b(k)-t subcarrier in the N_t l N OFDM symbols are

= T?h_k (b)s_k

(9)

J N N ^lN, IN^llk \^u^k

The above equations can be further written by

The selection of analog BF matrices _N needs to satisfy that the rank of the combined matrix T T and channel vector h_¾ (b) can be estimated from equation (10).

[0049] Step 2: estimate wideband channel covariance matrix R_¾ based on channel vector h_¾ (b) according to equation (5).

[0050] Step 3: derive analog BF matrix T according to equation (6) or (7) based on R_¾ . Quantization on analog BF matrix T is also required, which will be described below in detail.

Short-term-level estimation of the effective channel vector with the derived analog BF

[0051] Step 4: set the analog BF matrix as that derived in step 3, and send an uplink training signal to estimate the effective channel vector hk (b) = T^Hh_k (b) from all users.

[0052] Step 5: perform digital BF and user scheduling based on effective channel _k (b) .

FDD system or TDD system without precisely calibrated antennas

[0053] For FDD system or TDD system without precisely calibrated antennas, channel reciprocity can't be held well. Therefore, estimation and feedback of downlink channel covariance matrix and effective channel vector are required.

[0054] In this case, the channel state information estimation and feedback includes two parts. One part includes long-term- level estimation and feedback of wideband channel covariance matrix. The analog BF matrix is derived based on the feedback of wideband channel covariance matrix. The other part includes short-term-level estimation and feedback of the effective channel vector with the derived analog BF.

[0055] Following procedures are proposed by the embodiments of the present disclosure to obtain wideband channel covariance matrix and reduced-dimensional effective channel of all users.

Long-term-level channel estimation and feedback of wideband channel covariance matrix [0056] As shown in Fig. 4, a uniform planar array (UPA) of antennas is widely used for massive MIMO system due to practical limitation of geometry size. The UPA of N_t transmission antennas in Fig. 4 is composed of M_c row by M_r column antennas, i.e. N_t = M_rM_c . It is well known that the N, xl channel vector can be approximated by the kronecker product of the M_r xl horizontal sub-channel vector and the M _c xl vertical sub-channel vector, i.e.

h = h ®h _ΐ

(11)

The corresponding channel covariance matrix can be derived by

R = E(hh^H )

~ E h_r ®h )(h ®h )^B )

= R r ® R c

[0057] Apparently, the channel covariance matrix R of UPA can be approximated by the kronecker product of the sub-channel covariance matrix R_r associated with the horizontal linear array and the sub-channel covariance matrix R_c associated with the vertical linear array.

[0058] Hence, the feedback of N_t x N_t total covariance matrix R can be replaced by the feedback of two lower-dimensional covariance matrices. One is the M_r xM_r covariance matrix R_r of horizontal linear array, and the other one is the M_c xM_c covariance matrix R_c of vertical linear array. In this way, the feedback overhead of covariance matrix as well as the overhead of training signals for channel estimation can be significantly reduced.

[0059] Step 1 : each user estimates horizontal channel vectors on all sub-bands based on a first downlink training signal with specially-designed analog BF. Then, each user quantizes and provides feedback of horizontal wideband covariance matrix.

[0060] To estimate horizontal channel vectors, the base station sets a first initial analog BF matrix as below:

where e_; is the elementary vector with all zero elements except the i-th element being 1.

[0061] Then, the base station selects the i-th horizontal antenna array (i.e., the i-th row of antennas in Fig. 4) and M_r < N RF chains. With the first initial analog BF matrix in equation (13), the base station sends a downlink training signal (also referred to as a first downlink training signal) for horizontal channel vector estimation on the i-th horizontal antenna array. With the first initial analog BF matrix in equation (13), traditional training signal design and channel estimation for a full digital BF method can be reused.

[0062] As one specific example, suppose the antenna array in Fig. 4 is composed of 4 rows x 8 columns of antennas. First, the base station selects 8 antennas in the first row and 8 RF chains. Then, with the first initial analog BF matrix in equation (13), the base station sends the first downlink training signal on the 8 antennas in the first row and makes antennas in other rows silent. Next, a user estimates a horizontal channel vector corresponding to the 8 antennas in the first row based on the first downlink training signal, thereby obtaining an 8 x 1 horizontal channel vector. Then, the user calculates a covariance matrix corresponding to the antennas in the first row based on the 8 x 1 horizontal channel vector, thereby obtaining an 8 x 8 covariance matrix. Similarly, the base station and the user performs the above procedure with respect to 8 antennas in the 2^nd to 4^th rows respectively, finally obtaining four 8 x 8 covariance matrices. Later, by averaging the four 8 x 8 covariance matrices, a covariance matrix R_rwith respect to the four horizontal antenna arrays can be obtained.

[0063] Based on the estimated horizontal channel vectors on all sub-bands, a horizontal sub-channel covariance matrix can be estimated by each user according to equation (5). The quantization and feedback of horizontal sub-channel covariance matrix can reuse existing methods for linear arrays.

[0064] In one embodiment, the horizontal sub-channel covariance matrix R_r may be associated with a first transmission correlation coefficient. Taking a single-polarized linear array for example, the horizontal sub-channel covariance matrix R_ris approximated by

(14)

1 P

R P ,M -2

1

where p = ae^J >Θ is the transmission correlation coefficient. Therefore, the feedback of the horizontal sub-channel covariance matrix R_r may be simplified by the feedback of amplitude a and phase Θ of P .

[0065] Step 2: each user estimates vertical channel vectors on all sub-bands based on a second downlink training signal with specially designed analog BF. Afterwards, each user quantizes and provides feedback of a vertical wideband covariance matrix.

[0066] To estimate vertical channel vectors, the base station sets a second initial analog BF matrix as below:

O (15)

(i-l^')M _Cx(i-1)M _c

T = ^lM_rxM_r i = l, 2,- , _

o

[0067] As seen from equation (15), the second initial analog BF matrix consists of a M_C X M_C unit matrix, and the other elements are 0.

[0068] The vertical channel vector estimation and feedback of the vertical sub-channel covariance matrix R_c are similar to that in step 1, except that the base station selects the i-th vertical antenna array (i.e., the i-th column of antennas in Fig. 4) and M_C<N RF chains and uses the second initial analog BF matrix in equation (15) to send a downlink training signal (also referred to as a second downlink training signal) for vertical channel vector estimation.

[0069] In addition, in one embodiment, the vertical sub-channel covariance matrix R_c may be associated with a second transmission correlation coefficient. Therefore, the feedback of the vertical sub-channel covariance matrix R_c may be simplified by the feedback of amplitude and phase of the second transmission correlation coefficient. [0070] Step 3: the base station retrieves wideband channel covariance matrices of all users based on feedback of horizontal and vertical wideband covariance matrices in steps 1 and 2 according to equation (12). Subsequently, the base station derives an analog BF matrix T according to equation (6) or (7).

Short-term-level estimation and feedback of the effective channel vector with the derived analog BF

[0071] Step 4: the base station sets an analog BF matrix as derived in step 3, and sends a downlink training signal (also referred to as a third downlink training signal) for effective channel estimation.

[0072] Step 5: each user estimates, quantizes and feeds back the effective channel vector hk (b) = T^Hh_k (b) . At this point, traditional channel state information feedback methods can be used, such as scalar quantization, adaptive codebook, etc.

[0073] Step 6: the base station performs digital BF and user scheduling based on the feedback of effective channel h_k (b) from all users in step 5.

[0074] For steps 5 and 6, taking adaptive codebook for example, the final codeword is the multiplication of wideband effective channel covariance matrix

R_e and a predefined codeword W (e.g. DFT vector), i.e., = R_e W ^{( 16)}

[0075] The user can estimate its wideband effective channel covariance matrix based on the estimated effective channel k (b) , i.e.,

Since

= E (_T% (T% f )

= T^flR_tT

the base station can also derive the wideband effective channel covariance matrix by itself based on the knowledge of analog BF, and the derived wideband channel covariance matrix in step 3.

[0076] Consequently, there is no need to feed back the covariance matrix in adaptive codebook form, since it can be obtained by the user and the base station respectively. The user only needs to select and provide feedback of a best codeword W in equation (16).

[0077] With the above specific embodiments, detailed description has been presented to how to calculate the wideband analog BF matrix T and the sub-band digital BF matrix W(b). Hereinafter, detailed description is presented to how to obtain the quantized wideband BF matrix (J by quantizing the wideband analog BF matrix T.

[0078] As seen from above-described equations (6) and (8), each element in the analog BF matrix T as obtained based on channel estimation is an arbitrary complex number, whose phase and amplitude are random. However, analog BF is implemented using phase shifters which require each element in the analog BF matrix T to be a constant modulus and require the phase to be selected from a predetermined phase set. Thus, considering the complexity and cost of hardware implementation, the analog BF matrix T needs to be transformed to a matrix in compliance with existing hardware requirements, with the loss being minimized as much as possible. In the present disclosure, such transformation is also referred to as quantization of the analog BF matrix T for hardware impairment. An intuitive method is to independently quantize each of non-zero elements in the analog BF matrix T. However, this method may be not optimal considering channel capacity maximization. To this end, the embodiments of the present disclosure propose a phase search method according to maximum channel capacity criteria for the analog BF matrix T in equations (6) and (8).

[0079] As indicated in equation (3), the optimal analog BF matrix (i.e., a quantized analog BF matrix) with limited-phase resolution can be formulated as follows according to maximum effective channel capacity criteria:

(19) max det I + A &A

ft \ where denotes an determinant to calculate the matrix within

parentheses, and arg n¾x det 1 1 + A RA j denotes taking a matrix ^ that maximizes detjl + iA S.:A.j . The matrix ¾_. is referred to as a candidate matrix with amplitude of the non-zero elements being 1 and phase being selected from a predetermined phase set. To find the optimal analog BF tj¹ , full search on all candidate matrices is required. However, since the number of analog BF candidate matrices is d^NN' for architecture 100 and d^N' for architecture 200, the computational complexity to fulfill such a full search is too high for practical usage in large antenna systems. Consequently, it is important and necessary to have a sub-optimal method with significantly reduced computational complexity to search for analog BF matrices. To this end, the embodiments of the present disclosure propose a method of phase search according to maximum channel capacity criteria.

[0080] First of all, an amplitude of each element in the analog BF matrix T in equation (6) or (7) is normalized, and its phase is compared with phases in the predetermined phase set to select a nearest phase value. Consequently, an initialized analog BF matrix is formed and used as a seed for optimization. As one specific example, the predetermined phase set consists of 16 phase values obtained by equally dividing 360 degrees into 16 parts. It should be understood the predetermined phase set is selected depending on hardware requirements for analog BF, that is, any appropriate predetermined phase set may be selected according to specific hardware.

[0081] Then, several iterations are needed. In each iteration, phase search is performed one by one element in the above initialized analog BF matrix according to the maximum channel capacity criteria, so that one element in ¾_. is determined by each iteration. Specifically, in each iteration, only phase of one element in the above initialized analog BF matrix is changed while the other elements are kept unchanged.

Finally, with respect to an element, a phase value that maximizes det l + l & jin equation (19) can be selected from the predetermined phase set. Thereby, the number of analog BF candidate matrices can be significantly reduced to qdNN_t for architecture 100 and qdN_t for architecture 200, where q denotes the iteration number. Usually, a small value of q is enough for the convergence of search, such as 4 in following simulations.

[0082] It should be understood the above quantization procedure for the matrix T is only presented for the illustration purpose. Different quantization procedures may be employed for different hardware, and the scope of the present disclosure is not limited in this regard.

[0083] As may be understood from the foregoing description, the proposed analog BF indicates beams for all candidate users served within each cell, based on wideband channel covariance matrices of them. Then, the digital BF is performed based on the estimation of the reduced-dimensional effective channel after analog BF. Conventional precoding algorithms, such as the zero-forcing (ZF) algorithm, can be reused for digital BF design. Description is presented below to the design of user scheduling.

[0084] First of all, the embodiments of the present disclosure propose a joint analog BF and wideband user scheduling solution (hereinafter referred to as joint scheduling solution). Considering wideband user scheduling, analog BF can be jointly designed with user scheduling based on the criteria of maximum weighted sum capacity. The analog BF matrix is derived according to equations (4), (5) and (6) from the sum of wideband channel covariance matrices of scheduled users.

[0085] In the joint scheduling solution, since the design of analog BF depends on scheduled users, channel vector h_¾ (b) (1 k ≤Ξ S; 1 ≤Ξ b ≤Ξ B) of all S users on all B sub-bands should be estimated in each scheduling subframe.

[0086] In the joint scheduling solution, since analog BF takes user scheduling into consideration, there is a connection between analog BF and user scheduling. The wideband channel covariance matrix has a slow-varying property, while user scheduling is a fast-varying procedure. Once a result of user scheduling changes, the analog BF matrix needs to be modified. Since the measurement (channel estimation) of analog BF requires a huge overhead, the analog BF matrix desires to be a slow variable. In other words, analog BF matrix is preferred not to be changed once scheduled users vary. Therefore, the embodiments of the present disclosure further propose a separate analog BF and sub-band user scheduling solution (hereinafter referred to as separate scheduling solution).

[0087] In the separate scheduling solution, analog BF indicating beams for all S users are designed based on the sum of wideband channel covariance matrices of all S candidate users, i.e., o ^s (²⁰⁾ ft, =∑R_t

k=l

Then, sub-band user scheduling is performed based on the short-term-level estimation of the effective channel after analog BE Therefore, analog BF and user scheduling is performed separately, and sub-band user scheduling can be supported as well to get frequency selective gain.

[0088] In the separate scheduling solution, the analog BF matrix only needs to be modified when the channel covariance matrix of one user changes, regardless of user scheduling in each subframe. Thus, the update period of analog BF matrices could be slower than the joint scheduling solution. Another advantage of the separate scheduling solution is that only the effective channel vector T^H _k {b) {\ k ≤Ξ S; 1 ^ b ^ B) needs to be estimated in each subframe, instead of the estimation on channel vector h_¾ (b) as in the joint scheduling solution.

[0089] The performance comparison between two user scheduling solutions is shown in Table I. Simulation parameters and assumptions are summarized in Table II. As seen from the comparison result, the separate scheduling solution outperforms the joint scheduling solution with 20% cell-edge gain for 64 transmission antennas and 16 RF chains.

Table I

Performance comparison between wideband and sub-band user scheduling for hybrid architecture 100 (64Tx, 16 RF chains, with ideal analog BF)

Method Scheduling method Cell average capacity Cell edge user capacity

Hybrid architecture Joint scheduling 19.24 (Baseline) 0.409 (Baseline) 100 solution

Separate scheduling 19.71 (2%) 0.489 (20%) solution

[0090] In a second aspect, the embodiments of the present disclosure further propose a method for hybrid BF in a base station. Fig. 5 shows a flowchart of a method 500 for hybrid BF in a base station according to the second aspect of the embodiments of the present disclosure. Method 500 starts with step S510 in which a wideband analog BF matrix is calculated based on long-term-level estimation of a physical channel. In step S520, the wideband analog BF matrix is applied to the physical channel to obtain an effective channel of the physical channel. In step S530, a sub-band digital BF matrix is calculated and sub-band user scheduling is performed based on short-term-level estimation of the effective channel. In step S540, hybrid BF is performed on downlink signals for a scheduled user with the wideband analog BF matrix and the sub-band digital BF matrix.

[0091] In one embodiment, method 500 further comprises: sending to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively; receiving from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and constructing, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

[0092] It should be understood the above-described contents of user scheduling, training signal design, channel estimation and feedback of channel estimation results for the hybrid beamforming method according to the first aspect of the embodiments of the present disclosure are also applicable to method 500. For the brevity purpose, the details of the method 500 are omitted.

[0093] In a third aspect, the embodiments of the present disclosure further propose a method for hybrid BF in a mobile terminal. Fig. 6 shows a flowchart of a method 600 for hybrid BF in a mobile terminal according to the third aspect of the embodiments of the present disclosure. Method 600 starts with step S610 in which a horizontal sub-channel covariance matrix associated with a horizontal linear array of antennas of a base station is estimated based on a first training signal received from the horizontal linear array. Then, in step S620, a vertical sub-channel covariance matrix associated with a vertical linear array of antennas of the base station is estimated based on a second training signal received from the vertical linear array. In step S630, feedbacks of estimation of the horizontal sub-channel covariance matrix and estimation of the vertical sub-channel covariance matrix is provided to the base station.

[0094] In one embodiment, method 600 further comprises: associating the horizontal sub-channel covariance matrix with a first transmission correlation coefficient and associating the vertical sub-channel covariance matrix with a second transmission correlation coefficient; and wherein providing the base station with feedbacks of estimation of the horizontal sub-channel covariance matrix and estimation of the vertical sub-channel covariance matrix comprises: providing the base station with feedbacks of an amplitude and a phase of the first transmission correlation coefficient as well as an amplitude and a phase of the second transmission correlation coefficient.

[0095] In one embodiment, method 600 further comprises: estimating an effective channel based on a third training signal received from the base station; and providing the base station with feedback of estimation of the effective channel.

[0096] In a fourth aspect, the embodiments of the present disclosure further propose an apparatus for hybrid BF in a base station. Fig. 7 shows a block diagram of an apparatus 700 for hybrid BF in a base station according to the fourth aspect of the embodiments of the present disclosure. As shown in Fig. 7, apparatus 700 comprises: a long-term- level estimating unit 710 configured to calculate a wideband analog BF matrix based on long-term-level estimation of a physical channel; a quantizing unit 720 configured to quantize the wideband analog BF matrix to obtain a quantized wideband analog BF matrix; an effective channel obtaining unit 730 configured to apply the quantized wideband analog BF matrix to the physical channel to obtain an effective channel of the physical channel; a short-term-level estimating unit 740 configured to calculate a sub-band digital BF matrix based on short-term-level estimation of the effective channel; and a hybrid beamforming unit 750 configured to perform hybrid beamforming on downlink signals with the sub-band digital BF matrix and the quantized wideband analog BF matrix.

[0097] In one embodiment, quantizing unit 720 is further configured to: normalize an amplitude of each of non-zero elements in the wideband analog BF matrix; and with respect to the each of non-zero elements, perform phase search one by one element in a predetermined phase set to select a phase that maximizes capacity of the effective channel.

[0098] In one embodiment, apparatus 700 further comprises: a scheduling unit configured to perform sub-band user scheduling based on the short-term-level estimation of the effective channel.

[0099] In one embodiment, apparatus 700 further comprises: a sending unit configured to send to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively; a receiving unit configured to receive from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and a constructing unit configured to construct, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

[00100] In a fifth aspect, the embodiments of the present disclosure further propose an apparatus for hybrid BF in a base station. Fig. 8 shows a block diagram of an apparatus 800 for hybrid BF in a base station according to the fifth aspect of the embodiments of the present disclosure. As shown in Fig. 8, apparatus 800 comprises: a long-term- level estimating unit 810 configured to calculate a wideband analog BF matrix based on long-term-level estimation of a physical channel; an effective channel obtaining unit 820 configured to apply the wideband analog BF matrix to the physical channel to obtain an effective channel of the physical channel; a short-term-level estimating unit 830 configured to calculate a sub-band digital BF matrix and perform sub-band user scheduling, based on short-term-level estimation of the effective channel; and a hybrid BF unit 840 configured to perform hybrid BF on downlink signals for a scheduled user with the wideband analog BF matrix and the sub-band digital BF matrix.

[00101] In one embodiment, apparatus 800 further comprises: a training signal sending unit configured to send to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively; a feedback receiving unit configured to receive from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and a constructing unit configured to construct, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

[00102] In a sixth aspect, the embodiments of the present disclosure further propose an apparatus for hybrid BF in a mobile terminal. Fig. 9 shows a block diagram of an apparatus 900 for hybrid BF in a mobile terminal according to the sixth aspect of the embodiments of the present disclosure. As shown in Fig. 9, apparatus 900 comprises: a first estimating unit 910 configured to estimate a horizontal sub-channel covariance matrix associated with a horizontal linear array of antennas of a base station based on a first training signal received from the horizontal linear array; a second estimating unit 920 configured to estimate a vertical sub-channel covariance matrix associated with a vertical linear array of antennas of the base station based on a second training signal received from the vertical linear array; and a feedback unit 930 configured to provide the base station with feedbacks of estimation of the horizontal sub-channel covariance matrix and of estimation of the vertical sub-channel covariance matrix.

[00103] In one embodiment, apparatus 900 further comprises: an associating unit configured to associate the horizontal sub-channel covariance matrix with a first transmission correlation coefficient and associate the vertical sub-channel covariance matrix with a second transmission correlation coefficient; and the feedback unit 930 is further configured to provide the base station with feedbacks of an amplitude and a phase of the first transmission correlation coefficient as well as of an amplitude and a phase of the second transmission correlation coefficient.

[00104] In one embodiment, apparatus 900 further comprises: an effective channel estimating unit configured to estimate an effective channel based on a third training signal received from the base station; and the feedback unit is further configured to provide the base station with feedback of estimation of the effective channel. [00105] It should be understood that units comprised in apparatuses 700, 800 and 900 may be implemented in various forms, including software, hardware, firmware or any combination thereof. In one embodiment, one or more units may be implemented using software and/or firmware, for example, machine-executable instructions stored on the storage medium. In addition to or instead of machine-executable instructions, parts or all of the units in apparatuses 700, 800 and 900 may be implemented, at least in part, by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Application- specific Integrated Circuits (ASICs), Application- specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc.

[00106] Hereinafter, description is presented to system-level simulation results of the solution proposed by the present disclosure. In this section, the performance of the proposed hybrid BF method for massive MIMO is verified over 19 sites/57 pentagon- shaped cells. Simulation parameters and assumptions are summarized in Table II. Each base station is equipped with a single-polarized planar array with 8 row and 8 column antennas, and serves 10 single antenna users. Multi-user MEVIO with separate analog BF and sub-band user scheduling (i.e. above-described separate scheduling solution) is employed. Simulations results are shown in Table III- IV.

Hybrid BF vs. Digital BF

[00107] In Table III, hybrid architecture 100 achieves similar performance to digital BF with half RF chains. Further reducing the number of RF chains to a quarter, the performance loss is still limited within 9%. Hybrid architecture 200 has much larger performance loss over full digital BF, with 29% cell average loss and 38% cell-edge loss, mainly due to the less beamforming gain of simplified analog BF.

Table III

Performance impact of RF chain number on hybrid BF (64-Tx, ideal analog BF)

Method Antenna number RF chain Cell average Cell edge user number capacity capacity

Digital BF 64: 8 rows, 8 columns 64 21.56 (Baseline) 0.530 (Baseline) Hybrid 64 8 rows, 8 columns 32 21.48 (-0.4%) 0.525 (-1%) architecture 100 64 8 rows, 8 columns 16 19.71 (-9%) 0.489 (-8%)

Hybrid 64 8 rows, 8 columns 16 15.26 (-29%) 0.326 (-38%) architecture 200

Impact of hardware impairment

[00108] Hardware impairment of 4-bit phase resolution is shown in Table IV. Using the proposed phase search method, the performance loss of hybrid architecture 100 is limited within 4% over ideal analog BF. Hybrid architecture 200 achieves even better performance with up to 15% gain, since the analog BF matrix is further optimized during phase search according to maximum channel capacity criteria.

Table IV

Performance impact of hardware impairment on hybrid BF (64-Tx, 16 RF chains)

Table II

Simulation parameters and assumptions

Deployment case Homogenous network, 19 sites, 3 cells per site

Scenario 3GPP case 1 3D

Carrier frequency 2GHz

Duplex mode and bandwidth TDD downlink, 10MHz

Inter-site distance 500m

Traffic model Full buffer

User number per sector S=10 users per sector with uniform distribution

BS antenna 64Tx: 8 rows, each with 8 antennas, half-wavelength antenna spacing

User antenna One antenna

MIMO mode MU-MIMO

Precoding scheme Per subband precoding, 10 subbands, equal power allocation

- Digital BF: ZF method - Hybrid BF: ZF method for digital BF

Maximum scheduled users - Digital BF: min(N_t, S)

- Hybrid BF: min(N, S)

Channel estimation Non-ideal with SRS channel estimation error modeling

Scheduling Proportional fairness scheduling based on greedy search

Feedback scheme Interference-plus-noise power feedback, 5ms feedback period, 5ms feedback delay

Simulation time 300 sub-frames (1ms per sub-frame)

[00109] Although the present disclosure has been described with reference to specific embodiments, it is obvious to those skilled in the art that the present disclosure is not limited to details of the embodiments described above and the present disclosure may be implemented by various adaptations and modifications without departing the scope of the present disclosure. Therefore, the current embodiments are considered to be exemplary instead of limiting in any manner, the scope of the present disclosure is represented by claims instead of the foregoing description, and all adaptations falling within the equivalent meaning and scope of claims are thus included in the scope of the present disclosure.

Claims

I/We Claim:

1. A method for hybrid beamforming in a base station, comprising:

calculating a wideband analog beamforming matrix based on long-term-level estimation of a physical channel;

quantizing the wideband analog beamforming matrix to obtain a quantized wideband analog beamforming matrix;

applying the quantized wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel;

calculating a sub-band digital beamforming matrix based on short-term-level estimation of the effective channel; and

performing hybrid beamforming on downlink signals with the sub-band digital beamforming matrix and the quantized wideband analog beamforming matrix.

2. The method according to claim 1, wherein quantizing the wideband analog beamforming matrix comprises:

normalizing an amplitude of each of non-zero elements in the wideband analog beamforming matrix; and

with respect to each of the non-zero elements, performing phase search one by one element in a predetermined phase set to select a phase that maximizes a capacity of the effective channel.

3. The method according to claim 1, further comprising:

performing a sub-band user scheduling based on the short-term-level estimation of the effective channel.

4. The method according to claim 1, further comprising:

sending to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively;

receiving from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and

constructing, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

5. A method for hybrid beamforming in a base station, comprising:

applying the wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel;

calculating a sub-band digital beamforming matrix and performing a sub-band user scheduling based on short-term-level estimation of the effective channel; and

performing hybrid beamforming on downlink signals for a scheduled user with the wideband analog beamforming matrix and the sub-band digital beamforming matrix.

6. The method according to claim 5, further comprising:

7. A method for hybrid beamforming in a mobile terminal, comprising:

estimating a horizontal sub-channel covariance matrix associated with a horizontal linear array of antennas of a base station based on a first training signal received from the horizontal linear array;

estimating a vertical sub-channel covariance matrix associated with a vertical linear array of antennas of the base station based on a second training signal received from the vertical linear array; and

providing the base station with feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix.

8. The method according to claim 7, further comprising:

associating the horizontal sub-channel covariance matrix with a first transmission correlation coefficient and associating the vertical sub-channel covariance matrix with a second transmission correlation coefficient; and

wherein providing the base station with feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix comprises: providing the base station with feedback of an amplitude and a phase of the first transmission correlation coefficient as well as feedback of an amplitude and a phase of the second transmission correlation coefficient.

9. The method according to claim 7, further comprising:

estimating an effective channel based on a third training signal received from the base station; and

providing the base station with feedback of estimation of the effective channel.

10. An apparatus for hybrid beamforming in a base station, comprising:

a long-term-level estimating unit configured to calculate a wideband analog beamforming matrix based on long-term-level estimation of a physical channel;

a quantizing unit configured to quantize the wideband analog beamforming matrix to obtain a quantized wideband analog beamforming matrix;

an effective channel obtaining unit configured to apply the quantized wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel;

a short-term-level estimating unit configured to calculate a sub-band digital beamforming matrix based on short-term-level estimation of the effective channel; and a hybrid beamforming unit configured to perform hybrid beamforming on downlink signals with the sub-band digital beamforming matrix and the quantized wideband analog beamforming matrix..

11. The apparatus according to claim 10, wherein the quantizing unit is further configured to:

normalize an amplitude of each of non-zero elements in the wideband analog beamforming matrix; and

with respect to each of the non-zero elements, perform phase search one by one element in a predetermined phase set to select a phase that maximizes a capacity of the effective channel..

12. The apparatus according to claim 10, further comprising:

a scheduling unit configured to perform sub-band user scheduling based on the short-term-level estimation of the effective channel.

13. The apparatus according to claim 10, further comprising:

a sending unit configured to send to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively;

a receiving unit configured to receive from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and

a constructing unit configured to construct, based on the feedbacks, a kronecker product of the horizontal sub-channel covariance matrix and the vertical sub-channel covariance matrix as a wideband channel covariance matrix of the physical channel.

14. An apparatus for hybrid beamforming in a base station, comprising:

an effective channel obtaining unit configured to apply the wideband analog beamforming matrix to the physical channel to obtain an effective channel of the physical channel;

a short-term-level estimating unit configured to calculate a sub-band digital beamforming matrix and perform sub-band user scheduling based on short-term-level estimation of the effective channel; and

a hybrid beamforming unit configured to perform hybrid beamforming on downlink signals for a scheduled user with the wideband analog beamforming matrix and the sub-band digital beamforming matrix.

15. The apparatus according to claim 14, further comprising:

a training signal sending unit configured to send to a mobile terminal a first training signal and a second training signal respectively via a horizontal linear array and a vertical linear array of the base station, so that the mobile terminal estimates a horizontal sub-channel covariance matrix associated with the horizontal linear array and a vertical sub-channel covariance matrix associated with the vertical linear array based on the first training signal and the second training signal, respectively;

a feedback receiving unit configured to receive from the mobile terminal feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix, respectively; and

16. An apparatus for hybrid beamforming in a mobile terminal, comprising:

a first estimating unit configured to estimate a horizontal sub-channel covariance matrix associated with a horizontal linear array of antennas of a base station based on a first training signal received from the horizontal linear array;

a second estimating unit configured to estimate a vertical sub-channel covariance matrix associated with a vertical linear array of antennas of the base station based on a second training signal received from the vertical linear array; and

a feedback unit configured to provide the base station with feedback of estimation of the horizontal sub-channel covariance matrix and feedback of estimation of the vertical sub-channel covariance matrix.

17. The apparatus according to claim 16, further comprising:

an associating unit configured to associate the horizontal sub-channel covariance matrix with a first transmission correlation coefficient and associate the vertical sub-channel covariance matrix with a second transmission correlation coefficient; and wherein the feedback unit is further configured to provide the base station with feedback of an amplitude and a phase of the first transmission correlation coefficient as well as feedback of an amplitude and a phase of the second transmission correlation coefficient.

18. The apparatus according to claim 16, further comprising:

an effective channel estimating unit configured to estimate an effective channel based on a third training signal received from the base station; and

the feedback unit is further configured to provide the base station with feedback of estimation of the effective channel.