CN100493053C - Method for channel estimation in multi-antenna system - Google Patents

Abstract

The method of multi-antenna channel estimation system relates to a method in a multiple antenna system guide bis selective fading channel environment the design and pilot channel estimation. At the transmitting end, by using the discrete Fourier transform matrix least-squares sense the best constructed cyclic orthogonal pilot sequences, and inserted intermittently transmit data in the time slot structure composed of a double loop; at the receiving end, by the circulation n characteristics of the orthogonal sequence, to a low optimal implementation complexity least square channel estimation meaning of minimum mean square error, using the received pilot matrix decomposition, rapid channel estimation implemented using discrete cosine transform more band guide to accurately estimate the channel and noise variance estimates, and then using the discrete cosine transform domain interpolation to obtain channel parameters estimated data channel segment. This method can improve the channel estimation accuracy, improve the performance of a receiver in particular conventional channel estimation performance of the receiver and the transmission of high-speed movement of a method difficult to secure.

Description

Multi-antenna system channel estimation method

FIELD

The present invention relates to a broadband mobile communication system for high-speed data transmission by using a plurality of transmit / receive antennas, and particularly to a method for multi-antenna system in the dual guide selective fading channel environment the design and pilot channel estimation ·

Choi back *

To meet the needs of future development, ultra-generation mobile communication system must be able to: support full-speed IP packet data transmission, the data rate several ten trillion bps (bit per second) or even several hundred bps; support high terminal mobility, a moving speed several hundred kilometers per hour; child support high transmission quality, the error rate is lower than the data traffic ΙΟ "6; Gao frequency efficiency, the number of bits per hertz above; to provide high power efficiency, reduced transmit power 10 dB above, - effective support changes in user data rate, user capacity, service quality and moving speed ft large dynamic range.

In order to improve the spectrum utilization of the system, the use of multi-antenna transmission and multi-antenna receive air interface mechanism is an effective solution · However, even in a multi-antenna environment, reliable and effective support for high speed data transmission, ultra-generation mobile communication the system still be very Pei-bandwidth broadband transmission from increased frequency selectivity of the channel fading phenomenon, resulting in severe multipath interference; Gao and the speed of the mobile terminal caused by Doppler shift phenomenon is exacerbated channel time selective fading. Thus, the ultra-generation mobile communication system, the channel is off Gunn dual-selective

A receiver in a communication system into a coherent receiver and two kinds of non-coherent receiver. Pain known to one thousand with the receiver at the receiving end of the channel impulse response coefficients, requiring channel estimation at the receiving end; Pei noncoherent receiver is known to the receiving end of the channel impulse response coefficients are not, but requires the transmission signal quadrature modulation system, and there will be a loss in performance of 3 * 4dB. The present invention is primarily coherent receiver embodiment considered dominant ultra generation mobile communication system.

In order to achieve phase one thousand received, channel estimation is required at the receiving end. In order to timely and accurately estimate the channel parameters, the actual communication systems often employ channel estimation pilot sequences based · The basic idea is: inserting a pilot at the transmit end position intermittent receiving side using the pilot recovered pilot positions channel information, and then using some processing means (e.g., interpolation, filtering, transformation, etc.) to obtain channel information for all time periods. Here three main problems: (the I) the sending end a pilot selection and insertion: (2) receiving Yasushi guide way pilot position of the channel information acquired; (3) acquired by the pilot position channel information restoring all the time channels information. The present invention presents a approximate the optimal performance and low complexity of the technical solution to these three questions.

SUMMARY

Technical Q employed * object of the present invention is to provide a method for multi-antenna system channel estimation, the method can improve the channel estimation accuracy, improve speed and shift movement of the receiver performance in particular conventional channel estimation method is difficult to guarantee performance of the receiver.

Technical Solution: The method of multi-antenna system according to the present invention the channel estimate, at the transmitting end, by using a discrete Fourier transform (DFT) matrix constructed least squares (LS) in the sense of an optimal cycle orthogonal pilot sequences, and transmitting data intermittently inserted in slot structures of two-cycle: at the receiving end, the cycle characteristics by using the orthogonal sequence, to achieve low complexity the minimum mean square error (MMSE) sense optimal least-squares channel estimation using received pilot decomposition pilot matrix, channel estimation fast implementation using discrete cosine transform (DCT) for guiding the band more accurate channel estimation and noise variance estimates, and then using a discrete cosine transform (DCT) obtained by interpolation segment channel the parameter estimation method comprising the following steps:

The I multipath step number), the sending end, the multi-antenna system of transmitting antennas and the number of channels Gum

P, constructed length

Best least squares sense the best pilot sequence on the orthogonal pilot sequences cycle s, and follow these rules to generate each transmitting antenna:

Step 2), at the receiving end, the following equation is obtained by the impact of a channel time slot in response to each of the conduction band parameter estimation

Step 3), at the receiving end, using Step 2) estimated channel impulse response according to the following equation for each channel a guide band noise variance estimate, and to obtain the noise variance estimate for the current slot:

Step 4), at the receiving end, using step 2) all estimated pilot segment of the channel impulse response, which by-point processing in the DCT domain denoising and noise variance estimates, and by converting the DCT domain interpolation in the discrete cosine data segment obtained channel impulse response

The process cycle S orthogonal pilot sequences constructed from the Fourier transform matrix, and satisfies the orthogonality of the cycle: pilot sequences for each transmission antenna obtained by the shift cycle Shu; refers to the least squares estimation error and the minimum squared-estimation of the channel impulse response parameter of step 2 is carried out sub-slot, using the following matrix decomposition received pilot, estimate the channel impulse response parameter of the fast algorithm;

The process cycle S orthogonal pilot sequences constructed from the Fourier transform matrix, and satisfies the orthogonality of the cycle; pilot sequences for each transmission antenna obtained by the shift cycle s: the least squares error estimation means The minimum sum of squares. Step 2 The impact of channel estimation response parameter is a sub-time slot; using the received pilot matrix decomposition as estimated channel impulse response parameters fast algorithm:

The impulse response of the channel parameter estimation step 4 is performed within a time slot; the channel impulse response and the interpolation parameter denoising domain is a discrete cosine transform DCT performed.

Advantageous Effects: The present invention provides a method of constructing a channel estimation and a method for multi-antenna transmission system capable of channel estimation pilot sequences generated according to the method of the present invention the pilot sequence, can be minimal with low computational complexity the least squares sense the best channel estimation; while using time-domain correlation characteristics of the channel, the processing by the transform domain, obtaining a near-optimal solution of low complexity MMSE sense, to further improve the accuracy of channel estimation. The channel estimation method provided by the present invention compared to the prior art, can effectively provide Gao channel estimation accuracy, improve the performance of a receiver in particular conventional channel estimation method is difficult to ensure the performance of the receiver and the transmission of the movement of high speed. This method does not require channel estimation of the pilot sequence length, and the computation amount and memory capacity are very small, easy hardware implementation - Figure I is employed in the present invention an intermittent pilot frequency time slot structure, wherein the sugar ring is protected segments G, the guide band P, the number of data segments D · K sub-slot adaptively in accordance with the movement speed of the terminal.

FIG 2 is a schematic view of application of the method of the present invention is configured by a base sequence of each antenna pilot sequences.

FIG 3 is a schematic diagram of a present invention, the least squares channel estimation fast implementation apparatus which reordering module for adjusting the signal sequence; phase rotation module for rotating the phase of the signal: FFT and IFFT group set ·

DETAILED FIG. 4 is a block diagram of the present invention an apparatus for channel estimation. Including least squares estimation module for each frequency channel of the received pilot signal; DCT-block discrete cosine transform domain for all time estimates; single-point DCT domain denoising and noise variance estimation module: the signal inverse discrete cosine transform IDCT module.

Specific Choi formula · To make the objectives, technical solutions, and advantages of the present invention will become apparent Qing Chu, in conjunction with the accompanying drawings of embodiments of the technical solutions described in further detail:

At the transmitting end, by using a discrete Fourier transform (DFT) matrix constructed least squares (LS) in the sense of an optimal cycle orthogonal pilot sequences, and inserted intermittently transmit data in the time slot structure composed of a double loop; in the receiving end, the cycle characteristics by using the orthogonal sequence, to achieve low complexity the minimum mean square error (MMSE) sense optimal least-squares channel estimation using received pilot pilot matrix decomposition, channel estimation of fast implementation estimated using a discrete cosine transform (DCT) for guiding the band more accurate channel estimation and noise variance estimates, and then using a discrete cosine transform (DCT) obtained by interpolation segment channel parameters.

The method includes the following steps Ke:

The I step), the transmission side, the number of dead multipath multiple antenna system according to the number of transmission antennas and qe channel length construct optimal least-squares sense loop orthogonal pilot sequences s, press follows rule generating each transmission antenna pilot sequences:

Step 2), at the receiving end, the following equation is obtained by the impact of a channel time slot in response to each of the conduction band parameter estimation

Destroy Step 3), at the receiving end, the use of too-step 2) estimated channel impulse response according to the following equation ϋ each primary channel conduction band voice sound estimation variance, and the resulting current slot throat sound variance estimate:

The receiving end, using Step 2) all estimated pilot segment of the channel impulse response, which denoising and estimation of noise variance in the point-wise DCT domain, and impulse response data obtained by DCT domain interpolation segment in channel

The process cycle * orthogonal pilot sequences constructed from the Fourier transform matrix, and satisfies the orthogonality of the cycle: pilot sequence for each transmit antenna cyclic shifting by S; refers to the least squares estimation error The minimum sum of squares. The impulse response of the channel parameter estimation step 2 sub-time slot is performed, using the received pilot guide below the decomposition matrix, the estimated channel impulse response parameters fast algorithm; the channel impulse response estimate parameters step Midi 4 is performed within a time slot; channel impulse response parameter · denoising and interpolation are carried out in the DCT domain

I. Qi-type system

Figure I shows the transmitted pilot signal intermittently β slot structure is provided in each slot there are particular sub-slot, the number of its conduction band of a sufficient number of sub-slot + 1 · i: speed of the movable terminal in accordance with set.

In the MIMO system, the number of transmission antennas is set AV, the number of receiving antennas is the channel number of the channel impulse response length of the sequence of parameters P, then each receiving channel to be estimated for qe χΡ, correspondingly, the pilot sequences to make the length (wherein

1 represents the smallest integer not less X).

use

Eurya - ^ represents η transmit antennas pilot sequence, then S = [se, s, f represents transmission guide all antennas pilot signal which first it received signal conduction band after the receiving end to remove the loop protection may represent for:

[Formula I]

Wherein 0> and Zw are qe XLp matrix, respectively, of the received pilot signal and the variance of the additive white Gaussian tr2 of the voice sound;

Qe is due to make a matrix, Ag (p) represents the first Jfc a pilot field, first

w receive antenna corresponds to the "root path between the transmitting antennas of P channel taps coefficients;

L1P -

· S is a right circular column P bits obtained

make

All channel parameters indicates the k-th band guide,

Then the formula I can be rewritten as:

Equation 2

Obtained by the linear model described in Equation 2 channel parameters of the least squares (LS) estimation of:

[Formula 3]

When do> Λ ^ Ρ, you can get the noise variance unbiased estimator is:

[Formula 4]

|| F where F represents the norm of a matrix. Theoretical analysis shows that, at that time, the above estimate LS has the best performance and avoids matrix inversion ·

2. do the best guide column structure

Setting s = length of the cycle is to make the orthogonal sequence, as we will base sequence, each antenna configuration according to the following criteria pilot sequence

[Equation 5]

Wherein I * represents a modulo operation of # * can verify, when the "when the orthogonal sequence is a loop, according to the formula

5 pilot sequence constructed orthogonal to satisfy the foregoing conditions.

Construction related to the cycle of the orthogonal sequence, have been reported in the literature, we found 22 "and 22 = best" circulation orthogonal sequence Ip == -1 can be obtained from

Point DFT matrix element directly. AT Let W point DFT matrix, its elements and provided W = [W. W1], wherein W. And W1 are 2 "X2" · Sub moment Lp 1 = 22 "-1, so that W = (W. + jVix) / S, the s -vec {ViT) length 22 is" orthogonal circular-1 vector sequences. Here, for the stretch operator ·

3. The implementation method of least squares channel estimation Yoo

When orthogonal matrix X satisfying the conditions do XXw = Ia ^, Equation 3 can be simplified to:

Receiver can receive directly after the Yw equation 6 for channel estimation, compared with formula 3, to avoid the complexity of the matrix inversion. Further study of the structure of the matrix X, we found it to break down in two forms as follows:

In Equation 7, F is the best point DFT matrix, Γ = | also ig {Fa}, // dagger is generated by the permutation order WrP permutation matrix; In Equation 8, Λ is a diagonal order to make, PeiPp respectively by substitution ", order # Zp generated permutation matrix, Q-LJN. cushions fd Shu wherein Λ represents the main diagonal elements of the diagonal d, matrix ® denotes Kronecker product. Each permutation generation rule as follows:

8 diagonal matrix Λ may be extrapolated from its anti-diagonal elements of the formula. By Equation 6,7, 8, we obtain an initial three algorithms LS channel estimation:

(1) direct calculation (Equation 6), the plural number of multiplications is qe qe & corpse;

(2) a fast algorithm (equation 7), which is the number of complex multiplication (I + Iog2 (M));

(3) two fast algorithm (equation 8), which is the number of complex multiplication qe do (I + Iog2 V) ·

Figure 2 shows a view of a device realized by Equation 8 LS channel estimation "4 more accurate channel estimation given above and the noise variance estimation least square channel estimation method based on a single pilot band in the adaptive loop bis slot arrangement, a plurality of conductive bands, using the estimated time correlation of channel parameters may be more accurate channel estimation * Further, when ^ Lp-NtP, using equation 4 can not be estimated noise variance, then estimating channel parameters may be utilized correlation time voice sound variance.

Note (P) © (P)] ', represents ( ", m) of all channels available on the AT + 1 channel parameters obtained from the first ;; th path transmission channel, then the first (m, ri) transmission channel parameters can be written as:

There are:

Wherein, Iiitiw with the respective channel vectors over Yoo, IU-zero-mean white Gaussian noise vector, each element of which is the variance σ1, Lp ·

Seen from Equation 9, Ive minimum mean square error (MMSE) estimate is:

Wherein R = qhe> nIiiy · Here, we assume that each of the transport channels have the same power delay profile (PDP), further, the use of channel statistics in time and frequency domains detachable nature, the above correlation matrix R can be decomposed into : R = Rzs ^ Rowf, wherein Risf = also 4 is the power channel P-th path is: Rdp «statistical characteristics of the channel time-domain shift is determined by the Doppler-use decomposed R, the equation 10 may be dimension reduction realized, which is:

Wherein R17 = {h ugly knock (p) h ^ (/ »)} = P2pRdpk.

Since Rp is Hermite moo, wherein it can be decomposed into ·· Rp = UwApU, where U is an orthogonal matrix, Λρ = ^ {Λλ0, Λλ1, ..., Λ ^} ο decomposition With this feature, Equation 11 may be rewritten as _·among them

In order to achieve O ») * estimate of 1fflisE Pei Found correlation matrix to be decomposed and wherein Taking into account the different transmission channels with the same R ,, it can be directly on the sample space Rp is estimated, i.e.:

In order to avoid decomposition of the complex matrix eigenvalue calculation, the time domain correlation matrix of basic research on the properties, we use the discrete cosine transform (DCT) to approximate the correlation matrix of eigen-decomposition, i.e., instead of the above DCT matrix with n + 1 points Fan deposit matrix U, and point by point in the DCT transform domain denoising processing and noise variance estimates, and h "i (1 approximate solution (/>) MMSE estimate obtained by inverse DCT. theoretical analysis and simulation results have confirmed that this the method is effective.

Times the channel estimation section 5. Data

After acquiring the pilot channel parameter bands, parameter data required for the channel or track segments prediction ·

Commonly used methods are: linear interpolation, linear interpolation Gaussian weighted multi-slot averaging (WMSA), these methods are simple linear process, a common drawback thereof is that when the mobile station is too fast speed, fast channel fading transform , or non-linear change occurs, so that the channel using the pilot channel as the linear processing data obtained can not reflect the changes in the channel.

Here we interpolate the channel parameters in the DCT domain, the specific process is as follows: Firstly, especially with the + DCT 1 point (P) transformed into the DCT domain, in particular to obtain a +1 dimensional vector, in which the end of the complement (ft + 1 ) (1-1)

A zero element, dimensional vector obtained, wherein Z is the interpolation factor in the data segment, i.e., interpolated values ​​in each value ZI behind the sample, then the final point inverse DCT to transform it back into the time domain. Since the data edge effect DCT interpolation tail (II) is not very precise, taking into account not need these data in the subsequent detection, it is obtained after removing the trailing the (i-Ι) data-interpolated channel parameter length binding of KL + U equation 12, the above described process can be formulated as:

Where S is

Claims (5)

1, a multi-antenna channel estimation system, characterized in that: the transmitting side, using a discrete Fourier transform "DFT" constructed matrix least-squares "LS" optimal pilot cycle in the sense orthogonal pilot sequence S, and transmitting data intermittently inserted in the slot structure of the double loop consisting of; at the receiving end, the cycle characteristics by using the orthogonal sequence, with low implementation complexity on the minimum mean squared error "MMSE" best least square sense by channel estimation using the decomposition received pilot matrix quickly realize channel estimation using a discrete cosine transform "DCT" for guiding the band more accurate channel estimation and noise variance estimates, and then using a discrete cosine transform "DCT" obtained by interpolation estimating a channel parameter data channel segment, the method comprising the following steps: the I step), the transmission side, the number of multi-path multi-antenna system according to P and the number of channels transmitting antenna, configured to make the length of the least-squares sense optimal cycle turned on orthogonal pilot sequences S, and follow these rules to generate a pilot sequence for each transmit antenna:

Step 2), at the receiving end, the following equation is obtained by the impact of a channel time slot in response to each of the conduction band parameter estimation

Wherein: X is a matrix consisting of a pilot signal, Yw A first sub-slots of the received pilot signal; step 3), at the receiving end, using Step 2) estimated channel impulse response for each of the following formula a band guide channel noise variance estimates, and the resulting noise variance estimate for the current slot:

Wherein: represents the number of reception antennas outside; step 4), at the receiving end, using step 2) all estimated pilot segment of the channel impulse response, which by-point processing in the DCT domain denoising and noise variance estimates, and impulse response data segment obtained by DCT domain interpolation channel

Wherein: represents the "on / 7-path channel impulse response between the m-th transmitting antennas and receive antennas, U and denote DCT matrix Ξ and extended DCT matrix, Tp is the filter matrix.

2. The method of channel estimation in multi-antenna system according to claim I, wherein the method s circulating orthogonal pilot sequences constructed from the Fourier transform matrix, and satisfies the orthogonality of the cycle; each transmission antenna the pilot cycle s obtained from sequence shift.

3. The method of claim I of the multi-antenna system channel estimation, wherein the least squares means and the minimum square error estimation.

4. The method of claim estimated required channels in a multiple antenna system of claim I, wherein the step of the scale parameter of the channel impulse response estimation is performed sub slot; as follows using the received pilot matrix decomposition, channel impulse estimated response parameters of fast algorithm:

Wherein:? Pa, P ^, "are permutation matrix, W is the DFT matrix, Λ is a diagonal matrix, Q is a constant.

5. The method of claim estimated required channels in a multiple antenna system of claim I, wherein the step 2) of the impulse response of the channel parameter estimation is performed within one time slot; the channel impulse response and the interpolation parameter denoising are in the discrete Fourier transform "DFT" field carried out.