CN1281003C - Time-domain adaptive channel estimating method based on pilot matrix - Google Patents

Abstract

The present invention relates to a time domain self adaptive channel estimating method based on a pilot frequency matrix. A guiding ideology is the repression of noise horsepower, and the pilot frequency matrix with the deblocking cycle characteristics is combined with an LMS self adaptive arithmetic with minimum mean square error to realize the time domain self adaptive channel estimation of an MIMO system. First, a basic pilot frequency matrix block and an expand matrix are generated respectively according to channel realistic models and desired channel estimation lengths, then, the cross product of the basic pilot frequency matrix block and the expand matrix is computed to obtain the pilot frequency matrix which is used as a pilot frequency transmitting signal to be transmitted through a transmitting antenna, and a receiving antenna receives the pilot frequency matrix after going through a channel and uses the pilot frequency matrix as a pilot frequency receiving signal; then, the self adaptive time domain channel estimation of an actual channel is carried out by using the LMS self adaptive arithmetic through extracting the internal relation of the pilot frequency transmitting signal and the pilot frequency receiving signal. The present invention can compensate the influence caused by the signal synchronization error, the estimation frequency is not limited by the pilot frequency data frame length, and the estimation result has small influence of the Doppler shift. The present invention is suitable for being used in a high-speed moving environment.

Description

Time-domain adaptive channel estimation methods based on pilot matrix

Technical field

The present invention relates to a kind of time-domain adaptive channel estimation methods based on pilot matrix, can carry out channel estimating to the wireless communication system of multiple-input and multiple-output (MIMO), the information transmission technology during the field of wireless transmission that belongs to information, particularly mobile communication, Digital Television etc. are used.

Background technology

MIMO (multiple-input and multiple-output) system can effectively improve the availability of frequency spectrum of channel.Existing research points out, under the certain situation of required signal-to-noise ratio and spectrum efficiency, when reception antenna number during more than or equal to the transmitting antenna number, power system capacity will be with the number of transmit antennas linear growth.

The channel estimating of mimo system can be divided into time domain and estimate to estimate two classes with frequency domain.Li Y is at " ChannelEstimation for OFDM Systems with Transmitter Diversity in Mobile WirelessChannels " (IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL.17, NO.3, MARCH 1999) in, the frequency domain channel of mimo system is estimated to study.The relative time domain of frequency domain channel algorithm for estimating is estimated fairly simple, but generally only is applicable to multi-carrier transmission (for example OFDM) and supposes that channel parameter is constant in a bursty data piece (for example OFDM symbol); And time domain is estimated only to suppose that channel parameter is constant in one group of pilot frequency sequence transport process.With the ofdm system is example, and under the many situations of sub-carrier number, pilot sequence length is much smaller than the OFDM symbol lengths, so the estimated efficiency that time domain is estimated is higher than the frequency domain estimation, and for changing channel faster, its estimation effect is also better.But because time domain channel estimates inevitably to need to consider multipath effect (multipath effect can be avoided considering by adopting modulation systems such as OFDM) in frequency domain is estimated, when the multipath estimated length is big, its complexity is very high, so what present most of mimo system adopted is that frequency domain channel is estimated, under the situation of high data rate, fast moving, the effect of channel estimating is poor, exposed the frequency domain channel estimated efficiency low, be subjected to the big shortcoming of Doppler frequency shift (closely related) influence with translational speed.

Summary of the invention

The objective of the invention is to defective, provide a kind of complexity relatively low time-domain adaptive channel estimation methods, can carry out the channel estimating of mimo system quickly and accurately based on pilot matrix at the frequency domain estimation.Have and can suppress interchannel noise, estimate that frequency is not subjected to the restriction of pilot data frame length, fast convergence rate, the influence that can the compensating signal synchronous error causes is fit to be applied to advantages such as high-speed mobile environment.

To achieve these goals, the present invention is a guiding theory to suppress noise power, and the pilot matrix that will have the piecemeal cycle characteristics combines with LMS (least mean-square error) adaptive algorithm, realizes the time-domain adaptive channel estimating of mimo system.Method of the present invention generates basic pilot matrix piece and extended matrix according to channel realistic model and required channel estimating length at first respectively, asks both Kronecker long-pending (Kronecker product has another name called cross product) then, thereby obtains pilot matrix.Send pilot matrix (pilot transmission signal) by transmitting antenna afterwards, reception antenna receives by the pilot matrix (pilot reception signal) behind the channel, adopt the LMS adaptive algorithm again, by extracting the inner link between pilot transmission signal and the pilot reception signal, carry out the adaptive time domain channel estimating of actual channel.

Method of the present invention specifically may further comprise the steps:

1) choose size according to the number of antennas in the actual channel model (M sends out N and receives) and be the basic pilot matrix A of M*M, it is satisfied: the mould of each matrix element of the equal pairwise orthogonal of row, column (2) of (1) A is 1.

2) according to required multipath estimated length L, the structure twiddle factor ${\mathrm{\ω}}_0=e^{j2\mathrm{\π}/\left(L^{f\left(L\right)}\right)},$ Wherein f (L) is an elementary function about L, can select f (L)=L-1 usually.Use selected twiddle factor, ask its l ^{F (l+1)}(l=0,1 ..., L-1) inferior power makes up vector $B_0=[{\mathrm{\ω}}_0^0,{\mathrm{\ω}}_0^1,{\mathrm{\ω}}_0^{2^{f\left(2\right)}}...,{\mathrm{\ω}}_0^{{(L-1)}^{f\left(L\right)}}].$

3) with B ₀One of ring shift left is as B ₁, again with B ₁One of ring shift left is as B ₂, the rest may be inferred, up to finishing structure B _L-1With B ₀, B ₁... B _L-1Insert successively as the row of extended matrix B respectively, finish the structure of extended matrix.

4) according to extended matrix B and the long-pending B  A of the Kronecker of basic pilot matrix A, obtain pilot matrix, and it is sent by transmitting antenna as the pilot transmission signal, use reception antenna to receive by the pilot matrix behind the channel as pilot reception signal.

5) carry out the initializing set of LMS adaptive algorithm, comprising to the isoparametric initializing set of the step-length of weight matrix vector, channel parameter matrix vector and algorithm.The channel parameter matrix vector is according to pilot reception signal and pilot transmission signal, and the channel parameter that adopts LS (least square) algorithm to try to achieve is expected matrix and set.

6) right to use matrix vector and channel parameter matrix vector are estimated current channel parameter by transversal filter, and with the result as current estimation channel parameter matrix.

7) with the difference of current channel parameter expectation matrix and current estimation channel parameter matrix as error matrix, revise the weight matrix vector, the amplitude of correction is determined by the algorithm step-length.

8) upgrade the channel parameter matrix vector: with first parameter matrix deletion of channel parameter matrix vector, thereafter the parameter matrix room that previous matrix stays that fills vacancies in order of precedence, and current estimation channel parameter matrix inserted the tail of the queue of channel parameter matrix vector, constitute new channel parameter matrix vector.

9) with the result's output of the estimation channel parameter matrix of current time as channel estimating.

10) use revised weight matrix vector and new channel parameter matrix vector, repeating step 6) 7) 8) 9), carry out next channel parameter estimation constantly.

The present invention has the following advantages:

● the basic pilot matrix A that uses method of the present invention to try to achieve has the advantages that autocorrelation is strong, cross correlation weak, can suppress interchannel noise.

● owing to used the multipath channel estimation model of time domain, the influence that can the compensating signal synchronous error causes, and can save frequency domain estimate in necessary protection time slot.

● owing to carry out channel estimating in time domain, estimate that frequency is not subjected to the restriction of pilot data frame length, when adopting the LMS adaptive estimation method, its convergence rate is obviously estimated faster than the adaptive channel of frequency domain.

● owing to carry out channel estimating, only suppose that channel parameter is constant in one group of pilot frequency sequence transport process,, be fit to be applied to high-speed mobile environment so that estimation effect is influenced by Doppler frequency shift is little in time domain.

Description of drawings

Fig. 1 is the channel model of mimo system.

Fig. 2 is the flow chart of LMS algorithm of the present invention.

Embodiment

Below in conjunction with drawings and Examples technical scheme of the present invention is further described.

In mimo channel model shown in Figure 1, transmitting antenna has the M root, and reception antenna has the N root, the channel matrix H that to constitute a size like this be N*M.The matrix relationship formula that multipath situation lower channel is estimated is:

In Ψ=H ∑+v formula, Ψ=[Ψ ₀, Ψ ₁..., Ψ _L-1], Ψ wherein _k(k=0,1 ..., L-1) expression t _KMTo t _{(k+1) M-1}Received signal matrix constantly.H=[H ₀(t), H ₁(t) ..., H _L-1(t)] (H _i(t) time domain channel response of expression i bar multipath channel).In addition, ∑=(∑ _{M, n}) _{M * M}, wherein

∑ _m，n＝∑ _(m+n) m＝0，1，...L-1 n＝0，1，...L-1

∑ _k(k=0,1 ..., 2L-2) expression t _{(k-L+1) M}To t _{(k-L+2) M-1}Transmission signal matrix constantly.

With M=N=4, one 44 collection of letters road models are example, and the step that the present invention carries out channel estimating is as follows:

1) at first according to the number M=4 of transmitting antenna, selected basic pilot matrix A that size is 4*4,

$A=\left[\begin{array}{cccc}1& 1& 1& 1\\ -1& 1& -1& 1\\ -1& 1& 1& -1\\ -1& -1& 1& 1\end{array}\right]$

Basic pilot matrix A satisfies: the mould of each matrix element of the equal pairwise orthogonal of row, column (2) of (1) A is 1.

2) according to required multipath estimated length L, the structure twiddle factor ${\mathrm{\ω}}_0=e^{j2\mathrm{\π}/\left(L^{f\left(L\right)}\right)},$ Wherein f (L) is an elementary function about L, can select f (L)=L-1 usually.Use selected twiddle factor, ask its l ^{F (l+1)}(l=0,1 ..., L-1) inferior power makes up vector $B_0=[{\mathrm{\ω}}_0^0,{\mathrm{\ω}}_0^1,{\mathrm{\ω}}_0^{2^{f\left(2\right)}}...,{\mathrm{\ω}}_0^{{(L-1)}^{f\left(L\right)}}].$ In this example, twiddle factor is taken as ${\mathrm{\ω}}_0=e^{j2\mathrm{\π}/\left(L^{L-1}\right)},$ Accordingly, $B_0=[{\mathrm{\ω}}_0^0,{\mathrm{\ω}}_0^1,{\mathrm{\ω}}_0^2...,{\mathrm{\ω}}_0^{{(L-1)}^{L-1}}].$

3) with B ₀One of ring shift left is as B ₁Again with B ₁One of ring shift left is as B ₂, the rest may be inferred, up to finishing structure B _L-1With B ₀, B ₁... B _L-1Insert successively as the row of extended matrix B respectively, finish the structure of extended matrix.

4) the long-pending B  A of Kronecker that calculates B and A tries to achieve the pilot matrix ∑, and ∑ is sent by transmitting antenna, promptly sends the pilot transmission signal, uses reception antenna to receive by the pilot matrix behind the channel, i.e. pilot reception signal.

5) carry out the initializing set of the relevant channel parameter of LMS adaptive algorithm.Comprising to the weight matrix vector

The channel parameter matrix vector ${\hat{H}}_{n-K},{\hat{H}}_{n-K+1},...,{\hat{H}}_{n-1}$ And the isoparametric initializing set of step size mu of algorithm.In this example, μ is taken as 0.005, All put 1.With the pilot reception signal and the pilot transmission signal that obtain, try to achieve current channel parameter expectation matrix by LS (least square) algorithm

${\hat{H}}_d=\mathrm{\Ψ}{\mathrm{\Σ}}^{-1}$

Wherein Ψ is current pilot reception signal, and ∑ is current pilot transmission signal. ${\hat{H}}_{n-K},{\hat{H}}_{n-K+1},...,{\hat{H}}_{n-1}$ Expect matrix with the channel parameter that solves with the LS method that preceding K receives

Insert.

The setting of parameter influences the performance of algorithm, and when little signal to noise ratio, smaller step size μ can be so that convergence process be steady; During big signal to noise ratio, bigger μ can finish convergence quickly.Therefore, according to different noises choice of dynamical μ recently, can further improve the performance of LMS algorithm.

6) right to use matrix vector And channel parameter matrix vector ${\hat{H}}_{n-K},{\hat{H}}_{n-K+1},...,{\hat{H}}_{n-1},$ Estimate current channel parameter by transversal filter, and with the result

As current estimation channel parameter matrix, as shown in Figure 2.

Current estimation channel parameter matrix ${\hat{H}}_n=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}{\hat{w}}_n\left(i\right){\hat{H}}_{n-K+i-1}$

7) with current channel parameter expectation matrix

With current estimation channel parameter matrix

Difference as error matrix, revise the weight matrix vector Be the adjustment of self adaptation weight vector, as shown in Figure 2.The amplitude of revising is determined by the algorithm step-length:

In the formula

The expression point multiplication operation.

8) upgrade the channel parameter matrix vector: with first parameter matrix of channel parameter matrix vector Deletion, the room that previous matrix stays that fills vacancies in order of precedence of parameter matrix thereafter, and with current estimation channel parameter matrix

Insert the tail of the queue of channel parameter matrix vector, constitute new channel parameter matrix vector

${\hat{H}}_{n-K+1},{\hat{H}}_{n-K+2},...,{\hat{H}}_n.$

9) with the estimation channel parameter matrix of current time

Result's output as channel estimating.

10) use revised weight matrix vector and new channel parameter matrix vector, repeating step 6) 7) 8) 9), estimate next channel parameter constantly.

Claims (1)

1, a kind of time-domain adaptive channel estimation methods based on pilot matrix is characterized in that carrying out according to the following steps:

1) according to the number of transmit antennas M in the actual channel model, to choose size and be the basic pilot matrix A of M*M, it is satisfied: the mould of each matrix element of the equal pairwise orthogonal of row, column (2) of (1) A is 1;

2) according to required multipath estimated length L, the structure twiddle factor ${\mathrm{\ω}}_0=e^{j2\mathrm{\π}/\left(L^{f\left(L\right)}\right)},$ Wherein f (L) is an elementary function about L, selects f (L)=L-1, uses selected twiddle factor, asks its l ^{F (l+1)}(l=0,1 ..., L-1) inferior power makes up vector $B_0=[{\mathrm{\ω}}_0^0,]{\mathrm{\ω}}_0^1,{\mathrm{\ω}}_0^{2^{f\left(2\right)}}...,{\mathrm{\ω}}_0^{{(L-1)}^{f\left(L\right)}}];$

3) with B ₀One of ring shift left is as B ₁, again with B ₁One of ring shift left is as B ₂, the rest may be inferred, up to finishing structure B _L-1, with B ₀, B ₁... B _L-1Insert successively as the row of extended matrix B respectively, finish the structure of extended matrix;

4) according to the cross product B  A of extended matrix B, obtain pilot matrix, and it is sent by transmitting antenna as the pilot transmission signal, use reception antenna to receive by the pilot matrix behind the channel as pilot reception signal with basic pilot matrix A;

5) carry out the initializing set of least mean-square error LMS adaptive algorithm, the initializing set that comprises the step-length of weight matrix vector, channel parameter matrix vector and algorithm, the channel parameter matrix vector is according to pilot reception signal and pilot transmission signal, and the channel parameter that adopts the least square algorithm to try to achieve is expected matrix and set;

6) right to use matrix vector and channel parameter matrix vector are estimated current channel parameter by transversal filter, and with the result as current estimation channel parameter matrix;

7) with the difference of current channel parameter expectation matrix and current estimation channel parameter matrix as error matrix, revise the weight matrix vector, the amplitude of correction is determined by the algorithm step-length;

8) upgrade the channel parameter matrix vector: with first parameter matrix deletion of channel parameter matrix vector, thereafter the parameter matrix room that previous matrix stays that fills vacancies in order of precedence, and current estimation channel parameter matrix inserted the tail of the queue of channel parameter matrix vector, constitute new channel parameter matrix vector;

9) with the result's output of the estimation channel parameter matrix of current time as channel estimating;

10) use revised weight matrix vector and new channel parameter matrix vector, repeating step 6) 7) 8) 9), carry out next channel parameter estimation constantly.