CN1346187A - Known time delay based channel estimating method - Google Patents

Abstract

The present invention discloses a known time delay based channel estimating method. Under the condition of that the multipath parameter of current time-slot user channel is real-time estimated said method utilizes the conjugate gradient method to make channel estimation of structure matrix of real-time construction, and can obtain the channel parameter model approaching to actuality, at the same time, in the course of obtaining unbiassed channel estimation the massive calculation necessary for structure matrix inversion can be avoided, and the calculation effect obtained by direct inversion can be attained, therefore, at the same time of raising channel estimation accuracy it possesses higher channel estimation efficiency.

Description

A kind of channel estimation methods based on known time delay

The present invention relates to the channel estimation methods in a kind of SCDMA (S-CDMA) system, the channel estimation methods in the especially a kind of SCDMA system based on known time delay.

In existing research to the synchronous CDMA system channel estimating, the optimum and the suboptimum channel estimation methods of joint-detection " in the CDMA mobile communication systems up link based on " (" Optimum andSuboptimum Channel Estimation for Uplink of CDMA Mobile RadioSystem with Joint Detection of the 39th to 50 page be entitled as in european telecommunication and the correlation technique collection of thesis ", ETT, Vol.5, No.1, pp.39-50, jan-feb, 1994) proposed a kind of channel estimation scheme, its channel estimating adopts the method for estimation of the joint-detection of time delay the unknown.Because user's training sequence is derived by a basic cycle sign indicating number and obtained, the feasible channel estimating based on maximum-likelihood criterion of this special make can realize that hardware cost is low by shift register and multiplier.But this method defines the multipath number of estimated subscriber channel and the time-delay in each footpath, and for example, for 16 users' system, estimated channel number is defined as 8 footpaths, and the time-delay in these 8 footpaths is followed successively by 0,1,2,3,4,5, and 6,7 chips.Because the complexity and the unpredictability of practical communication environment, the multipath parameter of the subscriber channel of setting does not promptly very likely meet with actual conditions multipath number and time of delay.If carry out channel estimating according to such scheme, suppose the multipath number of the time delay number of channel greater than real channel, may occur so the situation of noise mistake as user profile; If the time delay number of supposition channel less than the multipath number of real channel, will be lost the information in some footpath so; Because the channel delay of supposition can only be an integer chip, but the real channel time-delay very likely is a non-integer chip, and may be half-chip, obviously supposition channel and truth are inconsistent, in this case, this will inevitably cause the inaccurate of channel estimating model, thereby makes estimated result inaccurate, causes systematic function to descend.

For addressing the above problem, the object of the present invention is to provide and a kind ofly can improve channel estimation accuracy and the higher channel estimation methods of efficient based on known time delay.

For achieving the above object, the technical solution used in the present invention is: a kind of channel estimation methods based on known time delay, and this method may further comprise the steps:

(1) according to the multidiameter delay structural texture matrix G of current time slots;

(2) the following parameter of initialization: autocorrelation matrix R, cross correlation vector b, weighing vector w, error vector g, conjugated gradient direction p, the initial value n=1 of iterations;

(3) set maximum iteration time n _MaxDraw with the error vector range parameter

(4) carry out the following calculating of channel estimating:

1. iteration step length

2. weighing vector w (n)=w (n-1)+a (n) p (n);

3. error vector g (n)=g (n-1)-a (n) R (n) p (n);

4. conjugated gradient direction

5. iterations n=n+1;

(5) if n less than maximum iteration time n given in advance _Max, and error vector ‖ g (n) ‖＞ε, then return above-mentioned steps (4) and proceed iterative computation, otherwise continue following steps;

(6) the weighing vector w (n) that will obtain through the finite iteration of above-mentioned steps (4) exports as channel estimation value, i.e. channel estimating

Wherein: T represents corresponding vector or transpose of a matrix.

The technical scheme that is adopted by the invention described above as can be seen, in order to obtain the multipath model more approaching with truth, to improve accuracy of channel estimation, essence of the present invention is under the known situation of time delay, utilizes conjugate gradient method to carry out channel estimating at the structure matrix of the statement time delay of real-time structure.Because the time delay of channel estimates that higher accuracy can be arranged, the channel parameter model ratio that makes the present invention obtain is supposed the more approaching reality of channel, and therefore channel estimation methods of the present invention can improve accuracy of channel estimation.

Because the present invention carries out channel estimating under the known situation of channel delay, real time delay can only be estimated according to the data in real time that current time slots receives, thereby the also necessary real-time structure of structure matrix G, thereby believe one side only by the nothing of directly asking generalized inverse to obtain to structure matrix G and to estimate that computed offline obtains in advance, can only calculate online.But ask the generalized inverse amount of calculation of higher dimensional matrix very big, the conjugate gradient method of utilizing that the present invention adopts can be avoided above-mentioned matrix inversion, simultaneously can reach the computational accuracy of directly inverting again, therefore the present invention is under the known condition of time delay, utilize conjugate gradient method to carry out channel estimating, when improving channel estimation accuracy, still has higher channel estimating efficient.

Below the present invention is further illustrated.

At first the present invention is described in detail in conjunction with TD SDMA (TD-SCDMA) system.

Supposing has K wireless user in the system, suppose that simultaneously the recovering pulse response of each subscriber channel is: $h^{\left(k\right)}={[h_1^{\left(k\right)},h_2^{\left(k\right)}...h_w^{\left(k\right)}]}^Y,k=1...K,$ Its length is W.K user's channel response is written as vector form:

H=[h ^{(1) T}, h ^{(2) T}H ^{(K) T}] ^T, the number altogether of Unknown Channel coefficient is:

U=KW, k user's training sequence code is: $m^{\left(k\right)}=[m_1^{\left(k\right)},m_2^{\left(k\right)}...m_{L+W-1}^{\left(k\right)}{]}^T,k=1...K.$

Because the length of supposition channel impulse response is W, initial W-1 sampled data in the received signal is subjected to the influence of the time delayed signal of other sequence, and therefore, initial W-1 element of each middle pilot code (midamble sign indicating number) also is not used in channel estimating.By the well-determined received signal of training sequence itself L element only arranged.If corresponding received signal is:

e＝[e ₁，e ₂…e _L] ^T。According to

K=1 ... training sequence m among the K ^(k), the matrix of structure L * W is: $G^{\left(k\right)}=\left(G_{\mathrm{ij}}^{\left(k\right)}\right),k=1...K,$ Wherein: $G_{\mathrm{ij}}^{\left(k\right)}=m_{W+i-j}^{\left(k\right)},i=1...L;j=1...W,$ That is: $G^{\left(k\right)}=\left[\begin{array}{cccc}{m}_W^{\left(k\right)}& m_{W-1}^{\left(k\right)}& ...& m_1^{\left(k\right)}\\ m_{W+1}^{\left(k\right)}& m_W^{\left(k\right)}& ..& m_2^{\left(k\right)}\\ ...& ...& ...& ...\\ m_{W+L-1}^{\left(k\right)}& m_{W+L}^{\left(k\right)}& ...& m_L^{\left(k\right)}\end{array}\right]$ Thereby the structure matrix G of L * U is:

G＝[G ⁽¹⁾，G ⁽²⁾…G ^(K)]。The additivity stationary noise of note zero-mean is:

N=[n ₁, n ₂N _L] ^T, then received signal can be expressed as:

E=Gh+n, thereby the problem of channel estimating in fact just is equivalent to the G among the known formula e=Gh+n, e obtains the process of h.

If use R _n ^-1Expression n=[n ₁, n ₂... n _L] ^TIn noise covariance matrix contrary, then do not have inclined to one side estimated channel estimated matrix to be: $M={\left(G^{*T}{R}_n^{-1}G\right)}^{-1}{G}^{*T}{R}_n^{-1},$ Thereby the maximal possibility estimation of channel h is: $\hat{h}=\mathrm{Me}.$ The covariance matrix of supposing noise satisfies R _n=E{nn ^{* T}}=σ ²I, thereby formula

Be M=(G ^{* T}G) ^-1G ^{* T}Utilize formula M=(G ^{* T}G) ^-1G ^{* T}The channel estimating that obtains $\hat{h}={\left(G^{*T}G\right)}^{-1}{G}^{*T}e$ Be not have partially to estimate, also be least-squares estimation simultaneously.

In case provided matrix G, channel estimating just can through type

Try to achieve.But Jia Ding model is inconsistent with real multipath model usually here, or multipath number does not meet, or deviation is arranged the time of delay of multipath, and this will inevitably reduce channel estimation accuracy, influences the whole receptivity of system.

In order to address this problem, improve accuracy of channel estimation, channel estimating should be carried out under the known situation of time delay.But, in order to obtain the multipath model more approaching with truth, must estimate the multidiameter delay of current time slots in real time, thereby matrix G must construct in real time according to the multidiameter delay of estimating, so, each time slot all must carry out the inversion operation of high level matrix, in order to reduce operand, utilize conjugate gradient method to carry out channel estimating, both can avoid matrix inversion, can reach the effect of directly inverting again simultaneously.Concrete implementation step is as follows:

(1) according to the multidiameter delay structural texture matrix G of current time slots;

(2) the following parameter of initializationization, that is: initialization autocorrelation matrix R=G ^HG, cross correlation vector b=G ^HE, weighing vector w (0)=0, error vector g (0)=b, conjugated gradient direction P (1)=g (0), the initial value n=1 of iterations;

(3) set maximum iteration time n _MaxWith error vector range parameter ε;

(4) carry out the following calculating of channel estimating:

1. iteration step length

2. weighing vector w (n)=w (n-1)+a (n) p (n);

3. error vector g (n)=g (n-1)-a (n) R (n) p (n);

4. conjugated gradient direction

5. iterations n=n+1;

(5) if n less than constant n given in advance _Max, and ‖ g (n) ‖＞ε, then return above-mentioned steps (4) and proceed iterative computation, otherwise continue following steps;

(6) the weighing vector w (n) that will obtain through the finite iteration of above-mentioned steps (4) exports as channel estimation value, i.e. channel estimating

Wherein: H represents the conjugate transpose of corresponding matrix, and e is a received signal, and T represents corresponding vector or transpose of a matrix.

Above described step (1) must at first finish, (2), (3) do not have inevitable ordinal relation.

In order to improve the efficient of said method, the initial value w (0) of described weighing vector can be set to the output of matched filter

, that is:

, the initial value g (0) of error vector multiply by initial value poor of weighing vector w (0) for the initial value of the initial value of cross correlation vector b and cross-correlation matrix R, that is: g (0)=b-Rw (0), and being provided with of all the other initial condition is constant.This kind practice utilizes traditional matched filter to carry out channel estimating earlier, and estimated value is , utilize conjugate gradient algorithms that this is worth then Revise, therefore can accelerate the convergence rate of the interative computation among the present invention, thereby can improve channel estimating efficient of the present invention.

Above described maximum iteration time n _MaxCan set flexibly according to the time length that system allows, specifically can be is arbitrary value between 1 to N, that is: 1≤n _Amx≤ N, wherein N is the dimension of autocorrelation matrix R, error vector range parameter ε should be set at greater than 0, and is far smaller than 1 arbitrary value, that is: 0＜ε＜＜1.

Be applied in the present invention below and further set forth the present invention in the TD-SCDMA system.

According to agreement TS_C102 V3.2.0 version, user's training sequence is that the midamble sign indicating number has 144 chip lengths, and the training sequence code of establishing k user is $m^{\left(k\right)}=[m_1^{\left(k\right)},m_2^{\left(k\right)}...m_{144}^{\left(k\right)}{]}^T,k=1...K$

Suppose each chip time span T _c, the time delay that estimates for 0,0.5T _c, 1T _c, 1.5T _c, 2T _c, 3T _c, the over-sampling rate of receiver section data is 2, promptly the sampling period is from T _cBecome T _c/ 2.In order to realize the time delay of 0.5 chip, we will carry out channel estimating on the basis of over-sampling data.The user's training symbol that is used for channel estimating should be by the well-determined received signal of training sequence itself, therefore, and the preceding WT that the data symbol that is postponed disturbs _cTraining sequence in time will be rejected.Because the total length of user's training sequence is 144 chip lengths, user's training symbol that therefore can be used for channel estimating is a lot, but we can only choose a part wherein, suppose that we carry out channel estimating with aforesaid received signal, from WT _cThe training sequence that constantly begins receives waveform r (t), the sequence e=[e that is used for channel estimating that obtains through as above process sampling ₀, e ₁E _2N-1] ^T, wherein, e _k=r (kT _c/ 2).Matrix G can be constructed as follows:

Then, with following parameter initialization, that is: initialization autocorrelation matrix R=G ^HG, cross correlation vector b=G ^HE, weighing vector w (0)=0, error vector g (0)=b, conjugated gradient direction p (1)=g (0), the initial value n=1 of iterations, and set the required maximum iteration time n of following calculating _MaxWith error vector range parameter ε, then carry out the following calculating of channel estimating:

As the n≤n that satisfies condition _MaxAnd during ‖ g (n) ‖＞ε, calculate:

{ 1. iteration step length

2. weighing vector w (n)=w (n-1)+a (n) p (n);

3. error vector g (n)=g (n-1)-a (n) R (n) p (n);

4. conjugated gradient direction

5. iterations n=n+1; }

The weighing vector w (n) that obtains at last obtaining by above-mentioned finite iteration exports as channel estimation value, i.e. channel estimating

In order to improve channel estimating efficient of the present invention, the w of initial condition (0) also can be set to the output of matched filter

, utilize conjugate gradient algorithms that this is worth then Revise.n _MaxThe length of numerical value time that can allow according to system set flexibly.The dimension of supposing autocorrelation matrix R is N * N dimension, then n _MaxDefinite scope be 1≤n _Max≤ N.

At this moment, the initial value of weighing vector is set to , the initial value of error vector is set to g (0)=b-Rw (0), and being provided with of all the other initial condition is constant.

The present invention goes for various mobile communication system, and provides accuracy higher channel estimating.

Claims (4)

1, a kind of channel estimation methods based on known time delay is characterized in that, this method may further comprise the steps:

(1) according to the multidiameter delay structural texture matrix G of current time slots;

(2) the following parameter of initialization: autocorrelation matrix R, cross correlation vector b, weighing vector w, error vector g, conjugated gradient direction p, the initial value n=1 of iterations;

(3) set maximum iteration time n _MaxWith error vector range parameter ε;

(4) carry out the following calculating of channel estimating:

1. iteration step length

2. weighing vector w (n)=w (n-1)+a (n) p (n);

3. error vector g (n)=g (n-1)-a (n) R (n) p (n);

4. conjugated gradient direction

5. iterations n=n+1;

(5) if n less than maximum iteration time n given in advance _Max, and error vector ‖ g (n) ‖＞ε, then return above-mentioned steps (4) and proceed iterative computation, otherwise continue following steps:

(6) the weighing vector w (n) that will obtain through the finite iteration of above-mentioned steps (4) exports as channel estimation value, i.e. channel estimating

Wherein: T represents corresponding vector or transpose of a matrix.

2, the channel estimation methods based on known time delay according to claim 1 is characterized in that: the following parameter of described initialization is that parameter autocorrelation matrix R is initialized as R=G ^HG, cross correlation vector b are initialized as b=G ^HE, weighing vector w be initialized as w (0)= , error vector g is initialized as g (0)=b-Rw (0), conjugated gradient direction p and is initialized as p (1)=g (0);

Wherein: H represents the conjugate transpose of corresponding matrix,

Be the output of matched filter, e is a received signal.

3, the channel estimation methods based on known time delay according to claim 1 is characterized in that, the following parameter of described initialization is that parameter autocorrelation matrix R is initialized as R=G ^HG, cross correlation vector b are initialized as: b=G ^HE, weighing vector w are initialized as w (0)=0, error vector g and are initialized as g (0)=b, conjugated gradient direction p and are initialized as p (1)=g (0);

Wherein: sheet is represented the conjugate transpose of corresponding matrix, and e is a received signal.

4, the channel estimation methods based on known time delay according to claim 1 is characterized in that: described maximum iteration time n _MaxBe the arbitrary value between 1 to N, that is: 1≤n _Max≤ N, wherein N is the dimension of autocorrelation matrix R; Described error vector range parameter ε is greater than 0, and is far smaller than 1 arbitrary value, that is, and and 0＜ε＜＜1.