CN1722623A - A method for fast channel estimation by means of training sequence - Google Patents

Abstract

This invention discloses a method for rapidly estimating communication channel with train sequence, which comprises the followings: a) making sure the widow length of the communication channel widow according to the base station; separating the train sequence receiving signal from the baseband receiving signal, and forming transmitting matrix according to all train sequence; b) cutting data from said receiving signal to form new data sequence, and cutting data from said transmitting matrix to form new matrix; c) deciding whether the widow length is larger than 16 cord sheet, substituting the element whose value is zero with the train sequence value of the corresponding user, making them meeting right circularity; c) operating Fourier transformation and inverse transformation, computing and separating the communication channel estimated value of each user. Said invention can be used when the widow length is larger or smaller than 16 cord sheet.

Description

A kind of method of utilizing training sequence to carry out channel estimating fast

Technical field

The present invention relates to the digital mobile communication field, adopt training sequence to carry out the method for channel estimating when being particularly related in the TD-CDMA system, relate in particular to grow up Fast Channel method of estimation when 16 chips of the user distributes in the TDS-CDMA system channel window.

Background technology

In existing TD-CDMA (TD-CDMA Time Division-Code Division Multiple Access) system and TD-SCDMA (Time Division-Synchronous Code Division Multiple Access) system, all users' of same sub-district training sequence is all basic by one, the training sequence in cycle (cycle is P) produces by a cyclic shift, and corresponding channel estimation methods generally all is a utilization Steiner method of estimation [document 1], and it is a kind of low-cost channel estimation methods.Yet the performance of Steiner estimator is subjected to the influence of receiving terminal additive noise, and its channel response that obtains responds with ideal communication channel and compares, and comprises noise component(s), and this is channel estimation errors just; Can enlarge noise power simultaneously, cause the relative input of output signal to noise ratio that a certain amount of loss is arranged.In the prior art [document 2], proposed a kind of noise reduction post-processing approach of crossing Threshold detection, its basic thought is to estimate original channel impulse response with the Steiner method of estimation earlier, uses a thresholding then, removes most of noise tap.Have very little amount of calculation in this way, and can make system obtain certain performance improvement.

Integrate, existing TD-SCDMA system is by the training sequence of designing user reasonably, as long as the given long W of window of each user is less than or equal to 16 chips, then use the Steiner method of estimation and just the linear convolution of complexity can be converted into simple circular convolution, can realize by FFT and IFFT computing, greatly reduce the algorithm complex of channel estimating.

Following document has been introduced existing channel estimation technique:

Document [1] STEINER B, BAIER P.Low cost channel estimation in theuplink receiver of CDMA mobile radio systems, Frequenz, 1993,47 (12): 292-298.

Document [2] Kang Shaoli, the improvement of low-cost channel estimation methods in the .TD-SCDMA systems such as Qiu Zhengding. communication journal, 2002,23 (10): 125-130.

Yet also obviously there is a defective in the producing method of above-mentioned training sequence: when the maximum delay expansion of the real multipath channel of certain user surpasses the given long W of window, this user's training sequence is through in the channel estimation window that can fall into a user thereafter after the multipath channel, thereby will cause the channel estimation error of two aspects: this can not be estimated (1) with all time delays expansions of corpse channel impulse response above W; (2) all time delay expansions of this user will show in next user's channel impulse window above the channel impulse response of W, thereby cause next user's the complete mistake of channel impulse possibility.Because this former thereby channel estimating that causes produces wrong and makes the demodulation performance rapid deterioration.This situation is called as channel estimating " window is fuzzy ".

Solve that above-mentioned " a direct and effective way of window fuzzy problem is that the window length of distributing to the user is long as much as possible, and this is fully possible the user more after a little while.For example: in the TD-SCDMA system, P=128,

Symbol Expression is less than or equal to the maximum integer of X, and K is for using user's number of same basic training sequences (Midamble sign indicating number); When K=8, W=16; During K=4, W=32; During K=2, W=64 etc.Yet, according to the Steiner method of estimation, use FFT and IFFT (Fourier transform and inverse transformation) computing to carry out channel estimating, require W to be less than or equal to 16 chips.

Summary of the invention

Technical problem to be solved by this invention provides a kind of method of utilizing training sequence to carry out channel estimating fast, solve prior art because the expansion of the maximum delay of the real multipath channel of user is long above given window, and the channel estimation error that produces and the problem of demodulation performance rapid deterioration.

For achieving the above object, the invention provides a kind of method of utilizing training sequence to carry out channel estimating fast, its characteristics are, comprise the steps: step 1, determine that according to the base station window of channel estimation window is long, from baseband receiving signals, isolate the training sequence received signal, and constitute transmission matrix according to all users' training sequence; Step 2, data intercept constitutes new data sequence from described received signal, and data intercept constitutes new matrix from described transmission matrix; Judging that described window is long whether greater than 16 chips, is execution in step three then, otherwise will described new matrix as newly transmission matrix and execution in step four; Step 3, to be 0 element to the value in the described new matrix replace with the training sequence value of relative users, and make the new matrix after the replacement satisfy right cycle characteristics, and with the new matrix after the described replacement as new transmission matrix; Step 4, carry out Fourier transform and inverse transformation, calculate all users' channel estimation value, and isolate each user's channel estimation value according to described all users' channel estimation value by described new data sequence and described new transmission matrix; Step 5, according to described channel estimation value, carry out channel estimating and processing, filtered noise tap, the new channel estimation value that obtains handling, and the new channel estimation value of described processing sent into follow-up data demodulation module.

The above-mentioned training sequence that utilizes carries out the method for channel estimating fast, and its characteristics are, in described step 2, are the 17th row beginning data intercepts from described received signal, until the 144th row, receive data by 128 row that intercepted and constitute described new data sequence; The 17th row from described transmission matrix begins intercepting accordingly, until the 144th row, and keeps all columns certificates, constitutes described new transmission matrix by 128 * 128 line data that intercepted.

The above-mentioned method of utilizing training sequence to carry out channel estimating fast, its characteristics be, in described step 4, and all users' channel estimation value $\hat{h}=\mathrm{IFFT}\[\frac{\mathrm{FFT}\left(e^{\′}\right)}{\mathrm{FFT}\left(G^{\′}(1,:)\right)}\];$ FFT, IFFT are Fourier transform and inverse transformation in the formula, and G ' (1 :) be the first columns certificate of described new transmission matrix, e ' is described new data sequence.

The above-mentioned method of utilizing training sequence to carry out channel estimating fast, its characteristics be, in described step 4, and the channel estimation value that estimates Isolate each user's channel estimation value

And it is concrete satisfied ${\hat{h}}_j^{\left(k\right)}={\hat{h}}_{j+(k-1)w}(j=\mathrm{1,2},\·\·\·,W;k=\mathrm{1,2},\·\·\·,k)$ Condition, wherein W is that described window is long, K is the maximum activation number of users of base station configuration.

The above-mentioned training sequence that utilizes carries out the method for channel estimating fast, and its characteristics are that described user is the user in the synchronization code multi-address division system, and the long W of described window is the maximum integer that is less than or equal to 128/K.

Technique effect of the present invention is:

The method of utilizing training sequence to carry out channel estimating fast of the present invention, channel estimation methods for a kind of any W length (can greater than, be equal to or less than 16 chips), can estimate the channel impulse response of each user in the TD-SCDMA system on the one hand accurately and effectively, solve " window fuzzy problem; the while can use FFT again and IFFT carries out quick computing; greatly reduced computation complexity, and finally improved the receptivity of uplink receiver.

Description of drawings

Fig. 1 is the FB(flow block) that channel estimation methods of the present invention is realized.

Embodiment

Further describe specific embodiments of the invention below in conjunction with accompanying drawing.

The channel impulse response vector of supposing different user (window long for W) is h (column vector of K * W dimension, K is the maximum activation number of users of base station configuration), and the transmission matrix that is made of the training sequence or the user data of different user is defined as G and (is (L _m+ W-1) * and the matrix of KW, in the TD-SCDMA system, L _m=144), according to transmission theory as can be known, receive data e=Gh+n (n is a noise matrix) by the training sequence that measures.Basic thought of the present invention is that any P among the intercepting transmission matrix G is (in the TD-SCDMA system, P=128) data of row constitute new matrix G ', 0 element among the matrix G ' is replaced with the training sequence of relative users, and the criterion of replacement is the new matrix that makes after the replacement G' satisfying right cycle characteristics, the P line data among the corresponding intercepting e is e', by e', G' obtain the estimated value of channel by the FFT/IFFT transformation calculations $\hat{h}=\mathrm{IFFT}\[\frac{\mathrm{FFT}\left(e^{\′}\right)}{\mathrm{FFT}\left(G^{\′}(1,:)\right)}\],$ Wherein FFT, IFFT are Fourier transform and inverse transformation, G' (1 :) be G' the first columns certificate.

The core algorithm of channel estimation methods of the present invention is: utilize the FFT conversion that the convolution algorithm in the time domain is transformed into and carry out product calculation in the frequency domain.

With reference to figure 1, be the FB(flow block) that channel estimation methods of the present invention is realized, detailed step is as follows:

Step 100, according to the long W of window that the base station is determined, (line number is 1, and columns is L to isolate training sequence received signal e from baseband receiving signals _m+ W-1)); Training sequence according to all users constitutes transmission matrix simultaneously, and (line number is (L to be defined as G _m+ W-1), columns is KW), detailed process is referring to list of references [1].

Step 200 is carried out channel estimating, estimates all active paths of all users; This step also comprises (a) and (b), (c), (d) four little steps;

According to the length of W, divide dual mode to handle: when W≤16, directly enter (a), the multipath data that can avoid data division fully are to the influence of the training sequence that receives, and channel estimating will not have error (if not considering noise effect).When W＞16, directly enter (b), channel estimation errors minimum (if not considering noise effect), lot of experiment validation, this error is insignificant.

(a) the 17th row from received signal e begins data intercept, until L _m=144 row receive data by this 128 row and constitute new data sequence e '.The 17th row from the G matrix begins intercepting accordingly, until L _m=144 row (all columns are according to remaining unchanged) constitute new transmission matrix G by this 128 * 128 line data, enter (c) then.

(b) the 17th row from received signal e begins data intercept, until L _m=144 row receive data by this 128 row and constitute new data sequence e '.The 17th row from the G matrix begins intercepting accordingly, until L _m=144 row (all columns are according to remaining unchanged), simultaneously this 128 * 128 line data is done following processing: being 0 element to its intermediate value promptly replaces with the training sequence value of relative users, the criterion of replacing is to make the new matrix G ' after the replacement satisfy right cycle characteristics, constitute new transmission matrix G ', enter (c) then.

(c) by e', G' obtain the estimated value of all users' channel by the FFT/IFFT transformation calculations

$\hat{h}=\mathrm{IFFT}\[\frac{\mathrm{FFT}\left(e^{\′}\right)}{\mathrm{FFT}\left(G^{\′}(1,:)\right)}\]$ (wherein FFT, IFFT are Fourier transform and inverse transformation, G ' (1 :) be first columns certificate of G '), enter (d) then.

(d) according to the channel estimation value that estimates ${\hat{h}}_i(i=\mathrm{1,2},\·\·\·,128),$ Isolate each user's channel estimation value Concrete satisfied: ${\hat{h}}_j^{\left(k\right)}={\hat{h}}_{j+(k-1)W}(j=\mathrm{1,2},\·\·\·,W;k=\mathrm{1,2},\·\·\·,K),$ Entered for the 3rd step then.

Step 300, the channel estimating original value according to step 200 obtains carries out the channel estimation value reprocessing according to the method that document [2] is introduced, filter most of noise tap, the new channel estimation value that obtains handling is sent into follow-up data demodulation module, and so far channel estimation works finishes fully.The step 100 of device of the present invention, step 300 can list of references [1] or [2], and this place emphasis is discussed the feasibility that step 200 of the present invention realizes.Short and sweet for what narrate, suppose that current excited user number is K, according to each user's of agreement window length be

For the theoretical derivation of W≤16, can be referring to document [1] and [2], the main situation of discussing W＞16 in this place, but for the situation of W≤16, also be one of protection of the present invention.

For K user, do not consider the influence of data division, then training sequence received signal part can be write as (specifically can referring to document [1] or document [2]): e=GH+n wherein: $e={\&lsqb;{\mathrm{<figure-callout\; id="1"\; label="e"\; filenames="A20041000932600111.png"\; state="\{\{state\}\}">e</figure-callout>}}_1,e_2,\&CenterDot;\&CenterDot;\&CenterDot;,e_{L_m+W-1}\&rsqb;}^T,$ H=[(h ⁽¹⁾) ^T, (h ⁽²⁾) ^T..., (h ^{(K) T}] ^T, h ^(k)=[h ₁ ^(k), h ₂ ^(k)..., h _W ^(k)] ^T, T representing matrix or vector change order; And transmission channel G is schematically as follows:

$G=\left(\begin{array}{ccccccccc}{m}_1\left(1\right)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0& \&CenterDot;\&CenterDot;\&CenterDot;& m_k\left(1\right)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ & & & & & & & & \\ m_2\left(2\right)& m_2\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;& m_k\left(2\right)& m_k\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& 0& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ & & & & & & & & \\ m_1\left(W\right)& m_1(W-1)& \&CenterDot;\&CenterDot;\&CenterDot;& m_1\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& m_k\left(W\right)& m_k(W-1)& \&CenterDot;\&CenterDot;\&CenterDot;& m_k\left(1\right)\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ m_1\left(L_m\right)& m_1(L_m-1)& \&CenterDot;\&CenterDot;\&CenterDot;& m_1(L_m-W+1)& \&CenterDot;\&CenterDot;\&CenterDot;& m_k\left(L_m\right)& m_k(L_m-1)& \&CenterDot;\&CenterDot;\&CenterDot;& m_k(L_m-W+1)\\ 0& m_1\left(L_m\right)& \&CenterDot;\&CenterDot;\&CenterDot;& m_1(L_m-W+2)& \&CenterDot;\&CenterDot;\&CenterDot;& 0& m_k\left(L_m\right)& \&CenterDot;\&CenterDot;\&CenterDot;& m_k(L_m-W+2)\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& \&CenterDot;\&CenterDot;\&CenterDot;& m_1\left(L_m\right)& \&CenterDot;\&CenterDot;\&CenterDot;& 0& 0& \&CenterDot;\&CenterDot;\&CenterDot;& m_k\left(L_m\right)\end{array}\right)$

For e and G, we until the 144th row, constitute new e ' and G ' since the 17th row data intercept, and then this moment, the received signal equation was:

E '=G ' H+n is wherein: $e={\&lsqb;e_{17},e_{18},\&CenterDot;\&CenterDot;\&CenterDot;{,e}_{L_m}\&rsqb;}^T,$ And transmission channel G ' is schematically as follows:

$G^{\&prime;}=\left[\begin{array}{ccccccccccc}{m}_1\left(17\right)& m_1\left(16\right)& \&CenterDot;\&CenterDot;\&CenterDot;& m_1\left(1\right)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0& m_2\left(17\right)& \&CenterDot;\&CenterDot;\&CenterDot;& m_x\left(17\right)& \&CenterDot;\&CenterDot;\&CenterDot;\\ & & \&CenterDot;& \&CenterDot;& & \&CenterDot;& & & \&CenterDot;& & \&CenterDot;\\ m_1\left(18\right)& m_1\left(17\right)& \&CenterDot;& \&CenterDot;& m_1\left(1\right)& \&CenterDot;& 0& m_2\left(18\right)& \&CenterDot;& m_x\left(18\right)& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ & & \&CenterDot;& \&CenterDot;& & \&CenterDot;& & & \&CenterDot;& & \&CenterDot;\\ m_1\left(W\right)& m_1(W-1)& \&CenterDot;& m_1(W-16)& m_1(W-17)& \&CenterDot;& m_{[}\left(1\right)& m_2\left(W\right)& \&CenterDot;& m_x\left(W\right)& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ m_1\left(L_m\right)& m_1(L_m-1)& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;& \&CenterDot;& m_1(L_m-W+1)& m_2\left(L_m\right)& \&CenterDot;& m_x\left(L_m\right)& \&CenterDot;\&CenterDot;\&CenterDot;\end{array}\right]$

Observe transmission channel G ', obviously G ' does not satisfy right cycle characteristics, therefore can not directly use FFT/IFFT to come the Fast estimation channel impulse response.

But observe again simultaneously; G ' intermediate value is that 0 element is still quite few; in actual channel; W also needn't be provided with too high; generally get W=32 be sufficient for sb.'s need (for W＞32 just error increase, do not influence the protection range of the inventive method), at this moment can replace with the training sequence value of relative users by being 0 element to value; the criterion of replacing is to make the new matrix G ' after the replacement satisfy right cycle characteristics, constitutes new transmission matrix G '.In practical operation, be actually and do not spend filling, as long as think populated,, be FFT and get final product because we when using FFT/IFFT to come the Fast estimation channel impulse response, only need use the data of first row of G '.

We just can use FFT/IFFT and came the Fast estimation channel impulse response this moment, embodied as follows:

$H=\mathrm{IFFT}\&lsqb;\frac{\mathrm{FFT}\left(e^{\&prime;}\right)}{\mathrm{FFT}\left(G^{\&prime;}(1,:)\right)}\&rsqb;\&CenterDot;\&CenterDot;\&CenterDot;\left(5\right)$

Wherein: FFT, IFFT are Fourier transform and inverse transformation, and G ' (1 :) be the first columns certificate of G '.Only note simultaneously $\frac{\mathrm{FFT}\left(e^{\&prime;}\right)}{\mathrm{FFT}\left(G^{\&prime;}(1,:)\right)}$ Be molecule with denominator in direct corresponding being divided by of element of sequence.

Existing a large amount of experiment confirms: (1) when the window length of the wireless channel environment of user's reality less than at 16 chips the time, adopt channel estimation methods disclosed by the invention, channel estimating will not have error (if not considering noise effect), and its performance is consistent with document [1] and the disclosed performance of document [2]; (2) when the window of the wireless channel environment of user's reality is grown up in 16 chips, channel estimation errors minimum (if not considering noise effect), lot of experiment validation, this error is insignificant, and adopt document [1] and document [2] disclosed method will produce influence greatly, cause the error rate to gather increase, do not increase any computation complexity simultaneously.The validity and the feasibility of the open method of the present invention have been confirmed by existing a large amount of experiments.

Compared with prior art, the invention provides a kind of method of utilizing training sequence to carry out channel estimating fast, can carry out channel estimating fast under the long W of window arbitrarily, when its performance was less than or equal to 16 chips at W, channel estimating did not have error (if not considering noise effect); At W during greater than 16 chips, channel estimation errors minimum (if not considering noise effect), lot of experiment validation, this error is insignificant.

The present invention is applicable to and adopts training sequence to carry out all systems of channel estimating among the TDMA/CDMA, especially is applicable to WCDMA TDD and TD-SCDMA system in the present 3-G (Generation Three mobile communication system).But be understood that, though technical scheme of the present invention is primarily aimed at the wireless communication system of code division multiple access, but be applicable to the frequency division multiple access and the time-division multiple address system that adopt similar transmission structure too, any have a signal processing, the engineer of knowledge background such as communication, can design corresponding channel estimating apparatus according to the present invention, it all should be included in thought of the present invention and the scope.Therefore, the above is preferred embodiment of the present invention only, is not to be used for limiting practical range of the present invention; Every according to equivalence variation and modification that the present invention did, all contained by claim of the present invention.

Claims (5)

1, a kind of method of utilizing training sequence to carry out channel estimating fast is characterized in that, comprises the steps:

Step 1, determine that according to the base station window of channel estimation window is long, from baseband receiving signals, isolate the training sequence received signal, and constitute transmission matrix according to all users' training sequence;

Step 2, data intercept constitutes new data sequence from described received signal, and data intercept constitutes new matrix from described transmission matrix; Judging that described window is long whether greater than 16 chips, is execution in step three then, otherwise will described new matrix as newly transmission matrix and execution in step four;

Step 3, to be 0 element to the value in the described new matrix replace with the training sequence value of relative users, and make the new matrix after the replacement satisfy right cycle characteristics, and with the new matrix after the described replacement as new transmission matrix;

Step 4, carry out Fourier transform and inverse transformation, calculate all users' channel estimation value, and isolate each user's channel estimation value according to described all users' channel estimation value by described new data sequence and described new transmission matrix:

Step 5, according to described channel estimation value, carry out channel estimating and processing, filtered noise tap, the new channel estimation value that obtains handling, and the new channel estimation value of described processing sent into follow-up data demodulation module.

2, the method for utilizing training sequence to carry out channel estimating fast according to claim 1, it is characterized in that, in described step 2, be the 17th row beginning data intercept from described received signal, until the 144th row, receive data by 128 row that intercepted and constitute described new data sequence; The 17th row from described transmission matrix begins intercepting accordingly, until the 144th row, and keeps all columns certificates, constitutes described new transmission matrix by 128 * 128 line data that intercepted.

3, the method for utilizing training sequence to carry out channel estimating fast according to claim 2 is characterized in that, in described step 4, and all users' channel estimation value $\hat{h}=\mathrm{IFFT}\left[\frac{\mathrm{FFT}\left(e^{\&prime;}\right)}{\mathrm{FFT}\left(G^{\&prime;}(1,:)\right)}\right];$ FFT, IFFT are Fourier transform and inverse transformation in the formula, and G ' (1 :) be the first columns certificate of described new transmission matrix, e ' is described new data sequence.

4, the method for utilizing training sequence to carry out channel estimating fast according to claim 3 is characterized in that, in described step 4, and the channel estimation value that estimates ${\hat{h}}_i(i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,128)$ , isolate each user's channel estimation value , and concrete satisfied ${\hat{h}}_j^{\left(k\right)}={\hat{h}}_{j+(k-1)W}(j=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,W;k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,K)$ Condition, wherein W is that described window is long, K is the maximum activation number of users of base station configuration.

5, the method for utilizing training sequence to carry out channel estimating fast according to claim 4 is characterized in that, described user is the user in the synchronization code multi-address division system, and the long W of described window is the maximum integer that is less than or equal to 128/K.

Images