CN102271103B - Multi-cell channel estimation method for time division-synchronous code division multiple access (TD-SCDMA) system - Google Patents

Abstract

The invention provides a multi-cell channel estimation method for a time division-synchronous code division multiple access (TD-SCDMA) system. The method specifically comprises the following steps of: performing rough channel estimation on each cell by adopting single-cell Steiner channel estimation; determining the total number of access users according to a rough channel estimation result; judging whether the total number of the access users meets multi-cell joint channel estimation conditions, performing multi-cell joint channel estimation on the access users of each cell of the total number of the access users meets the multi-cell joint channel estimation, otherwise performing multi-cell serial interference cancellation channel estimation on the access users of each cell; and performing de-noising processing on a multi-cell joint channel estimation result or a multi-cell serial interference cancellation channel estimation result. By the method, operational complexity can be greatly reduced and the needs of the access users in any number can be satisfied under the condition of ensuring high channel estimation accuracy, thereby better solving the problem of severe inter-channel interference of the TD-SCDMA system on cell edges and providing quality assurance for multi-cell joint detection.

Description

Multi-cell channel estimation method in a kind of TD-SCDMA system

Technical field

The present invention relates to mobile communication technology field, particularly relate to the multi-cell channel estimation method in a kind of TD-SCDMA system.

Background technology

TD-SCDMA(TD SDMA, Time Division-Synchronous Code Division Multiple Access) system is to be organized by CWTS(Chinese Industrial Standards (CIS), China Wireless Telecommunications Standards group) propose, one of and the 3G (Third Generation) Moblie standard of being accepted by ITU (International Telecommunication Union, International Telecommunications Union).Wherein, joint-detection is one of key technology being applied to TD-SCDMA system, and the performance of this technology depends on channel estimating fast and accurately.

TD-SCDMA system both can inter-frequency networking also can identical networking, wherein, the advantage of identical networking is that all communities all may be used all possible frequency resource, there is the higher availability of frequency spectrum, but it brings an inevitable problem, is also, adjacent cell subscriber signal near Target cell will bring serious co-channel interference, and therefore multi-cell joint detection seems especially necessary.A kind of the most direct method is the Steiner(stainer that adopts expansion) channel estimating, obtain respectively the CIR(channel impulse response of each community interference user by changing basic Midamble code (training sequence of TD-SCDMA system), Channel impulse response).But, because the Midamble matrix of each community is also non-orthogonal, therefore carry out in the process of individual channel estimation in Dui Ge community respectively, the access user signal of other community will bring the interference of can not ignore, this just makes the Steiner precision of channel estimation based on ZF principle greatly reduce, and the decline of precision of channel estimation will have a strong impact on the performance of joint-detection.

In order to improve multi-cell channel estimated accuracy, existing a kind of multiple cell combined channel estimation method, by changing the structure of channel estimate matrix, adjacent cell interference user is also processed as the subscriber signal that can detect, can reduce the noise power in channel estimating.

But above-mentioned multiple cell combined channel estimation method exists following technological difficulties in actual applications:

The one, computation complexity is high; Huge operand has restricted channel estimating speed, thereby causes the significantly minimizing of vacant time of joint-detection;

The 2nd, condition restriction harshness; Identical with tradition list community Steiner channel estimating, there is restriction to access customer number that can be satisfied at most in multi-plot joint channel estimating, for up link, this restriction is easy to be exceeded, if and the access user that exceeds process range is processed as white noise, can greatly reduce precision of channel estimation.

In a word, need the urgent technical problem solving of those skilled in the art to be exactly: how can be with lower computational complexity, solve low-rate users in many speed TD-SCDMA system and be subject to the problem of high rate user serious interference.

Summary of the invention

Technical problem to be solved by this invention is to provide the multi-cell channel estimation method in a kind of TD-SCDMA system, it can ensure under the high-precision condition of channel estimating, computational complexity can be greatly reduced and any access customer number object demand can be met again, thereby solve better TD-SCDMA system in the serious problem of cell edge co-channel interference, for multi-cell joint detection provides quality assurance.

In order to address the above problem, the invention discloses the multi-cell channel estimation method in a kind of TD-SCDMA system, comprising:

Adopt single community Steiner channel estimating, channel rough estimate is carried out in Dui Ge community;

According to channel rough estimate result, determine access user sum;

Judge whether described access user sum meets multi-plot joint channel estimating condition, described multi-plot joint channel estimating condition is, described access user sum is no more than P/W, and wherein, P is training length, the length of window that W is channel impulse response; If so, carry out multi-plot joint channel estimating for the access user of each community, otherwise, carry out many communities counteracting serial interference channel estimating for the access user of each community;

For the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, carry out Denoising disposal.

Preferably, described according to channel rough estimate result, determine the step of access user sum, comprising:

For the channel rough estimate result Zhong Ge each user's in community channel impulse response, the default thresholding of its power and community, place is compared;

If exceed the default thresholding of community, described place, this user is saved to access user set, and records this user place cell id and residing channel impulse response the window's position;

If lower than the default thresholding of community, described place, using this user as non-access user, and the full zero setting of channel impulse response tap to this user;

After comparison, using the number of users in described access user set as access user sum.

Preferably, the default thresholding of community, described place is β times of measurement noise power, wherein, determines β initial value in the time of networking or cell initial, or, in estimating, multi-cell channel dynamically adjusts described β initial value according to dynamic channel environment.

Preferably, the described access user for each community carries out the step of multi-plot joint channel estimating, comprising:

Pilot reception signal according to the each access user in each community is carried out matched filtering, obtains matched filtering Output rusults;

The correlation matrix that the pilot code of the each access user in structure You Ge community forms;

Calculate the Cholesky factor of this correlation matrix;

According to the described Cholesky factor, described matched filtering Output rusults is carried out to forward direction iteration one time, and backward iteration is carried out once in output to forward direction iteration, the channel impulse response family that obtains the each access user in each community is estimated.

Preferably, the described access user for each community carries out the step of many communities counteracting serial interference channel estimating, comprising:

Initialization iterations;

According to current channel estimation results, calculate the channel impulse response power of the current access user in each community, and select to have the access user of maximum channel impulse response power;

This access user is disturbed to recovery;

From current channel estimation results, balance out the interference of this user to other user;

Iterations adds 1;

Judge whether this iterations reaches preset value, if so, stop iteration, and using current channel estimation results as many communities counteracting serial interference channel estimation results, otherwise, the new channel impulse response power that calculates the current access user in each community returned to.

Preferably, employing FFT carries out the constructor of described matched filtering operation and correlation matrix, and, adopt Schur algorithm to calculate the Cholesky factor of this correlation matrix.

Preferably, before the access user for each community carries out multi-plot joint channel estimating, also comprise:

Judge the relation of current time slots access user and a upper time slot access user, if this relation is not that many new access users, a few former access user or access user are constant, carry out multi-plot joint channel estimating for the access user of each community.

Preferably, described method also comprises:

Because of the period of the day from 11 p.m. to 1 a.m, store this Cholesky factor at the Cholesky that calculates this correlation matrix;

If this relation is that a few former access user or access user are constant, the Cholesky factor of an above time slot is as the Cholesky factor of current time slots;

If this relation is many new access users, obtain the Cholesky factor of current time slots according to the Cholesky factor iteration of a upper time slot;

And, according to the Cholesky factor of current time slots, current matching filtering Output rusults is carried out to forward direction iteration one time, and backward iteration is carried out once in output to forward direction iteration, the channel impulse response family that obtains the each access user in each community is estimated;

Preferably, described Denoising disposal step comprises:

For the channel impulse response tap of access user in the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, if its mould value is less than the second default thresholding, by its zero setting.

Preferably, the α that the described second default thresholding is noise variance doubly.

Preferably, the single community of described employing Steiner channel estimating, the step that channel rough estimate is carried out in Dui Ge community comprises:

Pilot reception signal for current area is carried out FFT computing;

Use the FFT of the basic Midamble code in FFT operation result Dian Chuge community;

A division result is carried out to IFFT computing, obtain the channel rough estimate result of current area.

Preferably, obtain the basic Midamble code of adjacent cell by following steps:

The basic Midamble code of adjacent cell is transferred to base station by radio network controller;

Base station is transferred to terminal by signaling by the basic Midamble code of adjacent cell.

Compared with prior art, the present invention has the following advantages:

First, Dui Ge of the present invention carries out community after channel rough estimate, in the time that access user sum meets multi-plot joint channel estimating condition, just carry out many communities counteracting serial interference channel estimating, and in the time that access user sum does not meet multi-plot joint channel estimating condition, carry out many communities counteracting serial interference channel estimating; Therefore, can break away from the restriction of multi-plot joint channel estimating condition, meet any access customer number object demand;

Secondly, because multi-plot joint channel estimating is treated as desired signal processing adjacent cell signal, can greatly improve accuracy of estimation, Er Duo community counteracting serial interference channel estimating can be disturbed elimination to channel rough estimate result, obtains more accurate CIR; Therefore, the present invention can ensure the precision that multi-cell channel is estimated;

Moreover the operation of inverting that the present invention can adopt the Cholesky of correlation matrix to decompose alternate channel estimated matrix G, to reduce the computational complexity of multi-plot joint channel estimating;

In addition the main computing that, many communities of the present invention counteracting serial interference channel estimating relates to is the IDFT of a division result; Due to because the DFT of the basic Midamble code in community tries to achieve in channel rough estimate, or in the time of cell initial, store and all come again when channel estimating at every turn, so operand not quite; In addition, the mutual interference matrix element that belongs to each user of Yi Ge community together comes from same vector, if access community number is N, only needs

inferior IDFT operation; Therefore, many communities of the present invention counteracting serial interference channel estimating has low complexity;

Further, the present invention can also adopt FFT computing, and the matched filtering operation in simplification multi-plot joint channel estimating and the constructor of correlation matrix, to reduce operand;

Further, the Cholesky that the present invention can also adopt Schur algorithm to carry out correlation matrix decomposes, and reduces computation complexity;

In addition, in the process of multi-plot joint channel estimating, the present invention can also be by judging the relation of current time slots access user and a upper time slot access user, to the access user processing of classifying, to reduce operand;

Finally, if this relation meets the condition of access user many 1, the old Cholesky factor of available iteration calculates the new Cholesky factor, like this, with respect to again adopting Schur algorithm to calculate the new Cholesky factor, can greatly reduce amount of calculation.

Brief description of the drawings

Fig. 1 is the flow chart of the multi-cell channel estimation method embodiment 1 in a kind of TD-SCDMA system of the present invention;

Fig. 2 is the implementation procedure signal of a kind of step 105 of the present invention;

Fig. 3 is the flow chart of the multi-cell channel estimation method embodiment 2 in a kind of TD-SCDMA system of the present invention;

Fig. 4 is the implementation procedure example that a kind of access user for each community of the present invention carries out multi-plot joint channel estimating.

Embodiment

For above-mentioned purpose of the present invention, feature and advantage can be become apparent more, below in conjunction with the drawings and specific embodiments, the present invention is further detailed explanation.

With reference to Fig. 1, show the flow chart of the multi-cell channel estimation method embodiment 1 in a kind of TD-SCDMA system of the present invention, specifically can comprise:

Step 101, adopt single community Steiner channel estimating, channel rough estimate is carried out in Dui Ge community;

In a preferred embodiment of the present invention, the step that described Dui Ge carries out community channel rough estimate can specifically comprise:

Sub-step A1, carry out FFT(fast Fourier transform for the pilot reception signal of current area, fast Fourier transform) computing;

The FFT of sub-step A2, the basic Midamble code in use FFT operation result Dian Chuge community;

Sub-step A3, a division result is carried out to IFFT(inverse fast Fourier transform, Inverse fast Fourier transform) computing, obtain the channel rough estimate result of current area.

For Target cell, can obtain by following steps the basic Midamble code of adjacent cell: RNC (radio network controller, Radio Network Controller) the basic Midamble code of adjacent cell is transferred to base station; If use terminal, also can be transferred to terminal by signaling by the basic Midamble code of adjacent cell by base station.

Can obtain the rough estimate of the each user CIR of each community under full load (the Midamble code of community is used entirely) by step 101, and user's numerical value in its corresponding CIR window of access is not interference and the white noise of other community access user entirely.

Step 102, according to channel rough estimate result, determine access user sum;

In a preferred embodiment of the present invention, can determine as follows access user sum:

Sub-step B1, for the channel rough estimate result Zhong Ge each user's in community channel impulse response, the default thresholding of its power and community, place is compared;

Because the channel circumstance of different districts is not quite similar, therefore the present invention can be respectively for cell setting default thresholding separately, like this, the default thresholding of community, place can be determined and be passed through repeatedly to adjust and find Best Point by this cell channel environment.

In practice, the β that the default thresholding of community, described place can be set to measure noise power doubly, wherein, determines β initial value in the time of networking or cell initial, or, in estimating, multi-cell channel dynamically adjusts described β initial value according to dynamic channel environment.

If sub-step B2 exceeds the default thresholding of community, described place, this user is saved to access user set, and records this user place cell id and residing channel impulse response the window's position;

If sub-step B3 is lower than the default thresholding of community, described place, using this user as non-access user, and the full zero setting of channel impulse response tap to this user;

Sub-step B4, after comparison, using the number of users in described access user set as access user sum.

For example, suppose that CIR length of window is W=16, each community has 8 CIR windows (basic Midamble code length is P=128chips); This community sequence number is 0, and ground floor adjacent cell sequence number is 1 to 6, h (n) n=0,1 ..., the 6th, the full load CIR vector of community n; Noise power is σ ²; Access user set is U.Above-mentioned steps 102 can schematically show by following false code:

Upper example be intended that CIR window sequence number community n 1,2 ..., 8} be adjusted into 1+n × 8,2+n × 8 ..., 8+n × 8}, so just can read from any one CIR the window's position of its place cell id and community, place in the window sequence number after adjusting.After comparison, the number of users in access user set U is access user sum.

Step 103, judge whether described access user sum meets multi-plot joint channel estimating condition, if so, performs step 104, otherwise, execution step 105;

Because this community access user number can be controlled by this base station, but neighbor cell access user signal is uncontrollable for this community; Therefore, multi-plot joint channel estimating of the present invention adopts the identical access user restrictive condition of Liao Yudan community Steiner channel estimating.

In practice, described multi-plot joint channel estimating condition can be, described access user sum is no more than P/W, and wherein, P is training training length, the length of window that W is channel impulse response.

Step 104, carry out multi-plot joint channel estimating for the access user of each community;

The basic task of multi-plot joint channel estimating is that the MAI(multiple access of eliminating user self disturbs, Multiple Access Interference) and the ICI(co-channel interference brought of other user, Interchannel Interference), accurately estimate the CIR of access user.Be similar to single community Steiner channel estimation methods, multi-plot joint channel estimating also needs to construct channel estimate matrix G according to the Midamble code of access user.

The access user number of supposing n community (can be that sequence number is one of them community of 0 to 6) is K ⁽ⁿ⁾, access user set U comprises K element, has so

$K=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{K}^{\left(n\right)}---\left(1\right)$

Pilot reception signal can be expressed as:

$\begin{array}{c}{e}_m=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{G}_k{h}_k+z=\mathrm{Gh}+z\\ =\left(\begin{array}{cccc}{G}_1& G_2& ...& G_K\end{array}\right)\left(\begin{array}{c}{h}_1\\ h_2\\ .\\ .\\ .\\ h_K\end{array}\right)+z\end{array}---\left(2\right)$

Wherein, h is the CIR vector of K access user; G is channel estimate matrix; Z is noise vector; h _kbe the CIR of k access user; G _kbe the matrix being made up of this user's Midamble code, if the sequence number of its community, place is g, corresponding channel estimating window sequence number is l, has

Here $m_k^g={[m_{l,P-W+1}^g,...,m_{l,P}^g,m_{l,1}^g,m_{l,2}^g,...,m_{l,P}^g]}^T$ Community g(g=0,1 ..., 6) l Midamble code, W unit of the 1st the Midamble code cyclic shift in Ke Yougai community (k-1) obtains; G _kit is the row circular matrix on P × W rank.Be less than P/W if total access customer number that this community user detects is K, will there is unique maximum likelihood solution in h so:

$\hat{h}={\left(G^{H}G\right)}^{-1}{G}^H{e}_m---\left(4\right)$

If definition correlation matrix S=G ^hg, pilot reception signal is output as e after matched filtering _m=G ^he _m, formula (4) can be reduced to:

$\hat{h}=S^{-1}{e}_M---\left(5\right)$

Further, S is done to Cholesky decomposition computation and go out Cholesky factor R (upper triangular matrix), make S=R ^hr.Formula (5) can change into:

e _M＝R ^Hy （6）

$y=R\hat{h}---\left(7\right)$

Wherein, vectorial y is middle vector.Formula (6) and formula (7) illustrate, first by forward direction iterative complete on first after trigonometric equation again through second lower trigonometric equation of backward iterative once, also can obtain the maximum likelihood estimator of h.

Multi-plot joint channel estimating as desired signal processing, can improve accuracy of estimation adjacent cell signal greatly, but channel estimate matrix G is now no longer circular matrix, therefore can not adopt Fourier transform to substitute matrix inversion.

In order to reduce computation complexity, improve channel estimating speed, in a preferred embodiment of the present invention, can adopt Matrix C holesky to decompose and substitute the operation of inverting; Correspondingly, the described access user for each community carries out the step of multi-plot joint channel estimating, specifically can comprise:

The pilot reception signal of sub-step C1, the each access user in the each community of foundation is carried out matched filtering, obtains matched filtering Output rusults;

The acyclic structure of channel estimate matrix G, causes directly correlation matrix S being inverted or adopting the complexity of traditional C holesky decomposition up to O (K ³w ³).

Inventor herein notices, S has inherited G _ktoeplitz(Toeplitz matrix) structure, its be in fact one by K ²the Toeplitz-Block(Top profit thatch block matrix that individual W rank Toeplitz submatrix forms).At this moment, if adopt Schur Algorithm for Solving Cholesky factor R can make computation complexity be reduced to O (K ³w ²).

From matched filtering output, also must carry out KWP time multiplication, both amounts of calculation all exceed Schur algorithm, so that this operation becomes the bottleneck of restriction channel estimating speed.In order to address this problem, this preferred embodiment provides a kind of FFT of employing to calculate the shortcut calculation of matched filtering output, and principle and process are as follows:

For matched filtering operation, first calculate following formula with FFT:

Wherein,

it is the P dimension output vector that receives signal with the Midamble matrix matching of n community (being assumed to be sequence number r community) full load.So, K user's matched filtering output just can be constructed by above formula simply.For example, if access community is community 0, community 1 and community 3, and for simplicity, sets each access community and only have an access user, corresponding window sequence number is respectively Isosorbide-5-Nitrae, 7.Matched filtering output can be write:

$e_M={[{\mathrm{SQ}}_{1,e_m}{(1:W)}^T,{\mathrm{SQ}}_{2,e_m}{(3W+1:4W)}^T,{\mathrm{SQ}}_{3,e_m}{(6W+1:7W)}^T]}^T---\left(9\right)$

As seen from the above analysis, adopt FFT to calculate matched filtering output, operand is down to N (P+Plog ₂p).

The correlation matrix that the pilot code of sub-step C2, the each access user in structure You Ge community forms;

Textural from correlation matrix, need K if directly do Matrix Multiplication rule ²w ²p time, even if the Toeplitz-Block structure of known correlation matrix also at least needs K ²wP time; So that this operation becomes the bottleneck of restriction channel estimating speed.In order to address this problem, this preferred embodiment provides the shortcut calculation of a kind of FFT of employing structure correlation matrix, principle and process as:

Suppose that k1 and k2 access user are respectively in community n1, a w1 channel estimating window and community n2, a w2 channel estimating window, two users' Midamble correlation matrix is M _{k1, k2}, i.e. M _{k1, k2}=G _k1 ^hg _k2; Be similar to mutual interference matrix, M _{k1, k2}be W × W rank matrix with Toeplitz structure, and the capable j column element of its i can be tried to achieve by FFT, have

M _k1,k2(i,j)＝SQ _n1,n2((P-(w1-w2)W+i-j+1)modP) （10）

SQ _n1,n2＝IDFT[conj(DFT(m _n1)).×DFT(m _n2)] （11）

According to FFT character, frequency domain conjugate pair is answered the anti-pleat of time domain.If k1 and k2 access user community of living in are exchanged, two users' Midamble correlation matrix elements with do not exchange before all come from same sequence SQ _{n1, n2}, different is only to put in order: SQ _{n2, n1}(i)=SQ _{n1, n2}((P-i+2) modP).This point is that mutual interference matrix is not available.Therefore,, after simplifying by FFT computing, structure correlation matrix only needs

inferior multiplication, is far smaller than KWP time (N≤K).

For sub-step C1 and sub-step C2, its execution sequence is strictly restriction not, can arbitrarily exchange.

Sub-step C3, the Cholesky factor of calculating this correlation matrix storage;

Sub-step C4, the described Cholesky factor of foundation, carry out forward direction iteration one time to described matched filtering Output rusults, and backward iteration is carried out once in output to forward direction iteration, and the channel impulse response family that obtains the each access user in each community is estimated.

For example, can first adopt forward direction alternative manner to calculate e _m=R ^hmiddle vector y in y, then calculates with backward alternative manner

in channel impulse response family estimate

Step 105, carry out many communities counteracting serial interference channel estimating for the access user of each community;

The channel rough estimate result that this step mainly obtains according to step S101 is disturbed elimination, obtains more accurate CIR, specifically can comprise:

Sub-step D1, initialization iterations;

Sub-step D2, the current channel estimation results of foundation, calculate the channel impulse response power of the current access user in each community, and select to have the access user of maximum channel impulse response power;

Sub-step D3, this access user is disturbed to recovery;

Sub-step D4, from current channel estimation results, balance out the interference of this user to other user;

Sub-step D5, iterations add 1;

Sub-step D6, judge whether this iterations reaches preset value, if so, stop iteration, and using current channel estimation results as many communities counteracting serial interference channel estimation results, otherwise, return to the new channel impulse response power that calculates the current access user in each community.

Be similar to the Multi-User Detection mode of cdma system, what just adopt here is serial mode, and the unit of iteration is an access user, balances out the interference that this access user of recovery causes other non-its community, place access user at every turn.

Suppose that pilot reception signal is e _m; The basic Midamble code of community n is m _n, and the CIR of u window is

the access adjacent cell further recording in hypothesis access user set U is community 1 and 3, counteracting serial interference according to comprising following mutual interference formula:

$\mathrm{IDFT}[\mathrm{DFT}\left(e_m\right)./\mathrm{DFT}\left(m_0\right)]=h^{\left(0\right)}+\underset{u+8\×1\∈U}{\mathrm{\Σ}}{I}_u^{\left(\mathrm{0,1}\right)}+\underset{u+8\×3\∈U}{\mathrm{\Σ}}{I}_u^{\left(\mathrm{0,3}\right)}---\left(12\right)$

$\mathrm{IDFT}[\mathrm{DFT}\left(e_m\right)./\mathrm{DFT}\left(m_1\right)]=h^{\left(1\right)}+\underset{u+8\×0\∈U}{\mathrm{\Σ}}{I}_u^{\left(\mathrm{1,0}\right)}+\underset{u+8\×3\∈U}{\mathrm{\Σ}}{I}_u^{\left(\mathrm{1,3}\right)}---\left(13\right)$

$\mathrm{IDFT}[\mathrm{DFT}\left(e_m\right)./\mathrm{DFT}\left(m_3\right)]=h^{\left(3\right)}+\underset{u+8\×0\∈}{\mathrm{\Σ}}{I}_u^{\left(\mathrm{3,0}\right)}+\underset{u+8\×1\∈U}{\mathrm{\Σ}}{I}_u^{\left(\mathrm{3,1}\right)}---\left(14\right)$

Wherein,

for the mutual interference that the access user u of community n2 produces Dui community n1, can further be expressed as:

$I_u^{(n1,n2)}={\mathrm{MIM}}_u^{(n1,n2)}{h}_u^{\left(n2\right)}---\left(15\right)$

Here,

be P × W rank mutual interference matrix with Toeplitz structure, and the capable j column element of its i can be tried to achieve by FFT, have

${\mathrm{MIM}}_u^{(n1,n2)}(i,j)=\mathrm{IDFT}[\mathrm{DFT}\left(m_{n2}\right)./\mathrm{DFT}\left(m_{n1}\right)]\left((P-(u-1)W+i-j+1)\mathrm{mod}P\right)---\left(16\right)$

The main computing of formula (16) is the IDFT of a division result.Because the DFT(discrete Fourier transform of the basic Midamble code of community, Discrete Fourier Transform) in step 101, try to achieve, or stored in the time of cell initial and all come again when channel estimating at every turn, so operand little.In addition, can find out from formula (16), the mutual interference matrix element that belongs to each user of Yi Ge community together comes from same vector, if access community number is N, only needs

inferior IDFT(inverse discrete Fourier transform, Inverse Discrete Fourier Transform) operation.It is exactly more than the reason that many communities of the present invention counteracting serial interference channel estimating has low complex degree.

Below in conjunction with the flow chart shown in Fig. 2, the implementation procedure of step 105 is elaborated, supposes initial value=1 of iterations, specifically can comprise:

Step 151, by access user set U be assigned to set U ';

Access user set U is temporarily stored in set U ', in being convenient to like this operate afterwards, does not change original access user information.

In step 152, selection set U ', there is the access user of maximum peak power;

In practice, can be by each access user information of mark in pair set U ', comprise place cell id and the window's position, utilize CIR rough estimate result to calculate pilot power, then filter out the user u(hypothesis user u with maximum power and be in n2 community).

Step 153, recover the interference of this user to other non-its place community user;

The user profile screening for step 152, recovers the interference that this user produces other community

(n1 ≠ n2), namely calculates the product of mutual interference matrix and CIR.

Step 154, offset the interference of this user to other non-its place community user;

In practice, can balance out this part from other community CIR by following formula disturbs:

$h^{\left(n1\right)}=h^{\left(n1\right)}-I_u^{(n1,n2)}---\left(16\right)$

Step 155, this user is removed U ' from set.

So far, one time iterative process completes, and therefore in step 156, iterations is added to 1.

Step 156, iterations add 1;

Step 157, judge whether iterations reaches preset value N, if so, performs step 159, otherwise execution step 158;

Here, the numerical value of N can be chosen by emulation.

Step 158, judge that whether set U ' is empty, if so, return to step 151, otherwise, step 152 returned to;

Judge that whether set U ' is empty, checks whether the mutual interference of access user has all offset once.If it is accent is offset for the second time; Otherwise get next maximum peak power user and continue Interference Cancellation.

Step 159, many communities counteracting serial interference channel estimating are finished.

Step 106, for the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, carry out Denoising disposal.

In practice, described Denoising disposal step can comprise: for the channel impulse response tap of access user in the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, if its mould value is less than the second default thresholding, by its zero setting, otherwise the numerical value of this channel impulse response tap is constant.In practice, doubly, wherein, the numerical value of α can be by TD-SCDMA default for the α that can described second default thresholding is set to noise variance, and the present invention is not limited it.

With reference to Fig. 3, show the flow chart of the multi-cell channel estimation method embodiment 2 in a kind of TD-SCDMA system of the present invention, specifically can comprise:

Step 301, adopt single community Steiner channel estimating, channel rough estimate is carried out in Dui Ge community;

Step 302, according to channel rough estimate result, determine access user sum;

Step 303, judge whether described access user sum meets multi-plot joint channel estimating condition, if so, performs step 304, otherwise execution step 305;

Step 304, judge the relation of current time slots access user and a upper time slot access user, if this relation is not that many new access users, a few former access user or access user are constant, carry out multi-plot joint channel estimating for the access user of each community;

Step 305, carry out many communities counteracting serial interference channel estimating for the access user of each community;

Step 306, for the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, carry out Denoising disposal.

With respect to embodiment 1, the present embodiment is by judging the relation of current time slots access user and a upper time slot access user, to the access user processing of classifying, to reduce operand.It should be noted that, if the firm initialization of network, described judgement can be omitted, and directly carries out multi-plot joint channel estimating for the access user of each community; Otherwise, require to judge to classify processing.

The implementation procedure that the flow chart showing below in conjunction with Fig. 4 carries out multi-plot joint channel estimating to the access user for each community is elaborated, and specifically can comprise:

Step 401, matched filtering;

Step 402, correlation matrix structure;

Step 403, the Cholesky factor of calculating this correlation matrix storage;

Trigonometric equation group under step 404, forward direction iterative solution;

Trigonometric equation group in step 405, backward iterative solution.

In a preferred embodiment of the present invention, described method can also comprise:

If step e 1 learns that through step 304 judgement this relation is that a few former access user or access user are constant, the Cholesky factor of an above time slot is as the Cholesky factor of current time slots;

If step e 2 learns that through step 304 judgement this relation is many new access users, obtains the Cholesky factor of current time slots according to the Cholesky factor iteration of a upper time slot;

Step e 3, according to the Cholesky factor of current time slots, current matching filtering Output rusults is carried out to forward direction iteration one time, and forward direction iteration output is carried out once to backward iteration, the channel impulse response family that obtains the each access user in each community is estimated;

According to TD-SCDMA system, even if travelling carriage during with 120km/h high-speed mobile, distance mobile in each subframe time be only 0.16m, and therefore the during this period of time variation of interference user is also not obvious.If there is interference user interfering energy too little or do not produced interference, do not need to re-construct channel estimate matrix and calculate Cholesky factor R, can be directly alternative by the R originally storing, i.e. step e 2.If but the new interference user of coming in but still will cause very large error by the words of the R computing of previous moment, at this moment can adopt the R of a time slot in alternative manner utilization to solve the R of current time slots, i.e. step e 3.

Because step 403 is stored in the Cholesky factor of calculating this correlation matrix, the benefit of this storage is embodied in current time slots access customer number than in the situation of a upper time slot many 1, now, can in the time calculating the R of current time slots, simplify amount of calculation.Below in conjunction with example, alternative manner principle is described:

Suppose that the interference user u of new access is from community n, now total interference user sum becomes K+1, by the Midamble matrix of user u be put into G after, new channel estimate matrix G' becomes:

G'＝[G,G _u] （16）

To adding the correlation matrix S'=G' after new interference user ^hg' and its Cholesky factor R ' carry out same way piecemeal:

$S^{\′}=\left[\begin{array}{cc}{G}^{H}G& M_u\\ M_u& M_{u,u}\end{array}\right]---\left(17\right)$

$R^{\′}=\left[\begin{array}{cc}R& r\\ 0& r_{u,u}\end{array}\right]---\left(18\right)$

Wherein, R is the Cholesky factor of a upper time slot; M _u,ufor the auto-correlation Midamble matrix of user u; KW × W rank matrix M _ufor the cross-correlation matrix of user u and a upper time slot access user; r _u,ufor W rank upper triangular matrix; R is KW × W rank matrixes.By G' ^hg'=R' ^hr' relatively corresponding blocks position can obtain:

R ^Hr＝M _u （19）

r _u,u＝chol(M _u,u-r ^Hr) （20）

Two relational expressions represent r above, r _u,ucan and do Cholesky decomposition to a W rank matrix by forward direction iteration obtains.

More than explanation, if meet the condition of access user many 1, new Cholesky factor R ' comprised old R, and also remainder also can calculate by old R, like this, with respect to again adopting Schur algorithm to calculate R', can obviously reduce amount of calculation.

In a preferred embodiment of the present invention, after Duo community counteracting serial interference channel estimating is finished, also should empty access user set U.The object of so doing is the classification processing for the ease of multi-plot joint channel estimating, the condition of processing because can classify is the double combination channel estimating method that all adopts, and need to judge the relation of current time slots and a upper time slot access user, if and a upper time slot channel estimating is to adopt counteracting serial interference method, null set U can avoid makeing mistakes so.If certainly do not emptied, also can take other measure to avoid above-mentioned contingent mistake, because this problem does not belong to emphasis of the present invention, therefore not to repeat here.

Each embodiment in this specification all adopts the mode of going forward one by one to describe, and what each embodiment stressed is and the difference of other embodiment, between each embodiment identical similar part mutually referring to.

Above to the multi-cell channel estimation method in a kind of TD-SCDMA system provided by the present invention, be described in detail, applied specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment is just for helping to understand method of the present invention and core concept thereof; , for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention meanwhile.

Claims (12)

1. the multi-cell channel estimation method in TD-SCDMA system, is characterized in that, comprising:

Adopt single community Steiner channel estimating, channel rough estimate is carried out in Dui Ge community;

According to channel rough estimate result, determine access user sum;

Judge whether described access user sum meets multi-plot joint channel estimating condition, described multi-plot joint channel estimating condition is, described access user sum is no more than P/W, and wherein, P is training length, the length of window that W is channel impulse response; If so, carry out multi-plot joint channel estimating for the access user of each community, otherwise, carry out many communities counteracting serial interference channel estimating for the access user of each community;

For the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, carry out Denoising disposal.

2. the method for claim 1, is characterized in that, described according to channel rough estimate result, determines the step of access user sum, comprising:

For the channel rough estimate result Zhong Ge each user's in community channel impulse response, the default thresholding of its power and community, place is compared;

If exceed the default thresholding of community, described place, this user is saved to access user set, and records this user place cell id and residing channel impulse response the window's position;

If lower than the default thresholding of community, described place, using this user as non-access user, and the full zero setting of channel impulse response tap to this user;

After comparison, using the number of users in described access user set as access user sum.

3. method as claimed in claim 2, is characterized in that, the default thresholding of community, described place is β times of measurement noise power, wherein, in the time of networking or cell initial, determine β initial value, or, in estimating, multi-cell channel dynamically adjusts described β initial value according to dynamic channel environment.

4. the method for claim 1, is characterized in that, the described access user for each community carries out the step of multi-plot joint channel estimating, comprising:

Pilot reception signal according to the each access user in each community is carried out matched filtering, obtains matched filtering Output rusults;

The correlation matrix that the pilot code of the each access user in structure You Ge community forms;

Calculate the Cholesky factor of this correlation matrix;

According to the described Cholesky factor, described matched filtering Output rusults is carried out to forward direction iteration one time, and backward iteration is carried out once in output to forward direction iteration, the channel impulse response family that obtains the each access user in each community is estimated.

5. the method for claim 1, is characterized in that, the described access user for each community carries out the step of many communities counteracting serial interference channel estimating, comprising:

Initialization iterations;

According to current channel estimation results, calculate the channel impulse response power of the current access user in each community, and select to have the access user of maximum channel impulse response power;

This access user is disturbed to recovery;

From current channel estimation results, balance out the interference of this user to other user;

Iterations adds 1;

Judge whether this iterations reaches preset value, if so, stop iteration, and using current channel estimation results as many communities counteracting serial interference channel estimation results, otherwise, the new channel impulse response power that calculates the current access user in each community returned to.

6. method as claimed in claim 4, is characterized in that, employing FFT carries out the constructor of described matched filtering operation and correlation matrix, and, adopt Schur algorithm to calculate the Cholesky factor of this correlation matrix.

7. method as claimed in claim 4, is characterized in that, before the access user for each community carries out multi-plot joint channel estimating, also comprises:

Judge the relation of current time slots access user and a upper time slot access user, if this relation is not that many new access users, a few former access user or access user are constant, carry out multi-plot joint channel estimating for the access user of each community.

8. method as claimed in claim 7, is characterized in that, also comprises:

Because of the period of the day from 11 p.m. to 1 a.m, store this Cholesky factor at the Cholesky that calculates this correlation matrix;

If this relation is that a few former access user or access user are constant, the Cholesky factor of an above time slot is as the Cholesky factor of current time slots;

If this relation is many new access users, obtain the Cholesky factor of current time slots according to the Cholesky factor iteration of a upper time slot;

And, according to the Cholesky factor of current time slots, current matching filtering Output rusults is carried out to forward direction iteration one time, and backward iteration is carried out once in output to forward direction iteration, the channel impulse response family that obtains the each access user in each community is estimated;

9. the method for claim 1, is characterized in that, described Denoising disposal step comprises:

For the channel impulse response tap of access user in the result of described multi-plot joint channel estimating Huo Duo community counteracting serial interference channel estimating, if its mould value is less than the second default thresholding, by its zero setting.

10. method as claimed in claim 9, is characterized in that, the α that the described second default thresholding is noise variance doubly.

11. the method for claim 1, is characterized in that, the single community of described employing Steiner channel estimating, and the step that channel rough estimate is carried out in Dui Ge community comprises:

Pilot reception signal for current area is carried out FFT computing;

Use the FFT of the basic Midamble code in FFT operation result Dian Chuge community;

A division result is carried out to IFFT computing, obtain the channel rough estimate result of current area.

12. methods as claimed in claim 11, is characterized in that, obtain the basic Midamble code of adjacent cell by following steps:

The basic Midamble code of adjacent cell is transferred to base station by radio network controller;

Base station is transferred to terminal by signaling by the basic Midamble code of adjacent cell.

Images