Invention content
One object of the present invention is at least that provide a kind of time-domain channel estimating method and device based on MMSE.The base
The complexity of matrix inversion is at least reduced in the time-domain channel estimating method and device of MMSE so that MMSE's in the time domain
Channel estimation is possibly realized.
According to the first aspect of the invention, a kind of channel estimation methods of MMSE are provided, it is characterised in that:
Utilize formulaExecute the channel estimation of the MMSE of time domain, wherein logical
It crosses and the nonzero path L ' in multipath is selected to realize matrixCalculating;
Wherein,It is the LS channel estimation results of reference signal RS;F is Fourier transform matrix;L is that maximum channel is long
Degree;FPLIt is the P obtained by L is arranged before selecting corresponding to the row of the reference signal RS and the of the Fourier transform matrix
× L matrixes:FCLIt is C × L matrixes by selecting to obtain by L is arranged before the of Fourier transform matrix;σ2It is noise power;RhhIt is
The correlation matrix of time domain channel.
According to the preferred embodiment of the invention, the threshold value of the path energy in multipath is determined, and look for based on the threshold value
To the nonzero path L ' more than the threshold value.
According to the preferred embodiment of the invention, by only preservingThe first column element and according to corresponding to described
The row and column of nonzero path position constructs to obtain matrixInWherein Rh'h'It corresponds to non-
The correlation matrix of the time domain channel in zero path.
According to the preferred embodiment of the invention, pass throughCalculate influence of the path offset to channel estimation, whereinW is the matrix about the path offset, Rh'h'Correspond to nonzero path
The correlation matrix of time domain channel.
According to the preferred embodiment of the invention, described in acquisitionAs a result the Wiener filtering of time domain is executed, then is held
Fourier transformation F of the row from time domain to frequency domainCL。
According to the second aspect of the invention, a kind of channel estimation methods of MMSE are provided, it is characterised in that:
Obtain the LS channel estimation results of the frequency domain of reference signal RS
It willFourier inversion at rough time domain channel estimationWherein, FPLIt is by selection pair
L row before the of the row of reference signal RS described in Ying Yu and the Fourier transform matrix F and the P × L matrixes obtained;
Nonzero path L ' is selected from all path L of multipath;
Calculate the matrix under nonzero path L 'Wherein σ2It is noise power;Rh'h'Correspond to non-zero
The correlation matrix of the time domain channel in path;
Calculate the channel estimation of time domain
Influence of the path offset to channel estimation is calculated under nonzero path;And
The channel estimation Fourier transformation of the time domain for diameter bias effect of passing by one's way will be computed into the channel estimation of frequency domain.
According to the preferred embodiment of the invention, wherein the step for selecting nonzero path L ' from all path L of multipath
Suddenly, including:
It determines the threshold value of the path energy in multipath, and the non-zero road more than the threshold value is found based on the threshold value
Diameter L '.
According to the preferred embodiment of the invention, wherein by only preservingThe first column element and according to corresponding to
The row and column of nonzero path position constructs to obtain the matrixIn
According to the preferred embodiment of the invention, pass throughCalculate influence of the path offset to channel estimation,
WhereinW is the matrix about the path offset.
According to the preferred embodiment of the invention, time domain is executed before the Fourier transformation is at the channel estimation of frequency domain
Wiener filtering.
According to the third aspect of the invention we, a kind of channel estimator of MMSE is provided, it is characterised in that:
The channel estimator of the MMSE is configured as utilizing formulaIt executes
The channel estimation of the MMSE of time domain, wherein realize matrix by selecting the nonzero path L ' in multipath's
It calculates;
Wherein,It is the LS channel estimation results of reference signal RS;F is Fourier transform matrix;L is that maximum channel is long
Degree;FPLIt is the P obtained by L is arranged before selecting corresponding to the row of the reference signal RS and the of the Fourier transform matrix
× L matrixes:FCLIt is C × L matrixes by selecting to obtain by L is arranged before the of Fourier transform matrix;σ2It is noise power;RhhIt is
The correlation matrix of time domain channel.
According to the preferred embodiment of the invention, the channel estimator of the MMSE determines the threshold value of path energy, and is based on
The threshold value finds the nonzero path L ' more than the threshold value.
According to the preferred embodiment of the invention, the channel estimator of the MMSE is by only preservingFirst row member
Element and construct to obtain the matrix according to the row and column corresponding to the nonzero path positionInWherein Rh'h'Correspond to the correlation matrix of the time domain channel of nonzero path.
According to the preferred embodiment of the invention, the channel estimator of the MMSE passes throughTo calculate path offset pair
The influence of channel estimation, whereinW is the matrix about the path offset,
Middle Rh'h'Correspond to the correlation matrix of the time domain channel of nonzero path.
According to the preferred embodiment of the invention, the channel estimator of the MMSE is to described in acquisitionAs a result it executes
The Wiener filtering of time domain, then execute the Fourier transformation F from time domain to frequency domainCL。
According to the fourth aspect of the invention, a kind of channel estimator of MMSE is provided, including:
LS channel estimation modules are configured to obtain the LS channel estimation results of the frequency domain of reference signal RS
Fourier inversion module, be configured to byFourier inversion at rough time domain channel estimationWherein, FPLBe by selection corresponding to the reference signal RS row and the Fourier transform matrix F the
Preceding L is arranged and P × L matrixes of acquisition;
Path selection module is configured to select nonzero path L ' from all path L of multipath;
Matrix generation module is configured to calculate the matrix under nonzero path L 'Wherein σ2It is noise work(
Rate;Rh'h'Correspond to the correlation matrix of the time domain channel of nonzero path;
MMSE modules are configured to calculate the channel estimation of time domain
Path offset module is configured to calculate influence of the path offset to channel estimation under nonzero path;And
Fourier transformation module is configured to become the channel estimation Fourier for being computed the time domain for diameter bias effect of passing by one's way
Change the channel estimation of frequency domain into.
According to the preferred embodiment of the invention, wherein the path selection module is additionally configured to determine the road in multipath
The threshold value of diameter energy, and the nonzero path L ' more than the threshold value is found based on the threshold value.
According to the preferred embodiment of the invention, the matrix generation module is by only preservingThe first column element simultaneously
And construct to obtain the matrix according to the row and column corresponding to nonzero path positionIn
According to the preferred embodiment of the invention, the path offset module passes throughTo calculate the path offset pair
The influence of channel estimation, whereinW is the matrix about path offset.
According to the preferred embodiment of the invention, the fourier transformation module is additionally configured in the Fourier transformation
At the Wiener filtering for executing time domain before the channel estimation of frequency domain.
Specific implementation mode
Flow chart in attached drawing and block diagram, it is illustrated that according to the device of various embodiments of the invention, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part for a part for one module, program segment, or code of table, the module, program segment, or code includes one or more
Executable instruction for implementing the specified logical function.It should also be noted that in some implementations as replacements, institute in box
The function of mark can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are practical
On can be basically executed in parallel, they can also be executed in the opposite order sometimes, this is depended on the functions involved.Also it wants
It is noted that the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, Ke Yiyong
The dedicated hardware based system of defined functions or operations is executed to realize, or can be referred to specialized hardware and computer
The combination of order is realized.
The various embodiments of the present invention will only be described in detail by exemplary approach now.
As previously mentioned, present applicant proposes the MMSE channel estimation methods based on reference signal (RS).Fig. 1 and Fig. 2 shows
The cell dedicated RS patterns in LTE.
Fig. 1 is the mapping (normal cyclic prefix) of downlink reference signal according to example embodiment of the present invention.
Fig. 2 is the mapping (extended cyclic prefix) of downlink reference signal according to example embodiment of the present invention.
As shown in the figure, the blockage in figure indicates that known signal, the application i.e. will be in the letters estimated by the known signal
On the basis of road, the channel of other positions for being not inserted into reference signal is all estimated.Specifically, present applicant proposes one
Kind novel time-domain channel estimating method and device based on MMSE.
For example, under the weak channel of multipath, the time domain channel estimation of MMSE can be write as:
Wherein,It is the LS channel estimation results of reference signal RS;F is Fourier transform matrix;L is that maximum channel is long
Degree;FPLIt is the P obtained by L is arranged before selecting corresponding to the row of the reference signal RS and the of the Fourier transform matrix
× L matrixes:FCLIt is C × L matrixes by selecting to obtain by L is arranged before the of Fourier transform matrix;σ2It is noise power;RhhIt is
The correlation matrix of time domain channel.The RhhIt is unrelated with multipath, therefore RhhBe element on diagonal matrix and diagonal line it is multipath
Power has
Rhh=E (h*h)L*L (2)
Referring to formula (1), matrixSize be L × L.Any simplified processing is not being carried out to the matrix
In the case of, the complexity which directly calculates is difficult to receive, this be also in the prior art the channel based on MMSE estimate
The reason of meter focuses primarily upon frequency domain.It is well known that the channel estimation of time domain is more preferable than the channel estimation of frequency domain, in order to realize
The MMSE channel estimations of time domain, the application propose:Under the weak channel of multipath, the position of no any path energy is deleted, i.e.,
The nonzero path L ' in multipath is selected, so as to realizeSimplification.Namely formula (1) can be rewritten as
Wherein, L ' is the nonzero path in all paths, Rh'h'Correspond to the Correlation Moment of the time domain channel of nonzero path
Battle array.The meaning of formula (3) is at this time toSize be reduced to L ' × L ' from L × L, to greatly subtracting
It is fewThe calculation amount inverted, and such L ' × L ' matrix inversion calculation is existing computing capability to connect
It receives.
Fig. 3 shows the block diagram of MMSE channel estimations according to example embodiment of the present invention.MMSE channel estimators 300
May include LS channel estimation modules 310, Fourier inversion module 320, path selection module 330, matrix generation module
340, MMSE modules 350, path offset module 360, time domain Wiener filter 370 and fourier transformation module 380.
Wherein, LS (also referred to as least square method) channel estimation module 310 is configured for receiving the number of reference signal RS
According to the LS channel estimation results of the frequency domain of acquisition reference signal RS
Fourier inversion module 320, be configured to byFourier inversion at rough time domain channel estimationWherein, FPLIt is the row by selection corresponding to the reference signal RS and the Fourier transform matrix F
L row before the and the P × L matrixes obtained;According to an embodiment of the invention, which is preferably fast Fourier contravariant
Change IFFT.
Path selection module 330 is configured to select nonzero path L ' from all path L of multipath;
Matrix generation module 340 is configured to calculate the matrix under nonzero path L 'Wherein σ2It is to make an uproar
Acoustical power;Rh'h'Correspond to the correlation matrix of the time domain channel of nonzero path;
MMSE modules 350 are configured to carry out the channel estimation of time domain according to formula (4),
Path offset module 360 is configured to calculate influence of the path offset to channel estimation under nonzero path;And
Fourier transformation module 380 is configured to the channel estimation at frequency domain by the channel estimation Fourier transformation of time domain.
Each embodiment according to the present invention, MMSE channel estimators further preferably include time domain Wiener filter 370, by with
It sets for acquisitionAs a result the Wiener filtering of time domain is executed, wherein W is the matrix about the path offset.So
Afterwards, then by fourier transformation module 380 execute time domain to frequency domain Fourier transformation.The Fourier transformation is preferably in quick Fu
Leaf transformation (FFT).The Wiener filter of the time domain realizes more accurate channel estimation.
As described above, skilled addressee readily understands that the function that module is realized above.In order to allow the reality of the application
Existing mode is clearer, and the application will further introduce the calculating of several key points.
1st point:Time domain correlation matrix Rh'h'Calculating
It is calculated by IFFTLater, can be estimated by rough time domain channelObtain time domain phase
Close matrix Rhh。
Then R is usedhhGreatest member to calculate multipath threshold value, which can be indicated by formula (6):
Threshold value=threshold factor × Max (Rhh) (2)
The nonzero path L ' more than threshold value is found based on the threshold value, and thus to obtain the time domain Correlation Moment under nonzero path
Battle array Rh'h', matrix Rh'h'Diagonal item be made of the power of these multiple nonzero paths, and due to independently of multipath, the Rh'h'
Nondiagonal term all 0.
2nd point:Correlation matrixCalculating
In view of directly calculating correlation matrixComplexity it is still very huge, this application provides simpler
Mode calculatesIt considersStructure be as follows:
Above-mentioned matrix (7) is observed, it can be found that each element in matrix can be constructed to obtain by the element of first row,
Therefore the application preferably only preserves the first column element, and the element of other row is then obtained by the first column element.This is greatly saved
Memory space is saved.
Further, the application preferably constructs to obtain matrix by selecting the row and column corresponding to nonzero path positionIn each element.
3rd point:Path offset
Correlation matrixIt is related with the offset of RS.In LTE system, due to there are 6 kinds of different RS offsets (for example,
Offset=0,1 ..., 5), it is therefore desirable to calculate the matrix under different RS drift conditionsBut inventor has found in MMSE
The calculating that execution route deviates again after module calculates is also feasible, moreover, this is advantageous in that further
The calculating process for simplifying matrix reduces the complexity of operation matrix operation.Here is can be after MMSE module arithmetics
The proof of execution route calculations of offset again.
Assuming that the matrix that the offset of reference signal RS is 0 is FPLAnd the matrix that the offset of RS is k isThen have:
Wherein
Then formula (1) can be rewritten as:
Formula (10) means the channel estimation that identical method can be used to carry out MMSE, regardless of which kind of RS offset is just
It is used.Therefore it execution route can be deviated to obtain the estimated result of real channel after MMSE estimations.
4th point:Time domain Wiener filter
In LTE system, there are several OS (OFDM symbol) comprising reference signal in a subframe.It is terrible
To more accurate channel estimation, a kind of method appropriate be used together different OS channel estimation results be weighted it is flat
.
Wherein,Be formula (4) in an OS as a result, wiIt is the Wiener filter factor, i depends on system bandwidth.
In conjunction with reference to Fig. 3 it is found that formula (11) executes before Fourier transformation FFT.But those skilled in the art can be with
Understand, which can be calculate by the following formula with being carried out after FFT:
WhereinBut in formula (11),Only include a elements of L ', andIncluding N number of element, N
It is the number of LTE system sub-carriers, this is much larger than L '.Therefore computation complexity can be reduced using formula (11).
Fig. 4 shows the flow chart of the method for MMSE channel estimations according to a preferred embodiment of the invention.This method 400
Start from step S401.
Step S410 obtains the LS channel estimation results of the frequency domain of reference signal RS
Step S420, willFourier inversion at rough time domain channel estimationWherein, FPLIt is
P × L the squares obtained by L is arranged before selecting corresponding to the row of the reference signal RS and the of the Fourier transform matrix F
Battle array;
Step S430 selects nonzero path L ' from all path L of multipath.The step realizes matrix size from L × L
It is reduced to L ' × L ', to considerably reduceThe calculation amount inverted.According to an embodiment of the invention, the step
Suddenly it preferably includes:It determines the threshold value of the path energy in multipath, and the non-zero more than the threshold value is found based on the threshold value
Path L '.The threshold value can be for example calculated by above formula (5)-(6).
Step S440 calculates the matrix under nonzero path L 'Wherein σ2It is noise power;Rh'h'It is
Corresponding to the correlation matrix of the time domain channel of nonzero path.For matrix thereinThe present invention is preferably by only protecting
It depositsThe first column element and construct to obtain according to the row and column corresponding to nonzero path position.This reduce matrixes
Memory space.
Step S450 calculates the channel estimation of time domainBy above-mentioned steps
The optimization processing of S420-S440, the application can rapidly realize the time domain channel estimation of step S450.
Step S460 calculates influence of the path offset to channel estimation under nonzero path.The step is for considering path
Deviate the influence to MMSE channel estimations.As previously mentioned, the application confirms that no matter which kind of RS is deviated by proving, can use
Identical method executes the time domain channel estimation of MMSE, and then execution route deviates.
Step S480 will be computed the channel estimation Fourier transformation of the time domain for diameter bias effect of passing by one's way into the letter of frequency domain
Estimate in road.
Although step S480 can be carried out directly, as previously mentioned, the present invention can also preferably first carry out step S470,
The Wiener filtering of time domain is first carried out, then executes step S480 again, by the channel estimation Fourier transformation of time domain at frequency domain
Channel estimation.The meaning for executing step S470 is more accurately obtain channel estimation.
Step S490 terminates.
Given the specification of the present invention for the purpose for showing and describing, but its be not intended to it is exhaustive or
Person is limited to the invention of disclosed form.It may occur to persons skilled in the art that many modification and variation.Those skilled in the art
It should be appreciated that the method and apparatus in embodiment of the present invention can be realized with software, hardware, firmware or combinations thereof.
Therefore, embodiment is in order to better illustrate the principles of the present invention, practical application and makes people in the art
Other staff in member it will be appreciated that the following contents and select and describe, that is, without departing from the spirit of the invention, do
The all modifications and replacement gone out are fallen in the scope of the present invention that appended claims define.