CN101170525B - MLSE simplification detection method and its device based on blocked QR decomposition - Google Patents

Abstract

The invention relates to a MLSE simplification detection method based on the blocking OR decomposition and a device thereof; wherein the method comprises the following steps: the blocking OR decomposition based on the channel matrix H segments the user signal vector Y which is detected and the transmitting symbol series vector used for detection; maximum likelihood iterative detection is implemented in each section of vector from Z to A; the judgement result is combined and output. The device comprises a channel estimator(32), a channel matrix decomposition module(35), a block ML detector(31) and a segmenting length selection module(36) which is connected with or internally installed in the ML detector; wherein the channel estimator(32), the channel matrix decomposition module(35) and the block ML detector(31) are connected with each other in turn. The device and the method iterate and process the user signals in segmentation, which greatly reduces the complication of the MLSE algorithm. Based on the certain performance gain, the complication of detection is greatly reduced at the cost of small quantity of detection performance; therefore, the detection method and the device which are simplified need the general requirement for the hardware and the processing ability of the hardware and offer convenience for the practical application.

Description

A kind of MLSE simplification detection method and device thereof that decomposes based on piecemeal QR

Technical field

The present invention relates to the anti-interference detection of baseband transmission in the wireless telecommunication system, be specifically related under the relevant in a large number situations of detected data under single-carrier system, particularly the DFT-S-OFDM system, use maximum-likelihood sequence estimation Maximum likelihood sequence detection, be called for short MLSE, as the detection method and the device thereof of the main means of receiving terminal.

Background technology

For the baseband transmission system of reality, crosstalking is difficult to avoid.When cross talk effects is serious, must proofread and correct or equilibrium the transfer function of whole system, make it near undistorted transmission conditions.The equalizer of receiving terminal places the intermediate frequency or the base band of receiver usually, produce the characteristic opposite with the characteristic of channel, be used for offsetting channel the time become, interference that the multipath transmisstion characteristic causes, especially at the wireless transmitting system of High Data Rate, decline can bring serious intersymbol interference.

Present digital filter or equalizer can be divided into: time domain equalization and frequency domain equalization, linear equalization and nonlinear equalization, adaptive equalization and non-self-adapting equilibrium, transversary and lattice structure; Can be divided into zero forcing algorithm, MMSE algorithm and RLS algorithm and their various variants or the like on the algorithm.Different equalizers, its computation complexity, cost, convergence rate, stability, tracking performance etc. are also inequality, and the corresponding scope of application is arranged.

Maximum likelihood sequence detection is used for the detected symbol sequence as a kind of detection method of best performance.Its basic thought is all possible list entries of traversal, therefrom selects with receiving symbol to have the list entries of maximum comparability as testing result.Because MLSE does not adopt equalization filter, thereby having avoided strengthening the noise of having introduced, under complexity acceptable prerequisite good performance has been arranged, is first-selected detection method.But its complexity becomes the emphasis that algorithm design is paid close attention to.The Viterbi algorithm utilizes trrellis diagram that MLSE is simplified, under the situation that the complexity that sequence length, correlation length are brought increases gradually, can reduce time of delay and relevant memory spaces such as sequence path storage by the method for cutting of suboptimum, and in certain certain accuracy of assurance under the length parameter of blocking.But can not simply use above-mentioned algorithm for sequence, and need satisfy the situation decline low complex degree of performance requirement according to entire system structural design algorithm to pursue higher performance cost ratio through precoding.

The DFT-S-OFDM system adopts DFT transfer pair sequence to carry out precoding at transmitting terminal, makes being associated of mass data when improving performance, utilizes merely that the complexity of MLSE is too high to be required high and be difficult for realizing at receiving terminal hardware and disposal ability thereof.

Summary of the invention

The technical issues that need to address of the present invention provide a kind of MLSE simplification detection method and operative installations thereof, on the basis that guarantees certain performance gain, reduce complexity, can solve in the DFT-S-OFDM system because the related data that the transmitting terminal precoding is introduced, after through multipath channel and noise stack, use the high complexity issue that causes when the Maximum likelihood sequence detection algorithm detects at receiving terminal.

First technical problem of the present invention solves like this, a kind of MLSE simplification detection method that decomposes based on piecemeal QR is provided, be applied in the single-carrier system baseband transmission, detect according to the subscriber signal Y of channel matrix H, may further comprise the steps its extraction by receiving terminal:

1.1) piecemeal QR decomposition: the product that described channel matrix H is decomposed into an orthogonal matrix Q and a upper triangular matrix R is that unit is divided into the piecemeal upper triangular matrix by row and column with this upper triangular matrix R with certain-length CL also; With signal transformation column vector Y ', send column vector S again, white noise conversion column vector W ' is that unit is divided into the K section by the row correspondence with this length C L; Described transmission column vector S is the transmission symbol sebolic addressing vector of arbitrary possibility non-precoded, and described white noise conversion column vector W ' is by white noise vector W conversion, and described white noise vector W is the response of white noise AWGN at frequency domain, wherein, and Y '=Q ^-1* Y, W '=Q ^-1* W;

1.2) the maximum likelihood iterative detection: from the corresponding section of all described transmission column vector S of 1 section reverse traversal of K section to the, send corresponding section of sequence X according to the iterate conduct of therefrom selecting corresponding section maximum likelihood of maximum-likelihood criterion with subscriber signal conversion column vector Y ';

1.3) with each section of described transmission sequence X in order array output as final detection result.

According to MLSE simplification detection method provided by the invention, described K is the natural number more than or equal to two.

According to MLSE simplification detection method provided by the invention, subscriber signal Y is subscriber signal vector Y.

According to MLSE simplification detection method provided by the invention, described length C L can select and set, and is used to regulate the complexity and the performance of detection method.

According to MLSE simplification detection method provided by the invention, described receiving terminal also comprises interference signal restorer and arrester, described MLSE simplification detection method also comprises the signal that utilizes described restorer to recover the air interface form of described testing result correspondence, be considered as interference signal, eliminate by described arrester to other users.

According to MLSE simplification detection method provided by the invention, described single-carrier system includes, but are not limited to DFT-S-OFDM system or IFDMA system.

Another technical problem of the present invention solves like this, provide a kind of MLSE that decomposes based on piecemeal QR to simplify checkout gear, be built in the single-carrier system in the receiving terminal, detect according to the subscriber signal Y of channel matrix H by receiving terminal its extraction, comprise channel estimator and carry out the piece ML detector of Maximum likelihood sequence detection, also comprise with described ML detector being connected, carrying out channel matrix decomposition module that channel matrix QR decomposes and be connected, select that partition length CL's cut apart length selection module with described ML detector;

Described channel matrix decomposition module (35) is used for described channel matrix H is decomposed into the product of an orthogonal matrix Q and a upper triangular matrix R;

Described ML detector (31), being used for described upper triangular matrix R is that unit is divided into the piecemeal upper triangular matrix by row and column with partition length CL, again with signal transformation column vector Y ', send column vector S, white noise conversion column vector W ' is that unit is divided into the K section by the row correspondence with described partition length CL, corresponding section from all described transmission column vector S of 1 section reverse traversal of K section to the, send corresponding section of sequence X according to the iterate conduct of therefrom selecting section corresponding maximum likelihood of maximum-likelihood criterion with subscriber signal conversion column vector Y ', with each section of described transmission sequence X in order array output as final detection result; Wherein, described transmission column vector S is the transmission symbol sebolic addressing vector of arbitrary possibility non-precoded, and described white noise conversion column vector W ' is by white noise vector W conversion, and described white noise vector W is the response of white noise AWGN at frequency domain, Y '=Q ^-1* Y, W '=Q ^-1* W.

Simplify checkout gear according to MLSE provided by the invention, described channel matrix decomposition module is the standalone module between described channel estimator and piece ML detector.

Simplify checkout gear according to MLSE provided by the invention, described channel matrix decomposition module is built in the described channel estimator.

Simplify checkout gear according to MLSE provided by the invention, describedly cut apart length to select module be standalone module.

Simplify checkout gear according to MLSE provided by the invention, the described length of cutting apart selects module to be built in the described ML detector.

Simplify checkout gear according to MLSE provided by the invention, described single-carrier system includes, but are not limited to DFT-S-OFDM system or IFDMA system.

MLSE provided by the invention simplifies checkout gear and method thereof, make segmentation MLSE process user signal become possibility by piecemeal QR decomposition to channel matrix, greatly reduce the complexity of MLSE algorithm, on the basis that guarantees certain performance gain, with a small amount of detection performance is that cost greatly reduces detection complexity, make that detection method and device after simplifying are general to hardware and disposal ability requirement thereof, thereby be convenient to practical application.

Description of drawings

Further the present invention is described in detail below in conjunction with the drawings and specific embodiments.

Fig. 1 is a DFT-S-OFDM system transmitting terminal signal processing schematic diagram.

Fig. 2 is a DFT-S-OFDM system receiving terminal signal processing schematic diagram.

Fig. 3 is that MLSE provided by the invention simplifies the checkout gear structural representation.

Embodiment

At first, inventive concept is described:

(1) the invention provides a kind of MLSE detection algorithm of simplification, may further comprise the steps:

Channel matrix H is carried out QR decompose, be about to H and be divided into an orthogonal matrix Q and a upper triangular matrix R, and be that unit is divided into K * K piecemeal upper triangular matrix by row and column with the certain-length upper triangular matrix R; Again with column vector Y ', S, W ' is that unit is divided into the K section by the row correspondence with this certain-length all, every segment table is shown Y ' _k, R _k, S _k, W ' _k, k=1..K, S is the transmission symbol sebolic addressing vector of arbitrary possibility non-precoded herein, and W is the response of AWGN at frequency domain, and Y is the column vector of receiving symbol, Y '=Q ^-1* Y, W '=Q ^-1* W;=Q ^-1It is the inverse matrix of orthogonal matrix Q.

From selecting and receiving sequence column vector Y ' through candidate's list entries S of segment processing _kThe sequence that maximum likelihood is arranged is as court verdict, wherein one section X of the transmission sequence that obtains judging _K

Process and Y ' _K-1, Y ' _K-2... Y ' ₁The maximum likelihood decision that iterates can obtain X successively _K-1, X _K-2..., X ₁, with X _kK ∈ (1..K) combination can obtain final judgement sequence.

The second, the MLSE checkout gear that the invention provides a kind of simplification is described, as shown in Figure 3, comprise at least: channel estimator 32, channel matrix QR decomposing module 35, piece ML detector 31.Wherein, ML detector 31 connects or the built-in length selection module 36 of cutting apart.

The channel matrix H that channel matrix QR decomposing module 35 estimates channel estimator 32 is carried out QR and is decomposed, send into piece ML detector 31, cutting apart length simultaneously selects module 36 also parameter " cutting apart length selects " also to be sent into piece ML detector 31, the signal that participates in computing is cut apart, ML detector 31 is by inner segment iteration computing, and judgement is the correct input signal of array output also.

The MLSE checkout gear that the present invention simplifies is except parts separately, can also comprise interference signal restorer 33 and interference signal arrester 34, the signal that has ruled out recovers by interference signal, revert to the signal of air interface form, be considered as interference signal to other users, eliminate by interference signal arrester 34, thereby reduce interference other users.

The 3rd, describe the present invention to use the inventive method and device solves DFT-S-OFDM system practical problem in detail as example:

(1) the transmitting terminal data send modulated process, as shown in Figure 1, in the DFT-S-OFDM system, the bit stream of input is at first through ovennodulation, and the symbol that obtains is through the DFT precoding, carry out allocation of carriers after, send after adding Cyclic Prefix, through multidiameter fading channel, introduce noise, be received termination afterwards and take in row detection judgement.Special DFT-S-OFDM technology, in step 12: the step 11:DFT conversion that adds before the subcarrier mapping, realize the expansion of data symbol frequency domain, thereby providing better subcarrier mapping with when reducing peak-to-average force ratio PAPR, make between the big quantity symbol and produced correlation, be equivalent to increase greatly constraint length, thereby the trrellis diagram detection algorithm of Viterbi is no longer suitable here.

Considering has in U user's the DFT-S-OFDM system, the bit of each user's input serial to parallel conversion of flowing through, and every m bit divides one group into, carries out constellation mapping, obtains respective symbol, promptly modulates.User 1 symbol can be expressed as S ₁ ¹, S ₂ ¹...,

The symbol of user U can be expressed as S ₁ ^U, S ₂ ^U...,

Subscript is used to distinguish the user in the formula, and footnote is represented the sequence number of user symbol, M wherein _uThe sub-carrier number that expression user u distributes.The symbol of different user u is made M respectively _uThe DFT conversion of point obtains the symbolic vector through precoding.Symbol after the user u precoding can be expressed as C ₁ ^u, C ₂ ^u...,

With all users' conversion gained according to certain regular allocation to subcarrier, i.e. subcarrier mapping.Available centralised allocation, distributed allocation or other methods of salary distribution allocation of symbols after with precoding is to subcarrier.Sub-carrier vector after the employing centralised allocation

Sub-carrier vector after the employing distributed allocation

Symbol after distributing is made the DFT conversion that N is ordered, wherein N=M in regular turn ₁+ M ₂+ ...+M _UBe sent to multidiameter fading channel through the processing that adds Cyclic Prefix, serial to parallel conversion.

(2) receiving terminal Data Receiving demodulating process, as shown in Figure 2, receiving terminal at first carries out step 21:OFDM demodulation, at first remove the Cyclic Prefix of received signal in the demodulation after, it is made fast fourier transform fast Fourier transform, be called for short FFT; Carry out step 22 successively then: the user is carried out the carrier wave inverse mapping, step 23: therefrom extract targeted customer's signal, step 24: send MLSE algorithm detector, by branch's step 241: channel estimating, the channel estimation value of input can obtain to detect user's signal vector.Can obtain to detect user's signal vector.For the present invention, the MLSE algorithm detector is the MLSE algorithm detector through simplifying.More than complete process can be expressed as formula I:

y＝H*P*S+w I

Wherein, y is the column vector that receives, i.e. the matrix of N*1; S is the symbol rank vector after modulation, the i.e. matrix of N*1; P represents list entries is made precoding and allocation of carriers comprehensive function; H is a channel matrix, i.e. the N*N square formation; W is the response of white Gaussian noise at frequency domain.

Because each user's input bit stream is separate, and precoding carries out at unique user, so formula I can be reduced to formula II:

y _user[u]＝H _user[u]*P _user*S _user[u]+w _user[u] II

Wherein parameter all is at user u, y _UserBe the unique user column vector that receives, i.e. M _u* 1 matrix; S _UserBe the user symbol column vector after modulation, i.e. M _u* 1 matrix; P herein _UserOnly be precoding, i.e. M _u* M _uSquare formation, do not relate to the allocation of carriers process.Wherein, M _uBe the sub-carrier number of distributing to this user u, M _uIt is 25 multiple.

By the structure of above-mentioned DFT-S-OFDM as can be known, the existence of precoding makes the M of user u _uIndividual symbol associates.No matter to M _uThe distribution requirement, still consider from later development angle, use under the situation that the MLSE algorithm detects, all need a kind of reduction complexity, the optimized Algorithm of guaranteed performance adapts to the trend of development.This also is foothold of the present invention.

Precoding before the subcarrier mapping can realize the expansion of data symbol at frequency domain, thereby can provide better subcarrier mapping to reduce peak-to-average force ratio PAPR, meanwhile also makes data produce correlation.User u is long to be M _uSequence to be declared, M _uIndividual symbol is to be mutually related, and has brought the cost that detects time delay and register, has increased computation complexity greatly.

(3) in order not destroy such correlation as far as possible, we adopt the dimension with the feasible matrix of handling of method of matrix decomposition and iteration to reduce, and reduce complexity.

II can be rewritten as formula III in the receiving terminal formula:

Y＝H*S+W III

S is the transmission symbol sebolic addressing vector of arbitrary possibility non-precoded herein, and H represents the frequency domain response of channel and the acting in conjunction of DFT conversion, and W is the response of AWGN at frequency domain, and Y is the column vector of receiving symbol, i.e. the subscriber signal column vector.Decompose by H being carried out QR, H is divided into an orthogonal matrix Q and a upper triangular matrix R).Can get formula IV and formula V like this:

Y＝Q*R*S+W IV

Q ^-1*Y＝R*S+Q ^-1*W V

Might as well make Y '=Q ^-1* Y, W '=Q ^-1* W, then formula V can be rewritten as formula VI:

Y′＝R*S+W′ VI

R is a M herein _u* M _uSquare formation.Column vector Y ', R, S, W ' is a unit with a fixed length all is divided into plurality of sections by row, is divided in line number in one section herein and is called and cuts apart length, is designated as CL (error rate changes with CL).This sentences and divides four sections is example, and then formula VI can be represented by the formula:

$\left(\begin{array}{c}{y}_1^{\′}\\ y_2^{\′}\\ y_3^{\′}\\ y_4^{\′}\end{array}\right)=\left(\begin{array}{cccc}{R}_1& R_1^1& R_1^2& R_1^{3^{\′}}\\ 0& R_2& R_2^1& R_2^2\\ 0& 0& R_3& R_3^1\\ 0& 0& 0& R_4\end{array}\right)*\left(\begin{array}{c}{x}_1\\ x_2\\ x_3\\ x_4\end{array}\right)+\left(\begin{array}{c}{w}_1^{\′}\\ w_2^{\′}\\ w_3^{\′}\\ w_4^{\′}\end{array}\right)$

Can get formula VII and formula VIII: after the decomposition:

y′ ₄＝R ₄*x ₄+w′ ₄ VII

$y_3^{\′}=R_3*x_3+R_3^1*x_4+w_3^{\′}---\mathrm{VIII}$

But same recursion obtains the expression formula of other y.

Because said structure, and R estimates the given value that obtains, in formula VII, and y ' ₄, R ₄Known, according to maximum likelihood decision rule:

${\hat{S}}_{\mathrm{ML}}=\arg \max \left\{\mathrm{\Λ}\left(\stackrel{\·\·}{S}\right)\right\}---\mathrm{IX}$

Wherein

$\mathrm{\Λ}\left(\stackrel{\·\·}{S}\right)=2*\mathrm{Re}({\stackrel{\·\·}{S}}^H*H^H*y)-{\stackrel{\·\·}{S}}^H*H^H*H*\stackrel{\·\·}{S}---X$

From fixed candidate's list entries, select with receiving sequence maximum likelihood is arranged sequence as court verdict, obtain x ₄Wherein

Be log-likelihood function log-likelihood function, LLF, It is the candidate sequence column vector.

The x that will adjudicate by formula VIII ₄, and known y ' ₃, R ₃, R ₃ ¹Substitution through maximum likelihood decision, obtains x ₃Herein owing to brought court verdict x into ₄, may introduce decision error, and to x ₃Judgement impact.Because later judgement all is to be based upon on the former court verdict, thereby can cause the accumulation of error.When the decreased performance that influence caused of this error under the condition that we allow, this algorithm is adaptable.So just solved the problem of big quantity symbol association.

At last, summary description essence of the present invention: the Maximum likelihood sequence detection algorithm of the simplification that the present invention proposes, carrying out QR by the associating matrix with precoding and frequency domain response decomposes, will detect vector is divided into the less several vectors of dimension and makes the maximum likelihood iterative detection respectively, with a little performance is the complexity that cost greatly reduces Maximum likelihood sequence detection, thereby very high using value is arranged.

Claims (10)

1. a MLSE simplification detection method that decomposes based on piecemeal QR is applied in the single-carrier system baseband transmission, is detected according to the subscriber signal Y of channel matrix H to its extraction by receiving terminal, it is characterized in that, may further comprise the steps:

1.1) piecemeal QR decomposes: described channel matrix H is decomposed into the product of an orthogonal matrix Q and a upper triangular matrix R, and is that unit is divided into the piecemeal upper triangular matrix by row and column with certain-length CL this upper triangular matrix R; With signal transformation column vector Y ', send column vector S again, white noise conversion column vector W ' is that unit is divided into the K section by the row correspondence with this length C L; Described transmission column vector S is the transmission symbol sebolic addressing vector of arbitrary possibility non-precoded, and described white noise conversion column vector W ' is by white noise vector W conversion, and described white noise vector W is the response of white noise AWGN at frequency domain, wherein, and Y '=Q ^-1* Y, W '=Q ^-1* W;

1.2) the maximum likelihood iterative detection: from the corresponding section of all described transmission column vector S of 1 section reverse traversal of K section to the, send corresponding section of sequence X according to the iterate conduct of therefrom selecting corresponding section maximum likelihood of maximum-likelihood criterion with subscriber signal conversion column vector Y ';

1.3) with each section of described transmission sequence X in order array output as final detection result.

2. according to the described MLSE simplification detection method of claim 1, it is characterized in that described length C L can select and set.

3. according to the described MLSE simplification detection method of claim 1, it is characterized in that, described receiving terminal also comprises interference signal restorer and arrester, described MLSE simplification detection method also comprises the signal that utilizes described restorer to recover the air interface form of described testing result correspondence, be considered as interference signal, eliminate by described arrester to other users.

4. according to the described MLSE simplification detection method of claim 1, it is characterized in that described single-carrier system is DFT-S-OFDM system or IFDMA system.

5. a MLSE who decomposes based on piecemeal QR simplifies checkout gear, be built in the single-carrier system in the receiving terminal, detect according to the subscriber signal Y of channel matrix H by receiving terminal its extraction, comprise channel estimator (32), it is characterized in that, also comprise the piece ML detector (31) that carries out Maximum likelihood sequence detection, be connected, carry out channel matrix decomposition module (35) that channel matrix QR decomposes and is connected, selects that partition length CL's cut apart length selection module (36) with described ML detector (31) with described ML detector (31);

Described channel matrix decomposition module (35) is used for described channel matrix H is decomposed into the product of an orthogonal matrix Q and a upper triangular matrix R;

Described ML detector (31), being used for described upper triangular matrix R is that unit is divided into the piecemeal upper triangular matrix by row and column with partition length CL, again with signal transformation column vector Y ', send column vector S, white noise conversion column vector W ' is that unit is divided into the K section by the row correspondence with described partition length CL, corresponding section from all described transmission column vector S of 1 section reverse traversal of K section to the, send corresponding section of sequence X according to the iterate conduct of therefrom selecting section corresponding maximum likelihood of maximum-likelihood criterion with subscriber signal conversion column vector Y ', with each section of described transmission sequence X in order array output as final detection result; Wherein, described transmission column vector S is the transmission symbol sebolic addressing vector of arbitrary possibility non-precoded, and described white noise conversion column vector W ' is by white noise vector W conversion, and described white noise vector W is the response of white noise AWGN at frequency domain, Y '=Q ^-1Y, W '=Q ^-1W.

6. simplify checkout gear according to the described MLSE of claim 5, it is characterized in that described channel matrix decomposition module (35) is positioned between described channel estimator (32) and the piece ML detector (31).

7. simplify checkout gear according to the described MLSE of claim 5, it is characterized in that described channel matrix decomposition module (35) is built in the described channel estimator (32).

8. simplify checkout gear according to the described MLSE of claim 5, it is characterized in that, describedly cut apart length to select module (36) be standalone module.

9. simplify checkout gear according to the described MLSE of claim 5, it is characterized in that, the described length of cutting apart selects module (36) to be built in the described ML detector (31).

10. simplify checkout gear according to the described MLSE of claim 5, it is characterized in that described single-carrier system is DFT-S-OFDM system or IFDMA system.

Images