CN100382460C - Method of receiving radio transmission by multiple antenna array - Google Patents

Abstract

The present invention provides a method for a multi-aerial array reception and radio transmission. A plurality of transmitted signals are detected in a multiple-input multiple-output communication system; the transmitted signals are transmitted respectively by transmitting aerial elements and go through one channel; in each iterative process, when the estimation of selected one of the transmitted signals is calculated, pre-matched filtering results of multicollection signals and one intermediate medium matrix Q are used; therefore, in each iterative process, the method of the present invention does not directly use the multicollection signals but uses the pre-matched filtering results of the multicollection signals; in each iterative process, the method of the present invention directly uses the pre-matched filtering results of the multicollection signals. The method of the present invention can save the amount of calculation for the transformation of the multicollection signals and effectively solves the problem that the computation complexity of the existing multiple-input multiple-output communication system is increased as the number of aerials is increased.

Description

Multi-antenna array receives the method for wireless transmission

Technical field

The present invention relates to the twireless radio-frequency communication system, relate in particular to the signal acceptance method that all uses the digit wireless communication system of multi-antenna array at transmitting terminal and receiving terminal.

Background technology

Prediction according to information theory, determine the factor of the bit rate that digit wireless communication system transmission data finally can reach to comprise: total transmitting power of transmitting terminal, the number of transmitting terminal and receiving terminal antenna element, bandwidth, receiving terminal noise power, and the characteristic of communication environments.

Most of traditional systems use single transmission antenna unit and single receive antenna unit.Yet industry has realized that at transmitting terminal, receiving terminal, and perhaps these two ends use multi-antenna array can improve bit rate greatly simultaneously.

U.S. Patent No. 6097771 has proposed a kind ofly to use the multiple-input, multiple-output with Space-Time framework (MIMO) wireless communication system of multi-antenna array simultaneously at transmitting terminal and receiving terminal, and the framework that this patent is described has demonstrated the Shannon capacity that can reach many reflectors multiple collector system in theory.The Shannon capacity of a system refers to the maximum size of the information theory permission of this system.

An other U.S. Patent No. 6317466 has also proposed a kind of multiple-input, multiple-output (MIMO) system, the technology that the framework of describing with U.S. Patent No. 6097771 provides is compared, the technology that the framework that this patent is described provides has significant lower computation complexity, yet it still can reach the quite most of of Shannon capacity simultaneously.

Especially, in the system of Miao Shuing, a data sequence is divided into M incoherent code element subsequence in the above.Each subsequence is by an emission of M transmitting antenna.M subsequence received by N reception antenna at receiving terminal after the influence of the channel that is H through a channel matrix.The signal of transmitting terminal M antenna emission is detected according to optimal ordering one by one at receiving terminal, and the influence of the signal that newly detects is eliminated from the input problem of remainder one by one, with the dimension of the input problem that reduces next step; Detect one by one in the process transmit, at each signal to be detected, ZF vector is found and be used for eliminating the influence of all residual interference signals, to obtain the estimation to this signal to be detected.

Though compare with the technology that the framework of U.S. Patent No. 6097771 descriptions provides, U.S. Patent No. 6, the technology that 317,466 frameworks of describing provide has significant lower computation complexity, but the still suitable height of computation complexity of the technology that the framework that the latter describes provides.

U.S. Patent No. 6600796 puts forward for the problem that solves computation complexity, the technology that this patent provides can reduce matrix inversion operation, so that finally for once ask matrix generalized inverse computing, thereby and previous method ratio, computation complexity can reduce by an order of magnitude.This mode is efficient and stable in number simultaneously, because it uses unitary transformation to avoid matrix square operation and matrix inversion.

The technology of front, particularly the reduction that the technology of U.S. Patent No. 6600796 is suitable the computation complexity of above-mentioned multiple-input, multiple-output (MIMO) system signal testing process, but when the number of antenna increased, the computation complexity of these technology increased significantly.That is to say, although suitable efficient when these technology are used for the antenna of proper number, these technology trouble more that becomes, especially the number when transmitting antenna become for a long time (such as, bigger than 10).So, the signal detection technique of above-mentioned multiple-input, multiple-output (MIMO) system that is sought after improving, the computation complexity of this technology is slower than previous technology with the speed that number of antennas increases.

In U.S. Patent application 20040013212, a kind of method and apparatus of signal detection technique of above-mentioned multiple-input, multiple-output (MIMO) system of improvement is provided, use multiple receive antenna to detect pilosity and penetrate signal.Especially, according to an embodiment of foregoing invention, iterative process is multiple of penetrating in the signal of decoding in each iteration, and each iteration uses an intermediary matrix with decision decoded that to be transmitted.For each of series of iterations, this intermediary matrix can calculate easily with recursive mode, this recursive mode is used a Schur complement operation, and the execution of this Schur complement operation is based on the inverse matrix of correction form of the intermediary matrix of last iteration.A Schur complement is a well-known matrix manipulation concerning one of skill in the art.

Clearer and more definite, U.S. Patent application 20040013212 provides a kind of a plurality of method and apparatus that transmit that detect in multiple-input, multiple-output (MIMO) communication system, and these transmit, and other transmit antenna elements is plain launches by dividing, and through a channel.This method comprises following step: (a) in above-mentioned communication system from dividing other reception antenna element to collect a plurality of received signals; (b) determine a channel matrix H, it is made up of the channel coefficients that estimates, and the estimation of these channel coefficients is based on above-mentioned many received signals; (c) calculate selecteed one estimation in above-mentioned the transmitting, thereby cause detection selecteed in above-mentioned the transmitting, above-mentioned estimation is based on above-mentioned many received signals and an intermediary matrix Q, and wherein above-mentioned intermediary matrix Q is the function of channel matrix H; (d) repeating step (c) at least once, perhaps repeat more number of times to detect another one or a plurality of above-mentioned transmitting, wherein above-mentioned intermediary matrix Q in step (c) use, for such repeating each time, in this time carrying out, recomputated, and this recomputates a function of an inverse matrix that is based on a Schur complement, and above-mentioned Schur complement is at an element, in the inverse matrix of the correction form of the intermediary matrix Q that this element uses in the last execution of step (c).

The computation complexity of the technology that U.S. Patent application 20040013212 is introduced can be slower than previous technology with the speed that number of antennas increases.Therefore, still have living space and can improve its efficient, make the computation complexity of improvement technology increase with number of antennas and the speed that increases further reduce.

Summary of the invention

The invention provides the method that a kind of multi-antenna array receives wireless transmission, can solve existing multiple-input, multiple-output (MIMO)

In the communication system technology along with number of antennas increases the problem that its computation complexity decreases.

For achieving the above object, multi-antenna array of the present invention receives the method for wireless transmission, detects a plurality of transmitting in multiple-input, multiple-output (MIMO) communication system in the system, and these transmit by dividing other transmit antenna elements plain emission, and through a channel, this method may further comprise the steps:

(a) in above-mentioned communication system from dividing other reception antenna element to collect a plurality of received signals;

(b) determine a channel matrix H, it is made up of the channel coefficients that estimates, and the estimation of these channel coefficients is based on above-mentioned a plurality of received signals;

(c) calculate selecteed one estimation in above-mentioned a plurality of the transmitting, thereby cause selecteed one detection in above-mentioned a plurality of the transmitting, above-mentioned estimation is based on pre-matching filtering result and intermediary matrix Q of above-mentioned a plurality of received signals; The number of the item that the pre-matching filtering result of wherein above-mentioned a plurality of received signals has is identical with the not detected as yet number that transmits, and the pre-matching filtering result's of a plurality of received signals is every, respectively item by item corresponding to not detected transmit every as yet; And above-mentioned intermediary matrix Q is the function of channel matrix H;

(d) repeating step (c) at least once, perhaps repeat more number of times to detect another one or a plurality of above-mentioned transmitting, wherein above-mentioned intermediary matrix Q in step (c) use, for such repeating each time, in this time carrying out, recomputated, and this recomputates a function of an inverse matrix that is based on a Schur complement, and above-mentioned Schur complement is at an element, in the inverse matrix of the correction form of the intermediary matrix Q that this element uses in the last execution of step (c).

In iterative process each time, in calculating above-mentioned transmitting during selecteed one estimation, method of the present invention is used pre-matching filtering result and intermediary matrix Q of many received signals, and the method for prior art is used many received signals and an intermediary matrix Q.The intermediary matrix Q that method of the present invention is used is identical with the intermediary matrix Q of the method use of prior art.Therefore, in iterative process each time, method of the present invention is not directly used many received signals, and is to use the pre-matching filtering result of many received signals.

In iterative process each time, method of the present invention is directly used the pre-matching filtering result of many received signals, and art methods need be carried out same conversion to obtain the conversion of many received signals to many received signals.So method of the present invention can be saved the amount of calculation of many received signals being carried out above-mentioned conversion, efficiently solve existing MIMO communication system along with number of antennas increases the problem that its computation complexity decreases.

Description of drawings

Fig. 1 is to use the schematic diagram of multiple-input, multiple-output (MIMO) wireless communication system of a plurality of transmission antenna unit and a plurality of reception antennas unit.

Fig. 2 is that former U.S. Patent application 200400l3212 is docile and obedient preface ZF (Nulling) and eliminate the flow chart (being the FIG.5 of former patent application) of (Cancellation) scheme one by one.

Fig. 3 is the another kind statement of scheme shown in Figure 2, form with matrix has rewritten the formula that is used for the current detected signal of estimation of scheme in U.S. Patent application 20040013212 documents, that explains among the order of this figure step and Fig. 2 is slightly different, concerning one of skill in the art, the principle of the step foundation of explaining among the step of this figure statement and Fig. 2 is identical; Be illustrated in the former invention with " # " among the figure and carry out, but will no longer need the step carried out in the present invention.

Fig. 4 is that the present invention is docile and obedient the preface embodiment flow chart of the scheme of ZF and elimination one by one.Compare the peculiar step of the present invention with former invention with " # " expression.

Fig. 5 is under the situation of different transmit/receive antenna number of unit, when the scheme of using U.S. Patent application 20040013212 respectively and the solution of the present invention, asking the required amount of calculation of each data sample is that the result of the digital experiment of floating-point operation number contrasts schematic diagram.

Embodiment

At first introduce the embodiment of U.S. Patent application 20040013212 methods, this part is to the translation of the original text of this patent application document, and for the ease of readers ' reading, the row in this branch sort code, figure number and the original text number, figure number all are consistent.

As shown in Figure 1, use the scheme schematic diagram of multi-antenna array for what U.S. Patent application 20040013212 proposed simultaneously at transmitting terminal and receiving terminal.This scheme is operated in " Rayleigh scattering environment ", that is to say, in this signal communication environments, what each element of channel matrix can be similar to regards that statistics independently as.

As shown in Figure 1, s emission signal s ₁(k), s ₂(k) ..., s _M(k) respectively by M different antenna element 13.1,13.2 ..., the 13.M emission.Corresponding received signal x ₁(k), x ₂(k) ..., x _N(k) respectively from N different antenna element 16.1,16.2 ..., 16.N receives.In this scheme, it is 2 that transmission antenna unit is counted M minimum, and reception antenna unit number N minimum be M.Channel matrix H is the matrix of a N * M, and i reception antenna of element representation of the capable j of i row and j transmitting antenna are by the coupling of transmission channel in the matrix.Received signal x ₁..., x _NProcessed to produce the  that transmits that recovers in digital signal processor ₁...,  _MAlso shown summation composition 21-1 among this figure, 21-2 ..., 21-N, the unavoidable noise signal w that their representatives comprise ₁(k), w ₂(k) ..., w _N(k), these noise signals join reception antenna 16.1,16.2 respectively ..., in the signal that 16.N receives.

[0019] special, channel matrix H comprises channel vector h _{: 1}To h _{: M}, represent that respectively channel is to each the influence in M the transmission signals.Clearer and more definite, channel vector h _{: 1}Comprise channel matrix entry h ₁₁To h _N1, expression divides other in reception antenna 16-1 to 16-N on each, and channel is to s emission signal s ₁(k) influence; Channel vector h ₂Comprise channel matrix entry h ₁₂To h _N2, expression divides other in reception antenna 16-1 to 16-N on each, and channel is to s emission signal s ₂(k) influence; Channel vector h _{: M}Comprise channel matrix entry h _1MTo h _NM, expression divides other in reception antenna 16-1 to 16-N on each, and channel is to s emission signal s _M(k) influence.

More formal being described below of [0020] this multiple-input, multiple-output (using multi-antenna array simultaneously) structure at transmitting terminal and receiving terminal.

At first, at receiving terminal,, can obtain at sampling time slot k:

$x\left(k\right)=\underset{m=1}{\overset{M}{\mathrm{\Σ}}}{h}_{:m}{s}_m\left(k\right)+w\left(k\right)$

$=\mathrm{Hs}\left(k\right)+w\left(k\right)$

k＝1，2，…，K，

Here $x\left(k\right)={[x_1\left(k\right),x_2\left(k\right),...,x_N\left(k\right)]}^T$

$={[h_{1:}^{H}s\left(k\right)+w_1\left(k\right),h_{2:}^{H}s\left(k\right)+w_2\left(k\right),...,h_{N:}^{H}s\left(k\right)+w_N\left(k\right)]}^T$

[0021] be that the N dimension receives vector,

$={[h}_{:1},h_{:2},...,h_{:M}]$

$=\left[\begin{array}{c}{h}_{1:}^H\\ h_{2:}^H\\ \·\\ \·\\ \·\\ h_{N:}^H\end{array}\right]$

[0022] is a N * M complex matrix, supposes that it is a constant in the period of K symbol.Vector h _N:And h _{: m}Length be respectively M and N,

s(k)＝[s ₁(k)，s ₂(k)，...，s _M(k)] ^T

[0023] be M-dimension emission vector,

w(k)＝[w ₁(k)，w ₂(k)，...，w _N(k)] ^T

[0024] is additive white Gaussian noise (AWGN) vector of a zero-mean plural number, its variance

$R_{\mathrm{ww}}=E\left\{w\left(k\right)w^H\left(k\right)\right\}$

$={\mathrm{\σ}}_w^2{I}_{N\×N}$

[0025] and ^TWith ^HThe transposition and the conjugate transpose that divide other representing matrix or vector, and I here _{N * N}Expression N * N unit matrix.Suppose that here additive noise w (k) is to add up independently in time-domain and spatial domain.

In this system, the signal of a plurality of antennas of transmitting terminal emission is detected according to optimal ordering one by one at receiving terminal, and the influence of the signal that newly detects is eliminated from the input problem of remainder one by one, with the dimension of the input problem that reduces next step; Detect one by one in the process transmit, at each signal to be detected, ZF vector is found and be used for eliminating the influence of all residual interference signals, to obtain the estimation to this signal to be detected.

[0093] as shown in Figure 2, be docile and obedient the preface flow chart of the scheme of ZF and elimination one by one for former U.S. Patent application 20040013212.According to this patent application, more convenient with direct method with round-about way compute matrix G ratio.Especially, the least mean-square error of signal (MMSE) is estimated [1]:

$\hat{s}={(H^H\·H+{\mathrm{\αI}}_{M\×M})}^{-1}{H}^{H}x,$ Wherein α is based on the signal to noise ratio that transmits and predetermined constant.

Allow

G＝(H ^H·H+αI _M×M) ^-1H ^H

The G here can be known as least mean-square error and detect matrix, so

$\hat{s}=\mathrm{Gx}$

Definition

G＝R ^-1H ^H (21)

[0094] here

R＝(H ^H-H+αI _MxM) (22)

[0095] association's contrast matrix of error signal e (k)=s (k)-y (k) is:

R _ee＝E{e(k)e ^H(k)}

＝σ _w ²R ^-1

＝σ _w ²Q (23)

[0096] obvious, the highest signal to noise ratio (snr) element that has of y (k) is the element with minimum mean square error, so have:

${\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">p</figure-callout>}}_1=\underset{m}{\arg \min }{q}_{\mathrm{mm}}---\left(24\right)$

[0097] q here _MmMatrix Q=R ^-1Diagonal element.

[0098] matrix R can rewrite as follows:

$R=\underset{n=1}{\overset{l}{\mathrm{\&Sigma;}}}{h_{n:}}^H\&CenterDot;h_{n:}+{\mathrm{\&alpha;I}}_{M\&times;M}---\left(25\right)$

[0099] this means that R can pass through the calculating of N iterative recursive easily, as follows:

$R=\underset{n=1}{\overset{l}{\mathrm{\&Sigma;}}}{h_{n:}}^H\&CenterDot;h_{n:}+{\mathrm{\&alpha;I}}_{M\&times;M}$

$=R_{[l-1]}+h_{l:}{h_{l:}}^H---\left(26\right)$

[0100] and

R _[N]＝R，R _[0]＝α·I _M×M (27)

[0101] (being also referred to as " the second lemma inversion ", is the well-known mathematic(al) manipulation in this area with the Sherman-Morrison formula.), the calculating that Q also can recurrence, as follows:

$Q_{\left[l\right]}=Q_{[l-1]}-\frac{Q_{[l-1]}{h}_{l:}{h_{l:}}^H{Q}_{[l-1]}}{1+{h_{l:}}^H{Q}_{[l-1]}{h}_{l:}}---\left(28\right)$

[0102] initialization

Q _[0]＝(1/α)·I _M×M

[0103] obtains

Q _[N]＝[H ^H·H+αI _M×M] ^-1

[0106] well-known, any recursive calculation has been introduced possible numercal instability, because the limited accuracy of processor unit.But this instability is only a very just generation after the iteration of big figure.Because now in this case, the number of iteration that calculates Q is by the restriction of reception antenna number N, and so numercal instability unlikely takes place.Under any circumstance, numercal stable can the improvement easily is by increasing the value of α initialized the time.

[0107] notices that also in the iteration first time, formula (28) only is to calculate once easily according to a schematically realization of the invention of U.S. Patent application 20040013212 described herein.In case Q _[N]Calculated p ₁Can be easy to determine by top formula (24).

[0108] present, continue the schematic process of iteration for the first time, the estimation of input signal can be calculated as follows:

$y_{p_1}\left(k\right)=\underset{i=1}{\overset{\mathrm{<figure-callout\; id="1"\; label="M\; q\; p"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">M</figure-callout>}}{\mathrm{\&Sigma;}}}{q}_{p_{}}{{}_1,m{,l}_{m+1i}}_{}{h}_{:m}^{H}x\left(k\right)---\left(29\right)$

[0109] and

${\hat{s}}_{p_1}\left(k\right)=Q\left[y_{p_1}\left(k\right)\right]---\left(30\right)$

[0110] note decoding step final step according to the invention of U.S. Patent application 20040013212 be with U.S. Patent application 20040013212 in the final step (the 3rd step, [0070] OK) of the method introduced identical.

[0111] to the iteration below each (iteration is later for the first time), process is as follows: at first, note matrix Q dwindling by recurrence easily.For matrix Q easily by the method for dwindling of recurrence, [0111] row that the derivation of detailed principle according to Schur complement operation please refer to U.S. Patent application 20040013212 files has mainly only been listed the result of derivation herein to [0123] OK.

[0112] corresponding to element y with minimum variance _P1(k) p ₁Value determined after, can exchange the p of s emission signal s (k) easily ₁With the M item, so that M item signal is called current best estimation.Certainly, corresponding to the schematic realization of former invention, after such rearrangement, the index that transmits can be followed the trail of easily.Accordingly, the p of channel matrix H ₁Be listed as with M and exchanged, thus matrix R that derives by channel matrix H as defined above and the p of Q ₁Capable and the p with M ₁Also need to be exchanged with the M row.

Definition:

$R_{M-\mathrm{<figure-callout\; id="1"\; label="m\; +"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">m</figure-callout>}+1}=\left[\begin{array}{cc}{R}_{M-m}& v_{M-m}\\ v_{M-m}^H& {\mathrm{\&beta;}}_{p_m}\end{array}\right]$

$Q_{M-\mathrm{<figure-callout\; id="1"\; label="m\; +"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">m</figure-callout>}+1}=\left[\begin{array}{cc}{T}_{M-m}^{-1}& w_{M-m}\\ w_{M-m}^H& {\mathrm{\&eta;}}_m\end{array}\right]$

[0120] uses well-known Sherman-Morrison formula, obtain

$Q_{M-m}=T_{M-m}^{-1}-\frac{T_{M-m}^{-1}{v}_{M-m}{v}_{M-m}^H{T}_{M-m}^{-1}}{{\mathrm{\&beta;}}_{p_m}+v_{M-m}^H{T}_{M-m}^{-1}{v}_{M-m}}$

[0121] obvious, formula (38) display matrix Q is dwindling by recurrence easily.Especially, under general situation:

$Q_{M-m}=T_{M-m}^{-1}-\frac{T_{M-m}^{-1}{v}_{M-m}{v}_{M-m}^H{T}_{M-m}^{-1}}{{\mathrm{\&beta;}}_{p_m}+v_{M-m}^H{T}_{M-m}^{-1}{v}_{M-m}}---\left(39\right)$

$R_{M-m}=(H_{M-m}^H\&CenterDot;H_{M-m}+{\mathrm{\&alpha;I}}_{(M-m)\&times;(M-m)})---\left(40\right)$

[0122] notes R _M-mCan be obtained easily, do not needed direct calculating.Especially, according to the schematic realization of former invention, by deletion R _M+1-mLast column and last row, just can obtain R _M-m: R only _M=R was directly calculated during the iteration in the first time of the program of signal.

[0123] so, as shown in Figure 2, comprise following step according to multiple-input, multiple-output (MIMO) decoding technique of former invention:

[0124] initialization step (block diagram 50):

[0125] decision received signal x (k), and be that iteration is provided with x for the first time ₁(k)=x (k); With the initial matrix H of training sequence decision.

[0126] step 1 (block diagram 51):

[0127] the calculating R=R of recurrence _[N]And Q=Q _[N]--formula s (the 26) ﹠amp above using; (28)

[0128] step 2 (block diagram 52):

[0129] element with highest signal to noise ratio (SNR) of y (k) is detected.This element is related with the diagonal angle item of the minimum of the matrix Q of least mean-square error (MMSE) filter correspondence.If such row are p _m, the so selecteed estimation that transmits:

${\hat{s}}_{p_m}\left(k\right)=Q\left[y_{p_m}\left(k\right)\right]$

[0130] step 3 (block diagram 53):

[0131] supposes ${\hat{s}}_{p_m}\left(k\right)=s_{p_m}\left(k\right),$ S so _Pm(k) by from receiving vector x _m(k) deletion in, the result produces an amended reception vector, x _M+1(k).

[0132] step 4 (block diagram 54):

[0133] supposes l _mBe the selecteed index that transmits, at matrix R _M+1-mAnd Q _M+1-mMiddle exchange l _mAnd M-m+1 (last) row and column.Here and the part below this document, exchange x and y row and column in a matrix, the meaning is meant that in this matrix exchange x is capable and y is capable, and exchanges the x row and y is listed as.

[0134] step 5 (block diagram 55):

[0135] cuts apart matrix R _M+1-mDecision R _M-m, v _M-m, and β _Pm, and deletion Q _M+1-mLast delegation and one row with the decision TM _M-m ^-1

[0136] step 6 (block diagram 56):

[0137] calculates Q with formula (39) _M-m(recurrence).

[0138] step 7 (block diagram 57):

[0139] all M transmit decoded before, by successively at revised reception vector series x ₂(k) ..., x _M(k) enterprising line operate, repeating step 2-6 is to calculate p ₂..., p _M

[0140] according to an embodiment of former invention, multiple-input, multiple-output (MIMO) decoding technique is expressed as follows below:

[0141] initialization and iteration for the first time:

[0142]x ₁(k)＝x(k)

[0143] the calculating R=R of recurrence _[N]And Q=Q _[N], see formula (26) ﹠amp; (28)

[0144]f(k)＝[1，2，…，M] ^T

[0145] ${\mathrm{<figure-callout\; id="1"\; label="l"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">l</figure-callout>}}_1=\underset{i}{\arg \min }{q}_{M,\mathrm{ii}},{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">p</figure-callout>}}_1=f_{l_1}\left(k\right)$

$y_{p_1}\left(k\right)=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{q}_{M{,l}_{1}i}{h}_{:i}^H{x}_1\left(k\right)$

[0146] the item l of exchange vector f (k) ₁And M.

[0147] ${\hat{s}}_{p_1}\left(k\right)=Q\left[y_{p_1}\left(k\right)\right]$

[0148] recurrence (later iteration) is worked as m=1, and 2 ..., M-1:

[0149](a) $x_{m+1}\left(k\right)=x_m\left(k\right)-{\hat{s}}_{p_m}{h}_{p_m}$

[0150] (b) at matrix R _M+1-mMiddle exchange l _mAnd M-m+1 (last) row and column.

[0151] (c) at matrix Q _M+1-mMiddle exchange l _mAnd M-m+1 (last) row and column.

[0152] (d) as implied above, from R _M+1-mDecision R _M-m, v _M-m, and β _Pm

[0153] (e) by Q _M+1-mDecision T _M-m ^-1,

[0154] by deletion Q _M+1-mLast column and one row.

[0155] (f) as shown above, the calculating Q of recurrence _M-m, see formula (39)

[0156](g) $l_{m+1}=\underset{i}{\arg \min }{q}_{M-m,\mathrm{ii}},$ p _m+1＝f _lm+1(k)

(h) $y_{p_{m+1}}\left(k\right)=\underset{i=1}{\overset{M-m}{\mathrm{\&Sigma;}}}{q}_{M-m,l_{m+1i}}{h}_{{:f}_i\left(k\right)}^H{x}_{m+1}\left(k\right)$

[0157] (i) exchange vector f (k) the item l _M+1With item M-m.

[0158](j) ${\hat{s}}_{p_{m+1}}\left(k\right)=Q\left[y_{p_{m+1}}\left(k\right)\right]$

That [0159] tries to achieve separates:

[0160] estimation that transmits: ${[{\hat{s}}_{{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">p</figure-callout>}}_1},{\hat{s}}_{{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C20041005482700281.png"\; state="\{\{state\}\}">p</figure-callout>}}_2},\&CenterDot;\&CenterDot;\&CenterDot;,{\hat{s}}_{p_M}]}^T$

[0161] decoding order: f (k)=[p] _M..., p ₂, p ₁] ^T

Before introducing method of the present invention, introduce the notion that the present invention uses earlier, i.e. the pre-matching filtering of a plurality of received signals (Pre-match filtering) result.Corresponding to all m, m≤M not detected as yet transmitting, a plurality of received signals (are x ₁(k), x ₂(k) ..., x _N(k)) pre-matching filtering result is defined as follows:

$z_j\left(k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{h}_{\mathrm{ij}}^{*}{x}_i\left(k\right),$ J=1 wherein, 2 ..., m

In the top formula, ^*Conjugation is asked in expression.

Also be without loss of generality simultaneously for simple, suppose j here, j=1,2 ..., m transmits, from j transmission antenna unit emission.

Above formulate any one in transmitting for above-mentioned m, such as, j transmits, with the coupling coefficient h of the transmission antenna unit at i reception antenna unit and this place that transmits _IjConjugation, with the received signal x on same i.e. i reception antenna _i(k) multiply each other and obtain a product.For all N reception antenna, the operation above repeating obtains N product, again this N product addition, has just obtained the pre-matching filtering result corresponding to above-mentioned j a plurality of received signals that transmit.

Respectively the value of j is changed to 1,2 ..., m, and reuse method that the preceding paragraph describes to obtain the m item, they correspond respectively to each not detected as yet transmitting, the pre-matching filtering result of a plurality of received signals.

All m items with the pre-matching filtering result of the above-mentioned a plurality of received signals of vector representation claim this vector to represent vector for the pre-matching filtering result of a plurality of received signals.

During initialization, thereby promptly all M transmits all not detected when m=M, the pre-matching filtering result of above-mentioned a plurality of received signals represents all ordering of M item in this vector of vector, according to definition of the present invention, be according to M corresponding item by item respectively ordering that transmits in emission signal vector of this M item.When finding out initialization easily, the pre-matching filtering result of above-mentioned a plurality of received signals represents that vector is the product of the conjugate transpose of received signal vector and corresponding channel matrix H.Promptly initialized

z ₁＝H ^H·x

When m＜M, promptly transmitted when detected, the pre-matching filtering result of a plurality of received signals represents that vector can represent vectorial z by the pre-matching filtering result of initialized a plurality of received signals ₁Try to achieve.Provide in the paragraph that concrete method will be hereinafter be introduced the present invention.

In former invention, signal y _Pm+1(k) and x _M+1(k) divide other to be updated ([0156] row and [0149] is OK) in step (h) with (a), as follows

$y_{p_{m+1}}\left(k\right)=\underset{i=1}{\overset{M-m}{\mathrm{\&Sigma;}}}{q}_{M-m,l_{m+1}i}{h}_{f_s\left(k\right)}^H{x}_{m+1}\left(\text{k}\right)---\left(41\right)$

$x_{m+1}\left(k\right)=x_m\left(k\right)-{\hat{s}}_{p_m}\left(k\right){\&CenterDot;h}_{:p_m}---\left(42\right)$

But, shown in following method of the present invention, in conjunction with (41) and (42), y _Pm+1(k) and x _M+1(k) can more effectively be upgraded.

Here, the form rewriting formula (41) with matrix is more suitable.

At first definition correspondence is the channel matrix H of iteration each time _M-m+1, m=1,2 ..., M, when be m=1 in the iteration first time, H _M=H exchanges initial channel matrix H then _ML ₁Row (the pairing row of detected signal) and last M row are deleted last M then and are listed as to obtain being used for the channel matrix H of dwindling of next iteration _M-1In the iteration each time afterwards, all exchange the channel matrix H of dwindling _M+1-mL _mRow (the pairing row of detected signal) and last (M+1-m) row are deleted last (M+1-m) row then to obtain being used for the channel matrix H of dwindling of next iteration _M-mSimultaneously, exchange Q _M-mL _mRow and column and last (M-m) row and column are to obtain new Q _M-mFormula (41) can be rewritten as:

$y_{p_{m+1}}=q_{M-m}^H{\&CenterDot;H}_{M-m}^H{\&CenterDot;x}_{m+1}---\left(43\right)$

Here q _M-mExpression Q _M-m(M-m) row.

According to former U.S. Patent application 20040013212, sum up multiple-input, multiple-output (MIMO) decoding technique and be expressed as follows, and show corresponding flow chart below at Fig. 3.In the statement below, all are carried out in former invention, but will no longer need the step carried out in the present invention, mark with " # " in Fig. 3 accordingly.

In statement of the present invention, slightly different with the order of the step of former U.S. Patent application 20040013212 statements, concerning one of skill in the art, be easy to find out with the principle of the step foundation of former U.S. Patent application 20040013212 statements identical.In addition, introduced the formula that rewrites with the form of matrix above-mentioned herein.

Initialization step:

(i) (block diagram 60) decision received signal x (k); With the initial matrix of training sequence decision.

(ii) (block diagram 61) is that iteration is provided with x for the first time ₁(k)=x (k).

The (iii) calculating R=R of (block diagram 62) recurrence _[N]And Q=Q _[N], see formula (26) ﹠amp; (28).

(iv) (block diagram 63) defines initialized channel matrix H _M=H.

(v)f(k)＝[1，2，…，M] ^T

Recurrence (later iteration) is worked as m=1, and 2 ..., M-1:

(M_a) (block diagram 64) finds $l_m=\underset{i}{\arg \min }{q}_{M+1-m,\mathrm{ii}},$ Here q _{M+1-m, ii}Be Q _M+1-mDiagonal element.

(M_b) (block diagram 65) is then at matrix R _M+1-mMiddle exchange l _mWith M-m+1 (last) row and column, and at matrix Q _M+1-mMiddle exchange l _mAnd M-m+1 (last) row and column.Exchange l in vector f (k) _mItem and M+1-m item.

(M_c) (block diagram 66) is corresponding, in the original size or the channel matrix H of dwindling _M+1-mMiddle exchange l _mRow (the pairing row of detected signal) and last (M+1-m) row.

(M_d) (block diagram 67) p _m=f _M-m+1(k), p _mThe estimation of individual signal is $y_{p_m}=q_{M+1-m}^H\&CenterDot;H_{M+1-m}^H\&CenterDot;x_m,$ Here q _M+1-mExpression Q _M+1-mLast (M+1-m) row (block diagram 67); Detect signal ${\hat{s}}_{p_m}\left(k\right)=Q\left[y_{p_m}\left(k\right)\right]$

(M_e) (block diagram 68) $x_{m+1}\left(k\right)=x_m\left(k\right)-{\hat{s}}_{p_m}{H}_{(:,M+1-m)}$ (H here _{(:, M+1-m)}Be H _M+1-mLast M+1-m row).

(M_f) (block diagram 69) is corresponding, by deletion H _M+1-mLast M+1-m row are by H _M+1-mDecision H _M-m

(M_g) (block diagram 70) is as implied above, from R _M+1-mDecision R _M-m, v _M-m, and β _PmBy Q _M+1-mDecision TM _M-m ^-1, by deletion Q _M+1-mLast column and one row.

(M_h) (block diagram 71) is as shown above, the calculating Q of recurrence _M-m, see formula (39).

Iterative step when m=M:

Have only top step (M_a)-(M_d) to be performed to obtain this moment ${\hat{s}}_{p_M}\left(k\right)=Q\left[y_{p_M}\left(k\right)\right].$

That tries to achieve separates:

The estimation that transmits: ${[{\hat{s}}_{{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">p</figure-callout>}}_1},{\hat{s}}_{{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C20041005482700281.png"\; state="\{\{state\}\}">p</figure-callout>}}_2},\&CenterDot;\&CenterDot;\&CenterDot;{,\hat{s}}_{p_M}]}^T$

Decoding order: f (k)=[p _M..., p ₂, p ₁] ^T(from last detected signal to first detected signal).

Continue to introduce principle of the present invention below.Allow

$z_{m+1}\&equiv;H_{M-m}^H\&CenterDot;x_{m+1},---\left(44\right)$

By above as can be known to the pre-matching filtering result's of a plurality of received signals definition, the z in the following formula _M+1Represent with vector form, corresponding to all (M-m) individual not detected as yet transmitting, the pre-matching filtering result of a plurality of received signals.(44) substitution (43), can obtain:

$y_{p_{m+1}}=q_{M-m}^H\&CenterDot;z_{m+1}---\left(45\right)$

Can find out by (45), in formula (43) and (42), can use z _M+1Replace x _M+1Upgrade y _Pm+1But, if directly use (44) and (42) to upgrade z _M+1, computation complexity will remain unchanged, thereby will be identical with the computation complexity of (a) (going [0156] and row [0149]) with step (h).In fact, can directly not calculate (44) and upgrade z _M+1Because work as in conjunction with (43) (being the another kind of expression-form of formula (41)) and (42), some available intermediate object programs are such as vectorial H _M-m+1 ^HX _m, z just _m, and matrix Ф _M-m+1Can be repeated to utilize, wherein, for m=1,2 ..., M, above-mentioned matrix Ф _M-m+1By following formula definition: ${\mathrm{\&Phi;}}_{M-\mathrm{<figure-callout\; id="1"\; label="m\; +"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">m</figure-callout>}+1}\&equiv;H_{M-m+1}^H\&CenterDot;H_{M-\mathrm{<figure-callout\; id="1"\; label="m\; +"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">m</figure-callout>}+1}.$

(42) substitution (44) (with in conjunction with (43) and (42)), can obtain:

$z_{m+1}=H_{M-m}^H\&CenterDot;(x_m-{\hat{s}}_{p_m}\&CenterDot;h_{:p_m})---\left(46\right)$

$=H_{M-m}^H\&CenterDot;x_m-{\hat{s}}_{p_m}\&CenterDot;\left(H_{M-m}^H{\&CenterDot;h}_{:p_m}\right)$

(46) first of formula the right is one (M-m) long column vector.Confirm that easily this column vector is (M-m+1) long line vector z _mHead (M-m) OK, and the z here _mBe to former z _mBe the z that obtains after the sequence transformation operation identical with the step (i) of former invention ([0157] OK) _mIn other words, first on the right of (46) formula can be by z _mUpgrade, by to z _mCarry out sequence transformation and delete its last column then.

(46) second of formula the right also can be used the result of previous recurrence to obtain.But before so doing, need to introduce how to obtain Ф _M-m+1And without any more calculating.In fact, work as m=1, Ф _M-m+1=H ^HH, and it has been calculated when R ([0143] OK) is calculated in initialization.As m＞1, Ф _M-m+1Be matrix Ф _M-m+2Head (M-m+1) row and column, the Ф here _M-m+2Be to former Ф _M-m+2Be the Ф that obtains after the sequence transformation operation identical with the step (c) of former invention ([0151] OK) _M-m+2In other words, upgrade matrix Ф _M-m+1Can be by the Ф after the deletion sequence transformation _M-m+2Last row and delegation.

Ф has been arranged _M-m+1, can upgrade second (46) formula the right second.Because h:p _mIt is matrix H _M-m+1(at exchange l _M-1After row and (M-m+1) row) last row, verify easily H at second _M-m ^HH _{: pm}Be matrix Ф _M-m+1(M-m+1) row head (M-m) OK.In other words, can be by duplicating matrix Ф _M-m+1In element and obtain H _M-m ^HH _{: pm}

So, upgrade z by (46) _M+1The computation complexity that causes and in the step (a) of former invention ([0149] OK), upgrade x _M+1What cause is identical.In other words, in (43),, calculate H by reusing intermediate object program _M-m ^HX _M+1Cost can be saved, this cost is equivalent to (1/2) M ²N complex multiplication and addition.

Should be noted that and initialized the time, calculate z ₁=H ^HX ₁Can not avoid.z ₁Calculating need MN complex multiplication and addition.So, calculating y during corresponding to step (h) of former invention (row [0156]) and initialization _P1([0145] OK), the saving of total is (1/2) M ²N-MN complex multiplication and addition.

Below explain for embodiments of the invention.

As shown in Figure 4, be that the present invention is docile and obedient the preface flow chart of the scheme of ZF and elimination one by one.Among this figure, compare the peculiar step of the present invention with former invention, mark with " # " without exception.

Initialization step is as follows:

(i) (block diagram 80) decision received signal x (k); (Training Sequence) determines initial channel matrix H with training sequence, and it is made up of the channel coefficients that estimates, and the estimation of these channel coefficients is based on above-mentioned a plurality of received signals.This is a known technology, does not repeat them here.

(ii) (block diagram 81) is provided with x ₁(k)=x (k).

(iii) (block diagram 82) represents the initial value z of vector for the pre-matching filtering result of a plurality of received signals of the iterative computation first time ₁=H ^HX ₁

The (iv) calculating R=R of (block diagram 83) recurrence _[N]And Q=Q _[N], see formula (26) ﹠amp; (28).

(v) R=(H is calculated in (block diagram 84) initialization in the above ^HH+ α I _{M * M}) process in, obtain Ф in passing _M=H ^HH.

(vi)f(k)＝[1，2，…，M] ^T

Below recursive procedure (iteration) is described, work as m=1,2 ..., M-1,

(N_a) (block diagram 85) finds $l_m=\underset{i}{\arg \min }{q}_{M+1-m,\mathrm{ii}},$ Here q _{M+1-m, ii}Be Q _M+1-mDiagonal element.This step is selected to have transmitting of highest signal to noise ratio and is detected in this iteration according to the not detected as yet signal to noise ratio that transmits.The covariance matrix that can be derived the not detected as yet error signal e that transmits (k)=s (k)-y (k) by above-mentioned formula (23) is: $R_{\mathrm{ee}}=E\left\{e\left(k\right)e^H\left(k\right)\right\}={{\mathrm{\&sigma;}}_w}^2{R}_{M+1-m}^{-1}$ $={{\mathrm{\&sigma;}}_w}^2{Q}_{M+1-m}.$ Well known to those skilled in the art is above-mentioned matrix R _EeEach diagonal entry, be respectively the mean square error of the respective element of y (k); The element with highest signal to noise ratio (SNR) of y (k) has the element of least mean-square error exactly.So Q _M+1-mDiagonal element in have of minimum value, just corresponding to the element of y (k), promptly corresponding to not detected as yet transmitting with highest signal to noise ratio with highest signal to noise ratio (SNR).

(N_b) (block diagram 86) is then at matrix R _M+1-mMiddle exchange l _mWith M-m+1 (last) row and column, and at matrix Q _M+1-mMiddle exchange l _mAnd M-m+1 (last) row and column.Exchange l in vector f (k) _mItem and (M+1-m).

(N_c) (block diagram 87) is at matrix Ф _M+1-mMiddle exchange l _mAnd M-m+1 (last) row and column.

(N_d) (block diagram 88) represents vectorial z in the pre-matching filtering result of a plurality of received signals _mMiddle exchange l _mItem and last M+1-m item.

(N_e) (block diagram 8) p _m=f _M-m+1(k), p _mThe estimation of individual signal is $y_{p_m}=q_{M+1-m}^H{\&CenterDot;z}_m.$ Here q _M+1-mExpression Q _M+1-mM+1-m (at last) row.Detect signal ${\hat{s}}_{p_m}\left(k\right)=Q\left[y_{p_m}\left(k\right)\right]$

(N_f) the pre-matching filtering result of a plurality of received signals of (block diagram 90) deletion (M-m+1) length represents vectorial z _mLast obtain (M-m) long line vector (z _m) ^{Min us}

(N_g) (block diagram 91) Here (M-m) long line vector  _M-m+1Be matrix Ф _M-m+1(M-m+1) row head (M-m) OK.Verify z easily _M+1Corresponding to all (M-m) individual not detected as yet transmitting, the pre-matching filtering result of a plurality of received signals represents vector.

(N_h) (block diagram 92) deletion matrix Ф _M-m+1Last row and delegation, to obtain Ф _M-m

(N_i) (block diagram 93) is as implied above, from R _M+1-mDecision R _M-m, v _M-m, and β _PmBy Q _M+1-mDecision T _M-m ^-1, by deletion Q _M+1-mLast column and one row.

(N_j) (block diagram 94) is as shown above, the calculating Q of recurrence _M-m--see formula (39)

Iterative step when m=M:

Have only top step (N_a)-(N_e) to be performed to obtain this moment ${\hat{s}}_{p_M}\left(k\right)=Q\left[y_{p_M}\left(k\right)\right].$

That tries to achieve separates:

The estimation that transmits: ${[{\hat{s}}_{{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C20041005482700261.png,C20041005482700271.png"\; state="\{\{state\}\}">p</figure-callout>}}_1},{\hat{s}}_{{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C20041005482700281.png"\; state="\{\{state\}\}">p</figure-callout>}}_2},\&CenterDot;\&CenterDot;\&CenterDot;{\widehat{,s}}_{p_M}]}^T$

Decoding order: f (k)=[p _M..., p ₂, p ₁] ^T(from last detected signal to first detected signal).In a word, method of the present invention needs (1/2) M ³+ 2M ²N+O (M ²+ MN) (same number! ) individual complex multiplication and complex addition, if M=N, method of the present invention needs (5/2) M ³+ O (M ²) individual complex multiplication and complex addition.Need (11/3) M according to having provided the invention of the former U.S. in the 51st phase " IEEE signal processing journal " (IEEE Transactions on Signal Processing) " A fastrecursive algorithm for optimum sequential signal detection in a BLAST system " of publishing in July, 2003 ³+ O (M ²) individual complex multiplication and 3M ³+ O (M ²) complex addition, in " IEEE signal processing journal " " Comments on ' A fast recursive algorithm for optimumsequential signal detection in a BLAST system " ' that the inventor publishes as first author in August, 2004, only need to point out 3M ³+ O (M ²) individual complex multiplication just can realize the method for former invention.So method of the present invention to the acceleration of the method for former invention, with the standard of needed complex multiplication and complex addition number, is respectively 3/ (5/2)=1.2 and 3/ (5/2)=1.2.

As shown in Figure 5, under the situation of different transmit/receive antenna number of unit, when the mode of using U.S. Patent application 20040013212 respectively and mode of the present invention, ask the result of the digital experiment of the required amount of calculation of each data sample to contrast signal.Asking the required amount of calculation of each data sample is to measure with the number of required floating-point operation, and the number of above-mentioned required floating-point operation can be tried to achieve by the means of emulation with the respective function among the matlab.The result that can see Fig. 5 is consistent with theoretical analysis.

Claims (5)

1. a multi-antenna array receives the method for wireless transmission, transmit in order to detect M in multiple-input-multiple-output communication system, these transmit, and other transmit antenna elements is plain launches by dividing, and through a channel, it is characterized in that this method may further comprise the steps:

(a) in above-mentioned communication system from dividing other reception antenna element to collect a plurality of received signal x, and determine a channel matrix H, it is made up of the channel coefficients that estimates, and the estimation of these channel coefficients is based on above-mentioned a plurality of received signals;

(b) the pre-matching filtering result of a plurality of received signals of calculating represents the initial value z of vectorial z ₁=H ^HX; The initial value of compute matrix R and intermediary matrix Q, above-mentioned matrix R and matrix Q initial value are respectively: R=(H ^HH+ α I _{M * M}) Q=R ^-1, definition matrix Ф initial value is Ф=H ^HH, the initial value of this Ф obtains in calculating the R process, wherein, I _{M * M}Expression M * M unit matrix, α is based on the signal to noise ratio that transmits and predetermined constant, the conjugate transpose of H representing matrix or vector;

(c) select one calculating its estimation from above-mentioned a plurality of transmitting, and then detect this selecteed transmitting, the pre-matching filtering result that above-mentioned estimation is based on the pre-matching filtering result composition of above-mentioned a plurality of received signals represents vectorial z _M+1With intermediary matrix Q; Wherein, z _M+1Be expressed as the expression vector z that a preceding iterative process is used _mExchange is detected transmit corresponding row and column and last row and column and delete the vector that last column obtains and got matrix Ф _M-m+1The vector that obtains of the capable estimation of multiply by the signal that is detected that calculates in the preceding iterative process of head (M-m) of (M-m+1) row poor, the pre-matching filtering result of these a plurality of received signals represents that the number of the item that vector has is identical with the not detected as yet number that transmits, and the pre-matching filtering result of a plurality of received signals represents the every of vector, respectively item by item corresponding to not detected transmit every as yet; And above-mentioned intermediary matrix Q is the function of channel matrix H;

(d) repeating step (c) at least once, perhaps repeat more number of times detecting another one or a plurality of above-mentioned transmitting, the wherein above-mentioned intermediary matrix Q that uses in step (c) is for such repeating each time, matrix Q is all recomputated, Q=Q _M-1, here

$Q_{m-1}=T_{m-1}^{-1}-\frac{T_{m-1}^{-1}{v}_{m-1}{v}_{m-1}^H{T}_{m-1}^{-1}}{{\mathrm{\&beta;}}_{p_{(M-m+1)}}+v_{m-1}^H{T}_{m-1}^{-1}{v}_{m-1}},$

Here

$T_{m-1}=R_{m-1}-v_{m-1}{v}_{m-1}^H/{\mathrm{\&beta;}}_{p_{(M-m+1)}}$

Be at Q _m ^-1In β _{P (M-m+1)}Schur mend, the conjugate transpose of H representing matrix or vector wherein, and the correction form of intermediary matrix Q here when in the directly corresponding each time execution before above-mentioned repeating that is used in step (c), are Q _m, here:

$Q_m={\left[\begin{array}{c}{R}_{m-1}{v}_{m-1}\\ v_{m-1}^H{\mathrm{\&beta;}}_{p_{(M-m+1)}}\end{array}\right]}^{-1},$ It is at an element that above-mentioned Schur mends, last element of last column in the inverse matrix of the intermediary matrix Q that this element uses in the last execution of step (c).

2. method according to claim 1, it is characterized in that, the step that further comprises once or more frequently the pre-matching filtering result of above-mentioned a plurality of received signals is revised, by the influence of from the pre-matching filtering result of above-mentioned a plurality of received signals, eliminating the signal that is detected to small part, the influence of the signal that above-mentioned elimination is detected is based on the estimation of the above-mentioned signal that is detected that calculates, and the pre-matching filtering result of the above-mentioned therein a plurality of received signals that are corrected is the repetitions that are used to step (c) subsequently.

3. method according to claim 2 is characterized in that, the pre-matching filtering result of these a plurality of received signals is revised and estimates that with its step that transmits more specifically is:

(1) first time of step (c) carry out or the repeating each time of subsequently step (c) in, before the pre-matching filtering result who uses matrix Ф to above-mentioned a plurality of received signals represents that vector is revised, exchange

(i) pre-matching filtering result represents of vector, it corresponding to above-mentioned by detected the transmitting of previous execution of step (c), with

(ii) pre-matching filtering result represents last of vector, it is corresponding to above-mentioned last that transmits, and this transmits and also is not detected;

(2) use the pre-matching filtering result of above-mentioned a plurality of received signals to represent vector, calculate selecteed one estimation in a plurality of the transmitting;

(3) the pre-matching filtering result of a plurality of received signals of deletion represents last of vector, it corresponding to above-mentioned by detected the transmitting of previous execution of step (c);

(4) use above-mentioned matrix Ф that the pre-matching filtering result of a plurality of received signals is represented vector correction.

4. method according to claim 3, it is characterized in that, step (c) carry out for the first time or the repeating each time of subsequently step (c) in, before the matrix Ф that states in the use does to revise to the pre-matching filtering result of a plurality of received signals, above-mentioned matrix Ф is done the exchange of row and column, i.e. the previous delegation that carries out above-mentioned last that transmits that the detected delegation that transmits and matrix Ф correspondence also be not detected in the corresponding step of switching matrix Ф (c); One row of above-mentioned last that transmits that detected row that transmit of previous execution and matrix Ф correspondence also are not detected in the corresponding step of matrix Ф (c); Simultaneously, carry out for the first time or during subsequently step (c) repeats each time in step (c), after stating matrix Ф in the use the pre-matching filtering result of a plurality of received signals being done to revise, corresponding above-mentioned steps (c) had before been carried out detected last column that transmits and last row among the deletion matrix Ф.

5. method according to claim 1, it is characterized in that, the correction form of intermediary matrix Q is to be derived out by intermediary matrix Q, before carried out the delegation of above-mentioned last that transmits that a detected delegation that transmits and matrix Q correspondence also be not detected by the corresponding step (c) of switching matrix Q respectively; And the corresponding step (c) of switching matrix Q had before been carried out row of above-mentioned last that transmits that detected row that transmit and matrix Q correspondence also be not detected.

Images